# l0g.fr · complete English corpus > Full text of the published English analyses and reference guides. Infographics are removed. CC BY 4.0, attribution l0g.fr. Not investment advice. Generated 2026-08-10T06:32:22.700Z. 50 guides, 142 analyses. Concise map: https://l0g.fr/llms.txt French corpus: https://l0g.fr/llms-full.txt Normative editorial protocol: https://l0g.fr/en/editorial-protocol/ Stable protocol release 1.0.0: https://github.com/bluetouff/l0g/tree/main/releases/l0g-editorial-protocol-1.0.0 Licenses: MIT for code, tests and schemas; CC BY 4.0 for editorial text, public data and evidence artifacts. Citation metadata: https://github.com/bluetouff/l0g/blob/main/CITATION.cff ============================================================================ REFERENCE GUIDE: How to Read Corporate Debt: Five Numbers, One Maturity Wall, One Trap URL: https://l0g.fr/en/guides/read-corporate-debt/ Canonical French source: https://l0g.fr/guides/lire-la-dette-d-une-entreprise/ Date: 2026-08-07 (reviewed 2026-08-07) ---------------------------------------------------------------------------- A company can report abundant cash and still be fragile. It can also carry a high debt load without facing an immediate problem. The difference is not one isolated ratio. It is the alignment between three realities: debt to be repaid, cash that is genuinely available and the timetable of maturities. This guide offers a reproducible method built from financial statements and their notes, not a highlighted investor-presentation figure. Debt is a timetable before it is an amount The balance sheet answers a static question: what amount is recorded at the reporting date? Credit risk is dynamic: what payments fall due, with which resources, and when? That is why a serious reading starts with the balance sheet, moves into the debt note and ends with the statement of cash flows. For U.S. registrants, the SEC rule is explicit: management’s discussion must analyse the ability to generate or obtain adequate cash for the next twelve months and beyond, and specify material cash requirements from known obligations. It must also distinguish internal from external liquidity sources. The current text of Item 303 is useful reading even for companies that do not report in the United States. Under IFRS, financial-instrument disclosures must enable users to evaluate risk exposures and how the entity manages them. IFRS 7 therefore includes liquidity risk in the analytical perimeter. The logic is universal: a debt total never stands on its own without payment dates and terms. The five numbers to record The first reading sheet fits into five lines. All should use the same reporting date and cite their page or note of origin. 1. Gross interest-bearing debt. Add current and non-current borrowings, bonds, term loans, drawn revolving facilities and, depending on the chosen presentation, finance-lease liabilities. Do not confuse this total with all liabilities: trade payables, taxes payable and provisions are not automatically financial debt. 2. Cash that can actually be mobilised. Start with cash and cash equivalents, then identify restricted or pledged cash, funds held in a subsidiary that are difficult to move, or cash required for normal operations. Reported cash is an accounting fact; the share available to repay debt is a judgement that must be explained. 3. Net debt and its reconciliation. As a first pass, gross debt less cash. But the measure is not standardised: some companies also deduct liquid investments, while others exclude part of cash or leases. Preserve the issuer’s definition and rebuild the number from the accounts. 4. Principal due within twelve months and the full maturity schedule. The first number shows urgency. The later schedule reveals a potential maturity wall and therefore reliance on refinancing. 5. Interest burden and cash generation. Compare interest paid or incurred with operating profit, clearly reconciled EBITDA and cash flows. None of these denominators is interchangeable. Under U.S. GAAP, the cash-flow statement can reconcile cash, cash equivalents and amounts described as restricted cash separately. FASB Update 2016-18 sets out that reconciliation. It does not answer the economic question on its own: how much cash can actually service debt? Gross debt and net debt: keep the definition in view A company may present “net debt” in an earnings release. The measure can be useful, but it has no single accounting definition. The SEC says non-GAAP measures, and ratios built on them, must be clearly described and reconciled to the most directly comparable accounting measure. A ratio can mislead if its construction is not visible or if the GAAP measure vanishes behind it. See the SEC’s non-GAAP interpretations, especially Questions 100.05 and 102.10. The working formula is straightforward: net debt = selected financial debt − selected cash Its simplicity does not remove the need to inspect both sides. Debt can include secured or unsecured borrowings, fixed- or floating-rate instruments, convertibles, revolver drawings or related-party debt. Cash can include restricted funds or investments that cannot be sold immediately. Two companies reporting the same net-debt number may therefore have very different risk profiles. The practical rule is to copy the issuer’s definition, then build a more conservative version when the notes reveal restricted cash or debt omitted from the promotional figure. The point is not to replace management’s metric arbitrarily, but to make the difference explicit. The maturity schedule often decides the issue Ten-year debt and debt due in six months have the same balance-sheet amount, but not the same refinancing risk. Convert the debt note into a calendar: current portion, year two, year three, then later years. Separate interest from principal. An undrawn credit facility can provide room to manoeuvre, but it is not cash. Check its maturity, amounts already used, drawing conditions, potential security and covenants. In its reporting manual, the SEC notes that debt instruments, guarantees and covenants may need discussion when they affect liquidity or the capacity to raise additional funding. The SEC guide also cautions against merely repeating the cash-flow statement. Refinancing is not an economic repayment. It replaces one maturity with new debt, a new rate, new security and sometimes new restrictions. Treat an assumed refinancing as an assumption, never an established fact. Interest is paid in cash, not in EBITDA EBITDA is widely used to measure leverage, especially in credit agreements. But it does not replace operating cash flow or cash actually available after investment, interest, tax and working-capital needs. Free cash flow has no uniform definition either. The SEC makes this point in Question 102.07: free cash flow is often presented as operating cash flow less capital expenditure, but the definition must be explained and reconciled to the accounts. Crucially, it should not imply cash that is automatically discretionary, because mandatory debt service may not be deducted. The primary source is here. Two ratios can complete the reading if their formula is stated: - Net debt / EBITDA: a leverage order of magnitude relative to operating earnings before interest, tax and non-cash charges. The denominator should be the published, reconciled and period-comparable version. - EBITDA / interest or EBIT / interest: an indication of interest-burden coverage. The two ratios are not equivalent, because EBITDA excludes depreciation and amortisation. Covenant calculations can include further adjustments. Ratios ask questions; they do not deliver automatic verdicts. A stable company with low investment needs and long fixed-rate debt cannot be read like a cyclical, capital-intensive company exposed to floating rates or dependent on one funding market. To set the cost of credit in market context, see our guide to credit spreads. Three areas the debt number misses Covenants A covenant may require the borrower to stay within a leverage, interest-coverage or liquidity threshold. Its contractual calculation is often more adjusted than the ratio shown to investors. Record the threshold, the actual level, the headroom and the consequence of a breach. The SEC specifically recommends this information when a material covenant is needed to understand financial condition or liquidity. Funding-like commitments Finance leases, guarantees, factoring, purchase commitments and supplier-finance arrangements are not all financial debt in the same accounting sense. They can still absorb cash or alter payment priority. Reverse factoring illustrates the point: the IFRS Interpretations Committee notes that it can concentrate obligations with one financial institution and create liquidity risk. Its June 2020 analysis explains why the nature, amount and timing of liabilities matter as much as their label. Ranking and security Two equally sized debts can have very different implications: senior secured debt, unsecured bonds, subordinated debt, shareholder debt, debt at a subsidiary or debt against a specific asset. Reading the debt note also means identifying who gets paid first and which assets are pledged. Credit ratings offer a complementary lens, but they do not replace contracts and financial-statement notes. A thirty-minute reading method 1. Open the annual report, the most recent quarterly report and, for a U.S. issuer, the EDGAR filing. 2. Record current and non-current debt, then find every component in the debt note. 3. Record cash, equivalents, investments and restricted cash. Note perimeter differences instead of smoothing them away. 4. Build the principal maturity schedule and isolate the next twelve and twenty-four months. 5. Read management’s liquidity section: operating cash flow, capital expenditure, funding sources, covenants, guarantees and announced refinancings. 6. Rebuild ratios from published figures, showing the formula and any adjustment. 7. Compare the last two reporting years. Stable debt can become riskier when maturities move closer, interest rises or cash generation weakens. Our guide to reading a 10-K shows where to find these elements in a U.S. filing: balance sheet, cash-flow statement, debt note and MD&A. For acquisition-heavy issuers, pair it with the guide to CLOs and leveraged loans. Limits This framework is a reading method, not investment advice or a credit decision. Definitions of net debt, adjusted EBITDA and free cash flow vary across issuers and sometimes across an issuer’s own contracts. Published schedules are contractual at the reporting date: they do not establish a future issuance, renegotiation or market access. Banks, insurers and real-estate companies require specific analytical frameworks because debt and liquidity sit at the centre of their business model. Sources - SEC, 17 CFR §229.303, Management’s discussion and analysis, accessed 7 August 2026: liquidity, capital resources and cash requirements over the next twelve months and beyond. - SEC, Non-GAAP Financial Measures Compliance and Disclosure Interpretations, last updated 13 December 2022, accessed 7 August 2026: reconciliation of non-GAAP measures, EBITDA, free cash flow and covenants. - SEC, Financial Reporting Manual, Topic 9, accessed 7 August 2026: liquidity analysis, debt, guarantees and covenants. - IFRS Foundation, IFRS 7 Financial Instruments: Disclosures, accessed 7 August 2026: disclosures on financial-risk exposure and its management. - IFRS Interpretations Committee, IFRIC Update, June 2020, accessed 7 August 2026: reverse factoring, presentation of liabilities and liquidity risk. - FASB, ASU 2016-18, Statement of Cash Flows: Restricted Cash, accessed 7 August 2026: reconciliation of cash, cash equivalents and restricted cash in the U.S. GAAP statement of cash flows. - SEC EDGAR, filing search, accessed 7 August 2026: access to original annual and quarterly filings by U.S. issuers. ============================================================================ REFERENCE GUIDE: How to Read the Balance of Payments and the Current Account URL: https://l0g.fr/en/guides/read-the-balance-of-payments/ Canonical French source: https://l0g.fr/guides/lire-la-balance-des-paiements/ Date: 2026-07-28 (reviewed 2026-07-28) ---------------------------------------------------------------------------- import InfographicBars from '../../components/infographic/InfographicBars.astro'; The balance of payments is one of the rare places in economics where an accounting identity cannot lie. It records everything a country exchanges with the rest of the world, goods, services, income and capital, and these flows balance by construction. Its lesson is brutal and often misunderstood: a country that buys more from the world than it sells must, in the same motion, borrow the difference from it. The trade deficit and the capital inflow are two sides of the same coin. Reading the balance of payments means grasping that mechanism, and what it says about how an economy is financed. This guide runs through it, with the US accounts of 2026 as the backdrop. The three accounts The balance of payments reads in three blocks. The current account first, the most followed: it records exchanges of goods and services, investment and labour income, and transfers. Its balance says whether a country lives above or below its means vis-à-vis the world. The capital account next, marginal, recording a few wealth transfers. The financial account finally, which records cross-border asset flows: purchases and sales of bonds, equities, real estate, direct investment. The key is that these three accounts sum to zero, up to measurement error. It is not an economic law, it is an accounting identity: every unit spent abroad must be financed by something. Understanding the balance of payments means first internalising that balancing constraint. The iron identity From that identity flows the most counterintuitive truth of the subject. A current account deficit is not a hole opening into a void: it is necessarily offset by a financial account surplus, that is, by net capital inflows. A country that imports more than it exports borrows the difference from the rest of the world, by selling it assets. Current account deficit and capital import are the same reality seen from both sides. That mechanism has a second face, on the savings side. The current account balance equals, by identity, the gap between national saving and national investment. A country that invests more than it saves must fill the gap with foreign saving, which shows up as a current account deficit. The United States is the canonical example: they consume and invest more than they produce and save, and the world lends them the difference by buying their Treasuries, a flow we track in our piece on the marginal buyer of US debt. The US case The 2026 figures give the identity flesh. The US current account deficit widened to $226.8 billion in the first quarter of 2026, or 2.9% of gross domestic product, from 2.8% the previous quarter. In mirror, the financial account recorded about $209 billion of net inflows, the exact reflection of that deficit: the world lent the United States roughly what they overspent. The external position and the exorbitant privilege Summed year after year, these deficits build a debt: the net international investment position, the difference between what residents hold abroad and what foreigners hold at home. For the United States, it reached minus $21.27 trillion at end-March 2026, a colossal net indebtedness to the world. For a long time, that figure worried less than it should have, thanks to the exorbitant privilege: the United States earned more on its foreign assets, often risky and profitable, than it paid on its debt, largely low-yielding Treasuries. The income balance stayed positive despite a massive debtor position. That privilege is eroding. Recent revisions suggest the net international investment position has deteriorated to the point where the return differential no longer offsets the gap in holdings, a structural shift in how US investment income evolves. In other words, America's external debt is starting to cost more than it earns, a turn to watch closely, tied to the rise in rates and the swelling of the debt we track elsewhere. Reading pitfalls A few reflexes avoid misreadings. The first, and most important, is that the accounting identity says nothing about causation: it establishes that current account deficit and capital inflows coincide, not which causes which. A deficit can reflect a hunger for consumption, but also the attractiveness of a country that draws in the world's capital; the causal arrow does not read off the balance. The second is to scale the balances to GDP rather than in dollars, to judge their sustainability. The third is to distinguish, within the current account, the trade balance from the income balance, which tell different stories. The fourth is that the net international investment position contains, like reserves, valuation effects: it moves with equity markets and the exchange rate, not only with flows. Reading the balance of payments in practice The method fits in a few moves. Start from the identity: a current account deficit is always a capital inflow, and vice versa, which steers the question towards financing. Read the current account as the savings-investment gap, to connect the external balance to the domestic macro. Track the net international investment position and the income balance to judge sustainability, keeping in mind the erosion of the exorbitant privilege. Finally, cross the balance with foreign-holdings data, like the Treasury's TIC figures and the offshore dollar plumbing we describe in our analysis of eurodollars. The balance of payments does not judge, it constrains: it says no country can durably spend without someone, somewhere, agreeing to finance it. --- Sources - Bureau of Economic Analysis, "U.S. International Transactions and Investment Position, 1st Quarter 2026" (current account deficit $226.8bn, 2.9% of GDP; financial account; net international investment position -$21.27tn) - Peterson Institute for International Economics, "Don't blame America's current account deficit on the dollar", 2026 (causation and reading the identity) - Haver Analytics, "Q1 Current Account: Modest Slippage in Early 2026" (quarterly dynamics of the US current account) ============================================================================ REFERENCE GUIDE: How to Read Sovereign CDS URL: https://l0g.fr/en/guides/read-sovereign-cds/ Canonical French source: https://l0g.fr/guides/lire-les-cds-souverains/ Date: 2026-07-28 (reviewed 2026-07-28) ---------------------------------------------------------------------------- import InfographicBars from '../../components/infographic/InfographicBars.astro'; A bond spread tells how much a state pays to borrow; a CDS tells how much the market charges to insure it against default. These are two measures of the same risk, and they do not always coincide, which is what makes the second one interesting. The sovereign credit default swap is a thermometer of its own, more liquid on some names, sometimes faster to react, and governed by its own rules, those of ISDA. Reading it means having a second opinion on a country's solvency, provided one knows its mechanics. This guide runs through them, with Italy as the worked example. What a sovereign CDS is A sovereign CDS is a bilateral insurance contract. The protection buyer pays a periodic premium to the seller; in exchange, if the reference state suffers a credit event, the seller compensates the loss. The premium, called the CDS spread, is expressed in basis points of the insured notional, on an annual basis, and the reference maturity is five years, the most liquid. This premium is the direct translation of perceived default risk: it rises when the market judges default more likely, falls when it is reassured. The market is concentrated on a few names. Italy is the most traded sovereign CDS in the European Union, with an average daily notional of about $571 million, and its five-year CDS traded around 30 basis points in mid-2026. Thirty basis points is the annual price, $30,000, to insure ten million of Italian debt against default: a low level, consistent with the calm in spreads we described in our analysis of Italy's borrowed calm. The credit event and ISDA A CDS's whole value rests on the definition of what triggers payment, and that is where the contract turns technical. CDS are governed by ISDA documentation, which standardises credit events: failure to pay, restructuring, and for sovereigns repudiation or moratorium. Recognising an event is not left to the parties: it falls to an ISDA Determinations Committee, made up of ten sell-side and five buy-side members, which decides from public information. On an event, settlement most often runs through an organised auction that sets the recovery rate, hence the amount paid, equal to one hundred minus that recovery. That architecture has reading consequences. A CDS covers only ISDA-defined events: a state can see its debt fall sharply, inflicting heavy losses, without any credit event being declared, if the legal form does not fit the grid. The CDS insures against a characterised default, not against a mere mark-to-market loss. The CDS-bond basis The CDS comes into its own when set against the bond, and that gap has a name: the CDS-bond basis. In theory, the CDS premium and the bond's credit spread for the same issuer should coincide, since they measure the same risk. In practice, they diverge, and the divergence is a signal. A negative basis, a CDS cheaper than the bond, often betrays funding frictions or balance-sheet constraints that weigh on the cash instrument without touching the derivative. Reading the basis means crossing two markets to spot which moves first, and the CDS, more liquid and shortable without holding the bond, sometimes reacts before the cash spread we describe in our guide on credit spreads. The euro-area case: redenomination A euro-area sovereign CDS hides a subtlety no other carries: redenomination risk, the probability that a state leaves the euro and repays in a devalued national currency. Recent contracts include clauses treating a redenomination outside the reference currencies as a credit event, so the Italian or French CDS carries, on top of the classic default risk, an insurance premium against the bloc's break-up. That is why, in a stress episode, these CDS can jump faster than solvency alone would justify, and why the ECB's backstop, the TPI, weighs indirectly on their level, as we analyse in our guide on European sovereign debt. Reading pitfalls A few reflexes avoid errors. The first is not to confuse the CDS with an exact default probability: the conversion depends on an assumed recovery rate, and a liquidity or scarcity premium can inflate the level. The second is to account for market size: on a thinly traded sovereign, the CDS can overreact for lack of depth, and its move say more about positioning than about risk. The third is to remember that the CDS insures a legally defined event, not a market loss. The fourth, specific to the euro area, is never to forget the redenomination component, which can make the CDS diverge from the simple credit spread. Reading a sovereign CDS in practice The method fits in a few moves. Read the premium in basis points as a thermometer, converting it mentally into a default probability via recovery. Systematically compare the CDS with the bond spread to read the basis and spot the market that moves first. Check the name's liquidity before interpreting a move. And, for a euro-area sovereign, isolate the redenomination component from pure default risk. The CDS is not a truth superior to the spread, it is a second opinion, governed by its own rules, and its true usefulness is in the gap it keeps with the cash market. --- Sources - MacroMicro, "Italy 5-Year Credit Default Swaps" (Italy ~30 bp at 5 years, most traded EU sovereign CDS, daily notional ~$571m) - ISDA, "The Credit Event Process" (credit events, Determinations Committee, auction settlement) - ESRB, "Credit default swaps · analysis and policies" (European sovereign CDS market, notionals and concentration) ============================================================================ REFERENCE GUIDE: How to Read Central Bank FX Reserves (COFER and the gold share) URL: https://l0g.fr/en/guides/read-central-bank-fx-reserves-cofer/ Canonical French source: https://l0g.fr/guides/lire-les-reserves-de-change-cofer/ Date: 2026-07-28 (reviewed 2026-07-28) ---------------------------------------------------------------------------- import InfographicBars from '../../components/infographic/InfographicBars.astro'; Foreign exchange reserves tell what a central bank keeps at hand when everything goes wrong, and as such they betray its true preferences better than its speeches. But reading them is a minefield, because a number that moves is not necessarily a decision that was taken. The dollar's share can fall from one quarter to the next without any central banker having sold a single greenback, simply because the dollar depreciated. Reading reserves means first learning to separate what price did from what allocation intended. This guide gives that grid, with 2026 data and the de-dollarisation narrative as backdrop. What COFER measures The reference source is COFER, published each quarter by the International Monetary Fund. It measures the currency composition of the world's official foreign exchange reserves, the holdings of foreign currencies central banks keep to intervene and fund themselves. Two scope points are essential. First, COFER covers only so-called allocated reserves, those whose currency is reported to the IMF; a fraction stays unallocated, though coverage is now nearly complete. Second, and crucially, COFER excludes gold: it measures only currencies. The gold share is computed on total reserves, a different set. Within that scope, the dollar remains dominant, but its reign is slowly eroding. Its share went from close to 71% in 2000 to about 57% in the first quarter of 2026, a real but gradual decline, spread over a quarter of a century, nothing like the abrupt collapse sometimes announced. The valuation trap Here is the error that traps half the commentary. When the dollar's share moves, the first question is not "why did central banks sell", but "did the price change". Because reserves are valued in dollars, and the mere move of exchange rates shifts the shares without any allocation having changed. The 2026 data illustrate it perfectly: the dollar's share rose to 57.13% in the first quarter of 2026, from 56.42% at end-2025, but that rise was mostly down to the dollar's appreciation, the valuation effect explaining about half the move. A quarter earlier, the same mechanics ran the other way: when the dollar's share had fallen, 92% of the decline came from exchange rates, not portfolio decisions. The lesson is a discipline: never read a change in share without correcting for the valuation effect. A rising share can hide sales, a falling share can hide purchases. Only the measure at constant exchange rates reveals what central banks actually decided, and it almost always tells a duller story than the price. Gold, outside COFER The same trap replays, amplified, on gold. In 2025, gold overtook US Treasuries as a share of official reserves, a spectacular shift we discussed in our analysis of the debasement trade. But that crossover was driven almost entirely by the rise in the gold price, not by a rebalancing, and it does not show up in COFER's dollar share, which stayed broadly stable. When the metal jumps, its value in reserves swells mechanically, without an ounce being bought. That does not mean nothing is happening. Central banks are indeed buying gold at a sustained pace, a real flow we track in our piece on the tonnes accumulated, and to measure it one must look at tonnes, not value. The right reading therefore overlays two planes: valuation, which makes up most of the share moves, and flow, slower, which betrays intent. Confusing the two means taking a price rise for a geopolitical decision. Reading pitfalls Beyond valuation, a few reflexes are in order. The first is to distinguish allocated and total reserves: COFER covers the former, the gold share the latter, and mixing them produces nonsense. The second is to beware missing reporters: some large holders, China first, do not report the fine composition of their reserves, which makes measuring their choices indirect. The third is the lag: COFER data appears a quarter late. The fourth, finally, is narrative bias: de-dollarisation is a real but slow move, and the temptation is great to read every price wobble as the end of the dollar, when the true flows, corrected for valuation, sketch a gradual, partial diversification, consistent with our reading of the narrative against the numbers. Reading reserves in practice The method fits in four moves. Always correct a change in share for the exchange-rate effect before reading it as a decision. Measure gold in tonnes, never in value, to isolate flow from price. Cleanly separate COFER, which speaks currencies and allocated reserves, from the gold share, which speaks total reserves. And cross the hard data with declared intentions, like the World Gold Council's annual survey, to anticipate coming flows. Read this way, the 2026 picture is sober: the dollar around 57%, stable once valuation is neutralised; gold really accumulated but whose share surge owes almost everything to the price. De-dollarisation exists, it is slow, and it is read in the flows, not in the prices. --- Sources - IMF, "IMF Data Brief: Currency Composition of Official Foreign Exchange Reserves", July 2026 (dollar share 57.13% in Q1 2026, valuation effect, gold outside COFER) - BestBrokers, "US Dollar Share of Global Currency Reserves in 2026" (92% of the share decline due to exchange rates in Q2 2025) - World Gold Council, "Central Bank Gold Reserves Survey 2026" (central banks' purchase intentions) ============================================================================ REFERENCE GUIDE: How to Read a Consumer Credit Securitisation (Auto, Card, BNPL ABS) URL: https://l0g.fr/en/guides/read-a-consumer-credit-securitisation-abs/ Canonical French source: https://l0g.fr/guides/lire-une-titrisation-de-credit-a-la-consommation/ Date: 2026-07-27 (reviewed 2026-07-27) ---------------------------------------------------------------------------- import InfographicBars from '../../components/infographic/InfographicBars.astro'; An auto loan, a card balance, a pay-in-four: taken alone, each is a tiny bet on a household. Pooled by the thousand in a dedicated vehicle, they become a rated bond, bought by funds worldwide. This is consumer credit securitisation, the vehicle through which the borrower's risk leaves the lender's balance sheet and disperses, as we tracked in our analysis of the risk's journey, from the credit card to the annuity. Knowing how to read that ABS means answering three questions: what is in the pool, who absorbs losses first, and what triggers when things go wrong. This guide runs through the grid, with the 2026 market as the backdrop. What a consumer ABS is An ABS, for asset-backed security, is a security backed by a pool of non-mortgage receivables: auto loans, credit card balances, point-of-sale loans and instalment payments (BNPL). The mechanism is always the same. An originator transfers its receivables to a trust bankruptcy-remote from itself; that trust issues bonds whose cash flows, the borrowers' interest and principal, pay the investors. The difference from a CLO, which our guide details, lies in the collateral: the CLO is backed by corporate loans, the consumer ABS by household credit, smaller, more numerous and more granular. The market is huge and growing. US ABS outstanding is around $1.6 trillion, and 2026 issuance is expected near $385 billion, a post-financial-crisis record, with autos in the lead at more than a third of volumes. That is the scale of the channel through which consumer credit refinances itself. The waterfall and the four cushions The heart of an ABS is its protection structure, and that is where its soundness is read. The bonds issued are cut into ranked tranches, from the senior AAA to the equity tranche. Losses climb from the bottom up: the equity takes them first, the senior last. But before even the first tranche suffers, four cushions of credit enhancement absorb defaults, and they must be read in the order they are consumed. The first is excess spread, the gap between the interest collected on the pool, often high in subprime, and the interest paid to noteholders. This surplus mops up current losses month after month, without touching principal. The second is the reserve account, a cash cushion set up by the originator and refilled by excess spread. The third is overcollateralization: the pool of receivables is worth more than the bonds issued, and that excess collateral absorbs losses before the tranches. The fourth, finally, is subordination itself, the hierarchy where junior tranches protect senior ones. Reading an ABS means measuring the total thickness of these four cushions, because it is that, not the quality of the pool, that justifies the senior tranche's AAA rating. Reading the pool Then comes the examination of the collateral, where the real risk lies. A few metrics place a pool. The average credit score first: a prime pool shows a high FICO, a subprime pool runs around 550 or lower. Expected cumulative loss next, the most telling indicator: it measures the share of the pool the issuer expects to lose over the life of the deal, net of recoveries, and it can exceed 20% in a subprime auto pool against a few points in prime. Add loan maturity, seasoning, loan-to-value and concentration, geographic or by borrower. A single product, the auto ABS, thus covers two worlds that everything separates. Triggers and early amortisation An ABS is not a frozen structure: it carries circuit breakers. These are the performance triggers, thresholds on delinquencies or cumulative losses that, if breached, alter the waterfall mid-course. Concretely, a tripped trigger redirects cash flows to pay down the senior tranches faster, cuts payments to the lower tranches, or forces an early amortisation that winds down the structure ahead of term. These mechanisms protect the senior investor at the junior's expense at the very moment the pool deteriorates. Reading them means understanding that the AAA tranche's protection is not only static, in the thickness of the cushions, but dynamic, in these circuits that reconfigure under stress. What an ABS is not: the CDO of 2008 Here is the misconception to avoid at all costs, because it conditions any risk judgement. These securitisations are readily conflated with the instruments that blew up the system in 2008. That is wrong, and the distinction is structural. The CDO that exploded then was not backed by loans, but by tranches of other securitisations, most often subordinated tranches of mortgage credit, a securitisation of securitisations. When house prices fell everywhere at once, correlation wiped out all the tranches at the same time, including the highest. A consumer ABS is of another nature: it is backed directly by thousands of small loans, granular and weakly correlated with each other, short-dated, amortising fast and offering real recoveries, repossessing the car for instance. That is why auto and card ABS came through 2008 without the mortgage CDOs' catastrophe. Confusing the two is getting the wrong crisis. Reading pitfalls A few reflexes avoid the classic errors. The first is never to confuse the rating with the pool: a AAA does not mean the collateral is sound, but that the tranche is protected by everything beneath it; a mediocre pool can carry a very thick AAA tranche. The second is to watch excess spread, the first cushion consumed, which thins as funding cost rises: an ABS comfortable at low rates can tighten at high rates. The third is to doubt recoveries, which assume a liquid used-car market; in a recession, resale value collapses just as defaults climb. The fourth is valuation when these assets sit in private credit funds, where they are marked to model rather than to market, which we dug into in our analysis of one asset, two prices. The fifth, finally, is not to read an issuance record as a stress signal: the market can set volume records while pool quality erodes from the bottom. Reading a consumer ABS in practice The method fits in a few moves. Start with the pool, never with the rating: average score, expected cumulative loss, maturity, concentration. Then measure the thickness of the enhancement, the four cushions stacked beneath the tranche you are looking at, because that is the true measure of its protection. Spot the triggers, those circuit breakers that will redirect cash flows on deterioration. Place the pool in the cycle by crossing it with the household delinquencies tracked by the New York Fed, which give the collateral's macro context. And remember the founding distinction: a consumer ABS disperses a granular, amortising risk, it does not stack correlated tranches the way the 2008 CDO did. The rating is read only last, once one has understood what it protects and against what. --- Sources - Coinlaw, "Asset-Backed Securities in 2026: A $1.6 Trillion Market" (ABS outstanding ~$1.6tn, 2026 issuance ~$385bn, autos in the lead) - Auto Remarketing, "3 subprime ABS pool elements pushing losses to near recessionary peaks" (subprime FICO ~550 or lower, expected cumulative loss 20%+) - IMF, "US Asset-Backed Securities Monitor", October 2025 (prime auto delinquencies ~1.9%, subprime ~16%) - NAIC, "Auto Asset-Backed Securities Primer" (trust structure, subordination, overcollateralization, excess spread, reserve account) ============================================================================ REFERENCE GUIDE: How to Read a CBDC: the digital euro parameter by parameter URL: https://l0g.fr/en/guides/read-a-cbdc-the-digital-euro-parameter-by-parameter/ Canonical French source: https://l0g.fr/guides/lire-une-cbdc-euro-numerique-parametre-par-parametre/ Date: 2026-07-27 (reviewed 2026-07-27) ---------------------------------------------------------------------------- import InfographicBars from '../../components/infographic/InfographicBars.astro'; A central bank digital currency is not a thing, it is a list of decisions. Behind the acronym CBDC hides a dozen parameters, and each is a political trade-off disguised as a technical choice: how much can be held, does it pay interest, who distributes it, what does the central bank see of your payments, can its use be programmed. Reading a CBDC is therefore not judging it as a block, it is decoding these settings one by one, because they reveal the issuer's real priority, somewhere between reclaiming monetary sovereignty, protecting its banks and the temptation of control. This guide runs through the grid, with the digital euro as the worked example. Retail or wholesale: the first fork The first question to ask of any CBDC project is who it is for. A retail CBDC targets the general public: it is central bank money in households' wallets, a digital equivalent of the banknote. A wholesale CBDC stays reserved for banks and market infrastructure, where it serves as a settlement asset between institutions, notably on the tokenised platforms we describe in our analysis of the ECB's tokenisation projects. The two raise different stakes: the wholesale version is a matter of interbank plumbing, discreet and rarely controversial; the retail version touches the direct relationship between citizen and central bank, and concentrates the entire debate. The digital euro discussed here is a retail CBDC. Account or token: the architecture Next comes the architecture choice, which governs almost everything else. An account-based CBDC ties each unit to its holder's identity, like a bank deposit: ownership follows the entry in the ledger. A token-based CBDC works like cash: ownership follows possession of the token or knowledge of a key, without identity necessarily attached. This choice is not neutral, because it determines the degree of privacy possible and the ability to work offline. The digital euro combines both logics: an online leg, intermediated and identity-linked, and an offline, token-like leg, closer to cash. Who distributes: the two-tier model Third parameter, often underestimated: does the central bank serve the public directly, or through intermediaries? The direct model would make the central bank the account keeper of millions of citizens, an institutional revolution almost no Western project envisages. The digital euro instead adopts a two-tier model: it will rest on a partnership between the Eurosystem and payment service providers, commercial banks, payment institutions and e-money institutions. The central bank issues, the intermediaries distribute and keep the customer relationship. This choice is the first answer to the fear of disintermediation: banks stay in the loop and collect the fees, rather than being bypassed. The holding limit: the stability lock The most scrutinised parameter is the holding limit, the maximum an individual can keep. The ECB envisages it between €500 and €3,000, a deliberately low level compared with the British and Canadian projects, precisely to prevent a massive flight of deposits into central bank money. It comes with a waterfall mechanism, which sweeps any excess to the linked bank account, a reverse waterfall that reloads the wallet on the fly, plus a ban on holdings for companies. We detail the stakes of this lock in our analysis of the digital euro and its cap; the international comparison shows how low the euro-area bound is. Remuneration: the heaviest parameter Then comes a discreet but decisive setting: does the CBDC pay interest? It is arguably the most important design feature, because it decides whether the digital currency competes with savings. A remunerated CBDC would become a risk-free investment backed by the central bank, sucking in deposits; an unremunerated CBDC stays a mere payment instrument. The digital euro has settled on zero: no interest, so that it never rivals a savings account or a term deposit. That choice, combined with the cap, forms the core of the anti-flight design. Reading a CBDC's remuneration means reading how far its issuer accepts that it encroaches on the banking system. Privacy: the impossible triangle No parameter is more loaded than privacy, and it obeys a trade-off research sums up as a trilemma: one cannot maximise privacy, financial stability and regulatory compliance all at once. Strengthening anonymity weakens anti-money-laundering; strengthening traceability worries civil liberties. Each CBDC positions itself somewhere in that triangle, and its place is a choice of society. The digital euro sits there in tiers. Offline, it promises a cash-like confidentiality, where only payer and payee see the transaction; online, it applies data minimisation, where the intermediary sees only what is strictly necessary for anti-money-laundering checks, and where the ECB itself does not see individual payments. Small amounts would be the most protected. The CNIL and its counterparts follow the file closely, because it is here that the project's reputation is decided: a CBDC perceived as a surveillance tool would be rejected before it exists. Legal tender and non-programmability Two last parameters, often confused, close the grid. The first is legal-tender status: the digital euro would have it, obliging most merchants to accept it, guaranteeing its universality against private means of payment and preserving the singleness of money. The second is programmability, and it is the most misunderstood. The ECB insists: the digital euro would be non-programmable money, each unit staying fungible and spendable everywhere, without being restrictable to certain goods or a date. This choice answers head-on the suspicion of a control currency: programmable money could in theory condition spending, a non-programmable CBDC refuses to by construction. Distinguishing programmable money, ruled out, from programmable payments, which automate a transaction without constraining the money, is the key to reading this debate honestly. The timeline and reversibility One final reading reflex: place the project on its trajectory, because the parameters are not set in stone. For the digital euro, the legal framework is being negotiated in 2026, the ECB launched its call for expression of interest to providers in March 2026, a twelve-month pilot is targeted from the second half of 2027, and a first issuance remains possible in 2029. Above all, several settings are designed to evolve: the cap in particular is a deliberately low dial at the start, liable to be raised once stability has been tested. Reading a CBDC therefore means reading its margins for revision too, because a parameter calibrated to be loosened later says a lot about the issuer's uncertainty. Reading a CBDC in practice Faced with any central bank digital currency project, the same grid applies. First check whether it is retail or wholesale, since the whole public debate concerns only the former. Spot the architecture, account or token, which governs privacy and offline use. Read the distribution model, direct or two-tier, which tells the fate reserved for banks. Then the two stability locks, holding limit and remuneration, which measure how far the CBDC encroaches on deposits. Next place the project within the privacy trilemma, between privacy, stability and compliance. Finally check legal tender and programmability, which separate a universal public currency from a control instrument. Each setting read in isolation is technical; together they sketch an intent. And that intent is judged against the competition the CBDC claims to counter, foreign payment networks and private dollar-backed stablecoins, because a public digital currency is first an answer to the question of who, tomorrow, will issue the money we use. --- Sources - European Central Bank, "Preparation phase of a digital euro, Closing report", October 2025 (two-tier model, holding limit, tiered privacy, timeline) - MDPI, Journal of Risk and Financial Management, "Designing Retail Central Bank Digital Currencies" (privacy-stability-compliance trilemma; remuneration as the decisive parameter) - CNIL, "Confidentiality in the Digital Euro: Where are we?" (privacy tiers, data minimisation) - adesso, "The Digital Euro 2026: What decision-makers need to know now" (call for expression of interest March 2026, 2027 pilot, possible 2029 issuance, non-programmability) ============================================================================ REFERENCE GUIDE: How to Read the NY Fed Household Debt Report URL: https://l0g.fr/en/guides/read-the-ny-fed-household-debt-report/ Canonical French source: https://l0g.fr/guides/lire-le-rapport-dette-et-credit-des-menages/ Date: 2026-07-27 (reviewed 2026-07-27) ---------------------------------------------------------------------------- import InfographicBars from '../../components/infographic/InfographicBars.astro'; The American consumer carries two thirds of the world's largest economy, and their health shows up first in a quarterly document: the New York Fed's Household Debt and Credit Report. It tells how much households owe, on what, and above all how many are starting to fall behind. The trap is to stop at the aggregate number, the one that makes the headlines, when all the useful information hides in the breakdown and in the flows. This guide takes the report end to end, from its method to its pitfalls, using the first quarter of 2026 as a worked example. What the HHDC measures The report, produced each quarter by the New York Fed's Center for Microeconomic Data, is not a survey of opinions: it rests on real credit data. Its source is the Consumer Credit Panel, a random 5% sample of all Americans holding a social security number and a credit file, supplemented by the other people living at the same address to reconstruct the household. The whole covers roughly 44 million individuals per quarter, from anonymised Equifax records, going back to 1999. It is a longitudinal panel: it follows the same people over time, which allows observing not only levels but transitions. The data is cut off at quarter-end and published about six weeks later: the first-quarter report, cut off at the end of March, thus appeared on 12 May 2026. It covers the main categories of household debt, mortgage first, then auto loans, credit cards, student loans, home equity lines (HELOC) and other, with balances, new originations, credit scores and delinquencies for each. The composition of debt First rule of reading: US household debt is overwhelmingly housing. Of the $18.79 trillion outstanding at the end of March 2026, the mortgage alone weighs $13.19 trillion, about 70% of the total. The rest splits between auto, student, card and home equity lines. That structure has a direct consequence: when a headline says household debt hits a record, it is mostly talking about housing and demographics. The real stress signal never comes from the aggregate level, it comes from the fast-turning compartments and from delinquencies. Stock, flow, severity The heart of the report is delinquency, and it must be read at three distinct levels too often confused. The first is the stock: the share of balances past due at a given moment. At the end of March 2026, 4.8% of debt was at least thirty days late, and 3.36% seriously delinquent at ninety days or more, about $631 billion. This is the snapshot, useful but slow to move. The second, and the most important, is the flow: the transition rate into delinquency, that is, the share of previously current balances that tip into arrears over the quarter. The report defines it precisely as the balances that newly became at least ninety days late in the reference quarter, divided by the balances that were current or less than ninety days past due in the previous quarter. This flow turns before the stock: it rises when deterioration begins, well before the total share of arrears shifts. It is the leading signal. The third is severity. The report isolates serious delinquency, defined as having at least one account ninety days or more past due, in collections, or classified as severely derogatory. It is the mark of households unlikely to recover, and it comes with last-resort indicators: collections, home foreclosures and personal bankruptcies. Holding the distinction is vital: the stock tells where we are, the flow tells where we are going, severity tells who is already lost. The trap of the aggregate Here is the most common reading error. The aggregate delinquency rate, at 3.36% for serious arrears, looks benign, and the debt-to-disposable-income ratio, down to 79.9%, its lowest since 2003 outside pandemic-stimulus episodes, seems to confirm iron health. But the aggregate adds together products that have nothing to do with each other, and it hides a K-shaped economy where housing stays immaculate while other compartments catch fire. The rule is therefore always to go one level down: by product, then by credit-score tier, then by age. The report allows it, and that is where the K-shaped distribution becomes visible. The top of the distribution, homeowning and well-scored, pulls the average towards calm; the bottom, young and loaded with revolving debt, falls away without the aggregate flinching. Our piece on the average consumer who does not exist unpacks that fracture beneath the reassuring aggregate. The student-loan worked example No compartment better illustrates the report's pitfalls than student loans in 2026. After the federal repayment freeze ended in 2025, previously suspended balances began being reported to the credit bureaus again, and delinquencies resurfaced all at once. The transition rate into serious delinquency, measured as a four-quarter moving sum, jumped then began to recede, going from 16.2% at the end of 2025 to 10.9% in the first quarter of 2026, while the stock ninety days or more past due stayed elevated at 10.3%. The reading lesson is twofold. First, a spike in the transition rate can be partly a reporting artefact: it is not only that more households default, it is also that defaults previously invisible reappear in the statistics. Second, flow and stock can diverge: here the flow recedes while the stock stays high, the sign of a reset wave passing its peak without having cleared. Taking the rising flow for a catastrophe, or its recession for a cure, would be a double error. Reading pitfalls Beyond the misleading aggregate and the student artefact, a few reflexes avoid misreadings. The first is to distrust the nominal record. Household debt grows almost mechanically with prices and population; the dollar figure sets records nearly every quarter without meaning much. The relevant measure is the debt-to-income ratio, which at 79.9% tells the opposite story of relative deleveraging. The second is not to confuse the series: the HHDC, drawn from the Equifax panel, does not match the delinquency rates the Fed publishes on commercial-bank balance sheets, which have a different scope and definition. The third rests on a methodological subtlety: for mortgages, new delinquency is measured on the account balance at its entry into arrears, while for other loans it is measured on the net increase in the delinquent balance, which makes the compartments not strictly comparable. The fourth is the lag: the data is about six weeks old, and foreclosures and bankruptcies are late indicators, which confirm stress more than they announce it. First quarter 2026: the deceptive calm The May 2026 report is a textbook case of dissonance between the aggregate and the tail. On the surface, total debt barely rises, by 0.1% on the quarter, to $18.79 trillion, some 3.2% more year on year; card balances fall by $25 billion, auto gains $18 billion; the debt-to-income ratio drops to a two-decade low; the aggregate delinquency rate is stable. Nothing, at that level, to worry about. Below the surface, the picture shades. Home foreclosures touch 59,160 consumers and bankruptcies 124,020, slightly higher; collections rise to 5% of consumers; student loans stay at 10.3% serious delinquency, and subprime auto credit, outside the HHDC's scope alone, sits at delinquency levels unseen since the 1990s. The aggregate says calm, the distribution says fracture. The report, read correctly, says both at once, and that is precisely its value. Reading the HHDC in practice To draw the right signal from the report, a few moves suffice. Look first at the transition rates, not the delinquency levels: the flow leads the stock. Systematically go below the aggregate, by product then by score and age, to spot the tail falling away. Scale debt to income rather than to the dollar, to ignore the false nominal record. Treat severity, foreclosures and bankruptcies as late confirmations, not as alerts. Finally, cross the HHDC with the labor market, because employment remains the firewall of consumer credit: as long as it holds, delinquency stays manageable, and it is the labor market to watch to anticipate the next quarter. The risk does not vanish when it leaves the aggregate: it migrates towards non-bank intermediation and the securitisation that buy up these receivables, and that is where it must be tracked next. --- Sources - Federal Reserve Bank of New York, "Household Debt Balances Rise Slightly as Delinquency Transition Rates Hold Steady", 12 May 2026 (total debt $18.79tn, composition by category) - Federal Reserve Bank of New York, "Household Debt and Credit, Background" (Consumer Credit Panel methodology, 5% Equifax sample, panel since 1999) - Wolf Street, "Household Debts, Debt-to-Income Ratio, Serious Delinquencies, Foreclosures, Collections & Bankruptcies in Q1 2026" (90-day delinquency by category, debt-to-income 79.9%, foreclosures and bankruptcies) - Center for Microeconomic Data, New York Fed, Q1 2026 (student-loan transition rate, 16.2% to 10.9% as a four-quarter moving sum) ============================================================================ REFERENCE GUIDE: How to read an IMF program: quota, SDRs, facilities and repayment schedules URL: https://l0g.fr/en/guides/read-an-imf-program/ Canonical French source: https://l0g.fr/guides/lire-un-programme-du-fmi/ Date: 2026-07-21 (reviewed 2026-07-21) ---------------------------------------------------------------------------- When a country "calls the IMF," what unfolds is more precise than the headline. The International Monetary Fund lends to a state in balance-of-payments trouble, but through a codified, quantified and public mechanism, where every amount refers to one common measure, the country's quota. This guide walks through that mechanism end to end: what the Fund is for, how quota governs access, what the SDR is, which windows and instruments exist, how to read access limits and exceptional access, and above all how to read the data tables the IMF publishes for each member. The running case study is our X-ray of the IMF's record exposure to Argentina, where these figures come to life. What the Fund is for The IMF, founded in 1945, pursues three functions the Congressional Research Service sets out clearly. Surveillance: the Fund monitors the economic policies of its 190 members and flags risks. Capacity development: it trains and advises administrations. And loans, our focus here: the Fund finances a state facing a balance-of-payments difficulty, meaning it cannot pay for imports or service its external debt without exhausting its reserves. A point of vocabulary avoids much confusion. The IMF does not do development aid or project lending; it fills an external financing gap, until an adjustment restores balance. That is why its disbursements come in tranches, released after verifying that conditions have been met, the conditionality. Loan and reform move together, and the schedule of reviews paces the payout. The quota, key to everything Nothing at the IMF makes sense without the quota. Each member subscribes an amount, denominated in Special Drawing Rights, meant to reflect its weight in the world economy. This quota determines four things at once: the country's contribution to the Fund's resources, its voting power, its SDR allocation, and above all the limits of what it can borrow. The United States, the largest contributor, holds 17.43% of the votes, which gives it a de facto veto over major decisions, subject to an 85% supermajority. The practical consequence is that at the IMF everything is measured as a percent of quota, not in absolute dollars. Credit outstanding "at 1,335% of quota" says immediately that a country has drawn more than thirteen times its contribution, a level only extraordinary programs reach. Keeping this unit in mind is the first reading skill for any Fund file. The SDR, unit of account and reserve asset The Special Drawing Right, or SDR, is both the Fund's unit of account and an international reserve asset it created. Its value derives from a basket of five currencies, the dollar, euro, renminbi, yen and pound sterling, the renminbi having been admitted in 2016. The Fund publishes its value every business day: in mid-2026, one SDR is worth about $1.36. Since quotas, loans and repayment schedules are all denominated in SDRs, converting to dollars requires knowing that day's rate, which is why the same credit outstanding is expressed differently depending on the conversion date. The SDR also serves as a liquidity instrument. In general allocations, such as the $650bn distributed in 2021, each member receives SDRs in proportion to its quota, which it can exchange for currencies. A country that has "consumed" its SDR holdings, as the Fund's tables show, has therefore already mobilised that reserve. Windows and instruments The Fund lends through two main windows. The General Resources Account, the GRA, carries non-concessional loans, at market rates, open to all members. The Poverty Reduction and Growth Trust, the PRGT, carries concessional loans reserved for low-income countries. The distinction is decisive for reading a portfolio concentration: it is on the GRA alone that the weight of the largest borrowers is measured. Within these windows, several instruments meet distinct needs, per the CRS definitions: - Stand-By Arrangement (SBA): the historical instrument, for a short-term balance-of-payments imbalance, generally over one to two years. - Extended Fund Facility (EFF): for a deeper imbalance requiring structural reforms, over three years or more. This is the format of recent large programs. - Flexible Credit Line (FCL): a precautionary line for countries with strong fundamentals, drawable without new conditionality, on so-called ex-ante conditionality. - Precautionary and Liquidity Line (PLL): for countries close to but not quite eligible for the FCL. - Rapid Financing Instrument (RFI) and its concessional counterpart, the RCF: emergency assistance without a full program, useful against a sudden shock such as a natural disaster. Access limits and exceptional access A member cannot borrow without a ceiling. Access is bounded by two limits expressed as a percent of quota, one annual, one cumulative over time. As long as credit outstanding stays below the normal cumulative limit, currently set at 600% of quota, the program follows the ordinary procedure. Beyond it, the country enters exceptional access, a category that triggers reinforced safeguards: the Fund must verify four criteria, including debt sustainability and a reasonably assured capacity to repay, and produce a dedicated assessment of the risk it takes on. This threshold is not theoretical. In the Argentine case, credit reached 1,335% of quota, more than double the normal limit: the program falls entirely under exceptional access, which is why a special assessment of the Fund's exposure was published. Spotting whether a file is in normal or exceptional access is a reliable shortcut for gauging its intensity. The cycle: tranches, reviews, waivers A program is not paid out in one block. After Executive Board approval, the amount is disbursed in tranches, each conditioned on a review that verifies compliance with quantitative performance criteria, for example a floor on net international reserves or a ceiling on domestic credit. When a criterion is missed, two outcomes exist: the program derails, or the board grants a waiver, often paired with corrective measures and a recalibration of the following targets. Reading a review's press release therefore comes down to three questions: is the tranche disbursed, which criteria were met or missed, and which waivers were granted. This vocabulary is that of the official documents. In the Argentine file, the end-2025 reserve target was missed, a waiver granted and the following targets modified, a sequence typical of a program under strain that does not break. The borrower's bill The Fund's credit is not free. On top of a basic rate of charge, high or long-standing credit carries surcharges, additional fees above a quota threshold. Cut back in late 2024, they remain a significant source of Fund income, and the largest borrower is the largest contributor. A repayment schedule therefore reads in two components: principal, which repays the drawing, and charges, including interest and surcharges, which pay the lender. On large programs, the second component is far from trivial. On the Fund's side, two notions complete the reading. Precautionary balances are the capital cushion that absorbs a potential arrears event; when exposure to a single country exceeds them, concentration risk becomes explicit. Preferred-creditor status, finally, means the IMF is repaid before other creditors: that is why, historically, arrears to the Fund have been resolved without a definitive principal loss, a point any judgment on its exposure must factor in. Reading the IMF's tables For each member, the Fund publishes a financial-position sheet that condenses all of the above. Knowing how to read it is the final aim of this guide. It shows, in order: the quota in SDRs; the country's SDR holdings, where a near-zero level signals an already-mobilised reserve; outstanding purchases and loans by instrument; recent arrangements, with their type (EFF, SBA), amount approved and amount drawn; and finally the schedule of forthcoming payments, split between principal and charges, year by year. Three primary sources are enough to reconstruct a file. The member financial position gives quota, credit outstanding and the schedule. The GRA credit outstanding table lets you compute a country's weight in the non-concessional portfolio. The SDR valuation provides the day's conversion rate. With these three pages, you rebuild a program's figures yourself, without relying on a press summary. This is the approach we followed to quantify the Argentine case and, more broadly, to place the global safety net in our analysis of the emerging markets double squeeze. The model's limits Rigour demands naming the criticisms, since they structure the debate. Conditionality is regularly accused of imposing procyclical austerity; preferred-creditor status, of putting the Fund ahead of populations; portfolio concentration on a handful of very large programs, of blending financial risk with geopolitical considerations. These objections are beyond this guide, which describes a mechanism, but an informed reader keeps them in mind: reading an IMF program means understanding at once the sum a country receives, the sum it will owe, and the solidity the lender puts at stake. --- Primary and official sources: IMF, member financial position (Argentina example, 30 June 2026); IMF, GRA credit outstanding; IMF, SDR valuation; Congressional Research Service, "The International Monetary Fund" (IF10676) and detailed report R42019 for functions, quota, US voting share and instrument definitions; IMF, assessment of the Fund's financial exposure (April 2025) for the normal access limit, exceptional access, precautionary balances and surcharges. To apply: our X-ray of the IMF's record exposure to Argentina and our analysis of the emerging markets double squeeze against a strong dollar. Counterpoint on market-based sovereign debt: reading European sovereign debt. Instrument definitions are structural; the numerical thresholds (access limit, surcharges) are reviewed periodically by the Fund, and the levels cited are those in force in mid-2026. ============================================================================ REFERENCE GUIDE: Reading the ECB balance sheet and the Target2 balances URL: https://l0g.fr/en/guides/read-ecb-balance-sheet-target2/ Canonical French source: https://l0g.fr/guides/lire-le-bilan-bce-target2/ Date: 2026-07-19 (reviewed 2026-07-19) ---------------------------------------------------------------------------- import InfographicBars from '../../components/infographic/InfographicBars.astro'; The American central bank's balance sheet can be read every Thursday in the H.4.1. Its European counterpart exists, published every Tuesday, and almost nobody reads it: the Eurosystem's consolidated weekly financial statement. The difference fits in one word, consolidation: where the Fed is a single house, the Eurosystem aggregates the ECB and twenty national central banks, connected by an internal plumbing, Target2, whose balances have fed the wildest readings for fifteen years. This guide decodes the whole: the balance sheet, its rundown, the balances that alarm, the losses that worry and the framework that steers rates. It extends, on the euro side, our reading of the Fed's balance sheet. One balance sheet, twenty central banks The ECB holds only a fraction of monetary policy assets. The bulk sits on the balance sheets of the national central banks, the Bundesbank, the Banque de France, the Banca d'Italia and their peers, which execute the operations decided in Frankfurt. Together they form the Eurosystem, and its consolidated accounts appear every week. Asset purchases and the associated risks are mostly distributed according to the capital key, each country's share in the ECB's capital, computed from its population and GDP: roughly 26% for Germany, 20% for France, 17% for Italy. Orders of magnitude first. At the end of 2025, the consolidated balance sheet stood at €6,293 billion, against a peak of about €8,800 billion in the summer of 2022. On the asset side, two blocks dominate: the securities portfolios accumulated during the quantitative easing years, the historical APP (€2,322 billion at the end of 2025) and the pandemic-era PEPP (€1,423 billion), plus a residue of refinancing operations that have become marginal. On the liability side, banknotes in circulation (about €1,612 billion), minimum reserves (€172 billion) and above all excess liquidity, the cash banks park at the central bank beyond their requirements. The logic is the same as for the Fed: bank reserves are not a choice made by banks, they are the accounting residual of everything else. QT, the European way Since 2023, the Eurosystem has let its portfolios melt by no longer reinvesting maturing securities, the APP first, the PEPP since 2025. The pace is deliberately passive, "measured and predictable" in the formula the ECB repeats at every decision: no sales, only redemptions. Over 2026 this returns about €500 billion to the market, €330 billion of APP and €173 billion of PEPP, with the balance sheet expected around €5,800 billion by year-end. The contrast with America deserves emphasis: the Fed closed its quantitative tightening in December 2025 under pressure from repo market strains, while the ECB carries on without visible friction. The explanation is the cushion: euro area excess liquidity still stood at €2,358 billion in the spring of 2026, against a peak of €4,748 billion at the end of 2022. European banks are still swimming in cash: their recourse to regular refinancing operations averages a mere €24 billion, a crumb at the system's scale. The question for the coming years is the one the Fed just settled at its own expense: how far down can you go before the plumbing creaks? Our guide on US net liquidity details how elusive that threshold is. Target2, the plumbing that panics people Now for the most misunderstood piece. Target2, rebranded T2 in its modernised version, is the euro area's large-value payment system, the RTGS where cross-border interbank transfers settle in central bank money. When an Italian bank pays a German bank, the Banca d'Italia's position towards the ECB deteriorates and the Bundesbank's improves. Accumulated over time, these flows form the Target2 balances: as of 31 May 2026, the Bundesbank posted a claim of nearly €1,066 billion on the Eurosystem, while Spain carried a liability of about €489 billion and Italy a debtor position of the same order of magnitude. These vertiginous figures have fed an entire literature on the "hidden bailout" of the South by German savings, popularised by the economist Hans-Werner Sinn at the height of the debt crisis. The mechanical reading is more prosaic. A Target2 balance is not a loan granted by the Bundesbank: it is the accounting footprint of payment flows, which swells when capital flees a country (2011-2012), but also, and this is the neglected point, when the central bank buys securities from counterparties whose accounts sit elsewhere: a large share of the rise in balances after 2015 reflects the mechanics of QE, not capital flight. The German balance in fact peaked at €1,269 billion in December 2022, at the balance sheet's peak, before receding with QT. There remains the limit scenario, the only one in which these balances stop being an accounting entry: a country leaving the euro area, which would turn its debtor position into a real claim on a departed state, with recovery prospects uncertain at best. This is the bridge between Target2 and the redenomination risk we described in our guide on European sovereign debt: as long as euro membership is not in doubt, the balances are a thermometer of flows; the day it were, they would become one of the invoices of the rupture. Central banks in the red, so what? Another recurring source of alarm: the losses. The mechanics are simple to state. During the zero-rate years, the Eurosystem bought trillions of low-yielding securities, financed with central bank money. When policy rates rose, the remuneration paid to banks on that liquidity exceeded the portfolios' return: the interest margin turned negative. The Bundesbank thus recorded a seventh consecutive year without profit, with an €8.6 billion loss in 2025 and an accumulated loss carry-forward of €27.8 billion. The Banque de France, €7.7 billion in the red in 2024, returned to a profit of €8.1 billion in 2025, helped by an exceptional €11 billion gain from selling gold bars held in New York. The ECB itself shows negative accounting equity under the conventions it applies. Should this alarm anyone? A central bank is not a bank: it cannot run out of the money it issues, and no prudential rule imposes a solvency ratio on it. Its losses are a deferred cash-flow phenomenon: provisions built in the fat years absorb the shock, revaluation reserves (the Bundesbank's gold lifts its net equity to €363 billion) sleep beneath the negative carry-forward, and the return of a positive margin as QT proceeds will work off the stock. The real stake lies elsewhere, and it is political: a central bank durably in the red pays no dividend to the state, and becomes a convenient target for anyone keen to contest its independence. The ridge line is not an accounting one, it is institutional. The floor: rates and the operational framework Steering short-term rates changed in nature with the abundance of liquidity. In the operational framework revised in March 2024, the deposit facility rate became the central instrument: it sets the floor to which the overnight rate, the €STR, sticks as long as excess liquidity remains massive. The spread between the main refinancing rate and the deposit facility was narrowed to 15 basis points, so that banks will not hesitate to borrow at the window once liquidity grows scarce: the system is designed to evolve from a pure floor towards a demand-driven regime, without jolts. The levels, as of mid-2026: on 11 June, the ECB raised its three rates by 25 basis points, the first hike since 2023, taking the deposit facility to 2.25%, the main refinancing rate to 2.40% and the marginal lending facility to 2.65%. The motive: inflation pressures born of the war in the Middle East, with inflation expected at 3.0% in 2026 before returning towards 2.0% by 2028. Tightening rates in the middle of a balance sheet rundown illustrates the separation doctrine: rates steer inflation, the balance sheet follows its own path. We covered the Sintra turn in our analysis of the 2026 forum. The thermometers Monitoring the system comes down to a handful of indicators, all public. Excess liquidity, published in the weekly statements and the ECB's bulletins, tells the thickness of the cushion: €2,358 billion in the spring of 2026, falling by about €100 billion a quarter. Recourse to refinancing operations, minuscule today, will be the first signal of liquidity turning scarce: the European equivalent of banks tapping the Fed's standing facility. The gap between the €STR and the deposit facility, today glued a few basis points below the floor, will narrow and then flip above it as the cushion thins. The Target2 balances, published monthly by the national central banks, remain the thermometer of cross-border flows. And the consolidated weekly financial statement, every Tuesday, gives the overall photograph, the counterpart of the American H.4.1. Reading the balance sheet in practice Three misunderstandings organise most bad readings of the European balance sheet, and this guide equips you to avoid them. The first reads the Eurosystem as a single central bank: it is a federation of balance sheets, allocated by capital key, and this architecture explains both the Target2 balances and the distribution of losses. The second turns Target2 balances into enforceable claims: they are accumulations of flows, meaningful as a thermometer, dangerous only in a euro break-up scenario that belongs to redenomination risk, not to ordinary operation. The third treats central bank losses as bankruptcies in the making: the real constraint is political, not accounting. The 2026 thread assembles the whole: a balance sheet gliding towards €5,800 billion, a still-thick liquidity cushion that allows a drama-free QT, Target2 balances receding along with it, and a June rate hike reminding everyone that the floor is steered independently of the balance sheet. The same discreet infrastructure carries the other European files of the moment, from the Russian cash redeposited with central banks to the conditional net stretched beneath sovereign debts. For the terms used here, the glossary collects the definitions. --- Main sources: - ECB, consolidated weekly financial statement of the Eurosystem: the weekly balance sheet, European counterpart of the H.4.1. - ECB, 2025 annual accounts (26 February 2026): balance sheet size at the end of 2025, APP and PEPP portfolios. - ECB, Economic Bulletin, liquidity conditions from February to May 2026: excess liquidity (€2,358 billion), banknotes, minimum reserves, recourse to refinancing. - ECB, monetary policy decision of 11 June 2026: 25 basis point hike, rates at 2.25 / 2.40 / 2.65%, continued non-reinvestment. - ECB, speech "Striking the right balance" (18 February 2025): operational framework, role of the deposit rate, 15 basis point spread. - ECB, explainer "What are TARGET balances?": mechanics of the balances, link with QE and capital flows. - Bundesbank, Target balances: German claim of €1,065.7 billion as of 31 May 2026, historical series. - Bundesbank, 2025 annual accounts: €8.6 billion loss, €27.8 billion accumulated carry-forward, €363 billion net equity. - Banque de France, 2025 results: €8.1 billion profit, €11 billion exceptional gain on gold, €283 billion net equity. - PIIE, "How much money have central banks really lost?": loss comparisons and accounting conventions, the ECB's negative equity. - CPRAM, "The end of Quantitative Tightening is not on the agenda": 2026 redemptions (€330 billion of APP, €173 billion of PEPP), balance sheet trajectory. - Bruegel, "Finding the right balance (sheet): quantitative tightening in the euro area": framing of European QT. ============================================================================ REFERENCE GUIDE: Reading European sovereign debt: OATs, Bunds, BTPs and the spread URL: https://l0g.fr/en/guides/read-european-sovereign-debt/ Canonical French source: https://l0g.fr/guides/lire-la-dette-souveraine-europeenne/ Date: 2026-07-19 (reviewed 2026-07-19) ---------------------------------------------------------------------------- import InfographicBars from '../../components/infographic/InfographicBars.astro'; In the United States, a single issuer carries the federal debt, and the Treasuries market sets a rate without rival. In the euro area, nineteen countries share one currency but each issues its own debt, and the yield gap between them, the spread, becomes the true measuring instrument. This guide covers the European sovereign debt market end to end: its singular architecture, how to read the spread, the 2026 reversal of hierarchy, the safety net stretched by the ECB and the beginnings of a common safe asset. The permanent counterpoint is the Treasuries market, whose unity illuminates, by contrast, Europe's fragmentation. One currency, some twenty debtors The founding trait fits in one sentence: the euro area pooled its currency without pooling its debt. Each state borrows on its own account, with its own agency, calendar and signature. France issues OATs through Agence France Trésor, Germany issues Bunds through the Finanzagentur, Italy issues BTPs through the Rome Treasury. All denominate their debt in the same currency, but an investor never lends to the euro area: they lend to a specific country, with its own credit risk. This architecture creates a tension that national bond markets never face. A state issuing in its own currency can always, as a last resort, have its central bank print: nominal default is a choice, not a fate. A euro area member, by contrast, borrows in a currency it does not control, somewhat as if it borrowed in foreign currency. The European Central Bank serves nineteen states, not one, and has no mandate to guarantee the solvency of each. From this asymmetry flows the entire mechanics of the European market: since the risk of non-repayment becomes real again, it must be measured, and the price of that measurement is called the spread. The instruments: OAT, Bund, BTP, Gilt Four big names structure the market, and they are worth telling apart. The ten-year German Bund is the benchmark, the asset reputed safest in the area: it is the de facto risk-free rate, the one everything else is compared against. The French OAT and the Italian BTP are the other two markets of systemic size, France and Italy each carrying more than €2.5 trillion of negotiable debt. Alongside them, the British Gilt belongs to a different regime: the United Kingdom sits outside the euro area, issues in its own currency and therefore keeps its central bank as lender of last resort, which changes the nature of its risk, as we showed in our analysis of Gilt leverage and the Bank of England. The issuance mechanics resemble those of Treasuries: announced calendars, regular auctions, designated market makers, a deep secondary market. Agence France Trésor has programmed €310 billion of medium- and long-term issuance net of buybacks for 2026, a considerable volume that illustrates the weight of refinancing. But where the US Treasury sets a single price for federal debt, every European agency is permanently compared with its neighbours. Reading an auction therefore does not stop at the bid-to-cover or the tail: it factors in the level of the spread at the moment of issuance, which says at what premium the country funds itself relative to Germany. The spread, the central thermometer The spread is to European debt what the term premium is to Treasuries: the variable that condenses the information. It measures the yield gap between a country's debt and the German Bund of the same maturity, expressed in basis points. An OAT-Bund spread of 80 points means France borrows at ten years 0.80 percentage point more expensively than Germany. That premium prices a credit risk: probability of default, fiscal trajectory, political stability. Two things the spread mixes together must not be confused. In normal conditions, it captures a fiscal risk premium: the market demands more to lend to a state that is more indebted or less governable. In acute crises, as in 2011-2012, it can tip into redenomination risk, the fear that a country leaves the euro and repays in a devalued currency. The two mechanisms signal themselves differently, and telling them apart is decisive: we showed why the rise in French yields in 2026 belonged to the first and not the second in our piece on French rates and the Frexit myth. Mistaking an ordinary fiscal premium for the pricing of a monetary break-up is the most widespread reading error on this market. The great reversal of 2026 The year 2026 delivers a lesson no textbook would have dared to write ten years ago: the hierarchy of European sovereign risk has inverted. During the debt crisis, Italy embodied the weak link and France sat in the safe core, alongside Germany. The relationship has flipped. The Italian BTP-Bund spread, which had climbed to 251 points in September 2022, fell back to around 60 to 75 points in early 2026, a fifteen-year low, rewarded by seven rating upgrades in 2025, including a move from Baa3 to Baa2 at Moody's, a promotion the agency had not granted Italy in twenty-three years. Over the same period, France slid the other way. Its ten-year OAT reached 3.94% in the spring of 2026, its highest since 2009, and the OAT-Bund spread widened to around 80 points, including a political risk premium estimated at 20 to 25 points over fair value. The three major agencies downgraded the French signature within twelve months, joined by KBRA, which cut the rating to AA-. French state interest costs are set to reach €59.3 billion in 2026, against €36.2 billion in 2020: the cost of the slide reads directly in the budget. To interpret these rating moves, our guide on reading a credit rating lays out the agencies' grid. The ECB's safety net: the TPI Behind every European sovereign debt market stands an implicit question: how far would the ECB let a spread widen? Since the "whatever it takes" of 2012, the market knows the central bank holds a net. That net now has a name, the Transmission Protection Instrument, or TPI, announced on 21 July 2022. Its principle: the ECB may buy, on the secondary market, the debt of a country whose financing conditions deteriorate in an "unwarranted" way, that is, out of line with its fundamentals, when that deterioration threatens the uniform transmission of monetary policy across the area. The subtlety lies in the conditionality. The TPI is no blank cheque: a country is eligible only if it complies with the EU fiscal framework, is not subject to an excessive imbalance procedure, has debt judged sustainable and pursues sound macroeconomic policies. These criteria create a useful paradox: the net deploys only for countries that do not really need it, and closes on those whose fiscal drift is the cause of the spread. That lock is nonetheless its deterrent strength: its mere existence has long been enough to contain the gaps, without ever having been activated. The day a market seriously tests a large issuer, the ambiguity of this conditionality will become the real subject. Common debt: the embryo of a safe asset Since 2021, a new issuer has slipped into the landscape: the European Union itself. To finance the post-pandemic recovery plan, NextGenerationEU, the Commission received a mandate to borrow up to €750 billion on the markets on behalf of the 27 member states. By mid-2026, the EU's outstanding debt stood near €827 billion, and the Commission should approach €1 trillion by year-end, ranking it among the continent's largest supranational issuers. Is this the birth of the European safe asset the area lacks, a credible counterpart to the US Treasury? The answer remains cautious. These bonds do attract central banks, pension funds and sovereign wealth funds hunting for high-grade euro-denominated paper. But they carry no unlimited joint and several guarantee like a true federal debt: each state remains liable for its share, the programmes are temporary and capped, and index providers classify them as "supranational" debt, alongside the European Investment Bank, rather than as full sovereign paper. Europe has thus built a quasi-safe asset without genuine mutualisation, a compromise that reflects the political balance of the moment rather than a completed fiscal union. We devote a dedicated analysis to this common debt that appears in no ratio, and to its repayment starting in 2028. The same logic of discreet but systemic European infrastructure runs through our investigation into Euroclear and the immobilised Russian assets. The German brake comes off The last shift of 2026 comes from the heart of the system. Germany, historic guardian of fiscal orthodoxy, reformed its constitutional debt brake, the Schuldenbremse, on 21 March 2025. The reform exempts defence spending above 1% of GDP and creates a special infrastructure fund of more than €500 billion. The consequence is mechanical: Berlin moves from frugal management to massive borrowing, with a 2026 funding requirement of about €174 billion, more than triple that of two years earlier, and a defence budget above €100 billion. The market effect is twofold, and it touches everyone. On one side, a far more abundant supply of Bunds lifts the area's benchmark rate: when the safest asset yields more, the whole European curve adjusts. On the other, that same supply could, in time, deepen the Bund market and strengthen its safe-haven status. The paradox is that the euro area's anchor of stability is itself becoming a heavy borrower, reshuffling relative scarcity: the Bund is no longer just the low point of the curve, it becomes a supply engine too. For the general mechanics of debt supply and the term premium, the parallel with the Treasuries market remains the best point of comparison. The weight of debt, country by country The spread does not say everything: it must be read alongside the stock of debt relative to GDP, which sets the long-run trajectory. At the end of 2025, the euro area posted public debt of 87.8% of GDP according to Eurostat, but the dispersion is the real story. Greece peaks at 146%, Italy at 137%, France at 116%, Belgium at 108%, Spain at 101%, while Germany stays below two-thirds of GDP. This map explains why the market now watches each country's speed of drift more than the old North versus South divide: Greece is deleveraging, France is slipping, and the ranking is being reordered. Reading the European market in practice Read properly, the European sovereign debt market is not a juxtaposition of national curves but a three-level system. The first level is the architecture: a single currency, some twenty issuers, hence a relative credit risk that does not exist for a classic sovereign issuer. The second is the spread, which measures that risk day by day, provided one distinguishes the ordinary fiscal premium from crisis-time redenomination risk. The third is the net and its limits: the ECB's TPI, deterrent but conditional, and the EU's common debt, a safe asset in the making but without full mutualisation. The thread of 2026 ties the three levels together: France becomes the point of tension while Italy normalises, Germany abandons its frugality and swells the supply of Bunds, and Europe inches towards a common safe asset without taking the federal leap. For the terms used here, the glossary collects the definitions, and the spread logic extends to corporate and private credit in our guide on credit spreads. The other half of the European machinery, the Eurosystem balance sheet and the Target2 balances, is covered in our guide on the ECB balance sheet. --- Main sources: - Eurostat, euro area government debt in Q4 2025: euro area debt-to-GDP ratio (87.8%) and country figures (Greece, Italy, France, Belgium, Spain). - European Central Bank, announcement of the Transmission Protection Instrument (TPI), 21 July 2022: purpose, eligibility criteria and activation conditions. - Deutsche Bundesbank, note on the TPI: how the instrument works and the notion of fragmentation. - Agence France Trésor, indicative state financing programme for 2026: €310 billion of net medium- and long-term issuance, OAT issuance strategy. - European Commission, the EU as a borrower (NextGenerationEU): outstanding common debt, supranational status, absence of unlimited joint guarantee. - Bruegel, "What does German debt brake reform mean for Europe?": the debt brake reform of 21 March 2025, infrastructure fund and defence exemption. - Il Sole 24 Ore, BTP-Bund spread and ten-year BTP yield: level of the Italian spread in 2026. - Ideal Investisseur, OAT-Bund spread: level of the French spread, Banque de France and Bundesbank data. - MNI Markets, French 2026 budget deficit: deficit target and France's debt trajectory. - Morningstar, EU bonds and the European safe asset: treatment of EU bonds as supranational debt, outstandings heading towards €1 trillion. ============================================================================ REFERENCE GUIDE: Reading a credit rating: scales, default and the issuer-pays conflict URL: https://l0g.fr/en/guides/read-credit-ratings/ Canonical French source: https://l0g.fr/guides/lire-une-notation-de-credit/ Date: 2026-07-16 (reviewed 2026-07-16) ---------------------------------------------------------------------------- A credit rating looks like a school grade, and that is the first trap. It does not say whether a bond is a good investment, nor whether its price is fair. It says one thing, with deliberately limited precision: how likely the borrower is not to repay. This opinion, produced by a handful of agencies paid by the very issuers they rate, governs trillions of regulated allocation. Understanding it means knowing what it measures, what it is worth, and above all what it does not say. This guide lays out the scales, the default rates, the model's conflict and the fracture between public and private ratings, as an extension of our guide on credit spreads. What a rating measures, and what it is not A credit rating is an opinion on a borrower's ability and willingness to service its debt, summed up in a symbol. It bears on default risk, sometimes complemented by an estimate of the loss given default. It is neither a price, nor a buy recommendation, nor a guarantee. A bond rated AAA can lose 30% of its value if rates rise, without any default occurring: the rating says nothing about market risk, only about credit risk. This distinction is the source of the costliest misunderstandings. In 2008, securitisation tranches rated AAA collapsed, not because the agencies had lied about default risk in the strict sense, but because the market had taken the rating for a guarantee of overall safety. Reading a rating means first bringing it back to its exact perimeter: a probability of default, over a given horizon, as of the date of the analysis. The scales: AAA down to D, and the decisive frontier The three big agencies use parallel scales. S&P and Fitch share the same alphabet: AAA, then AA, A, BBB, and so on down to D for default, with intermediate notches marked by a plus or a minus. Moody's uses a distinct notation: Aaa, Aa, A, Baa, with numeric modifiers 1, 2, 3. The frontier that matters is not at the top of the scale, but in its middle. An issuer is investment grade as long as it is rated BBB- or better at S&P and Fitch, Baa3 or better at Moody's. Below begins high yield, also called speculative or "junk". This line is not gradual: crossing it downward, becoming a "fallen angel", suddenly excludes the issuer from many regulated portfolios constrained to investment grade, forcing sales and widening its spread. A notch around this frontier therefore weighs far more than a notch elsewhere on the scale. Default rates: what the rating actually predicts A rating is only worth anything if it predicts. History shows it does, in broad strokes. The one-year default probability climbs sharply as the rating falls: close to zero for a AAA, on the order of 0.2% for a BBB, 0.6% for a BB, 3% for a B, and up to 26% for a CCC or below. Cumulatively over five years, the gap widens further: less than 2% defaults for investment grade, more than 20% for an issuer rated B. This non-linear profile is a rating's real message. The gap between AAA and BBB, both investment grade, is modest. The gap between BB and CCC, both speculative, is dizzying. The rating does not measure a risk proportional to the position on the scale: it measures a risk that explodes at the bottom. That is why a portfolio can easily absorb a few BBBs but comes apart if it accumulates CCCs. To grasp the dynamics rather than the snapshot, analysts use the transition matrix: a table that gives, for each rating, the probability of migrating to another over one year, up, down or into default. It is what reveals the relative stability of high ratings and the volatility of low ones, and what feeds credit-risk models. The conflict of the issuer-pays model That leaves the question that undermines the credibility of the whole: who pays for the rating? The three approved agencies, the NRSROs in the SEC's sense, hold about 95% of the world market. Their dominant model is issuer-pays: the company or government issuing the debt pays the agency to be rated. The agency's customer is therefore the very entity it is supposed to judge without indulgence. This conflict is not theoretical. It pushes toward a bias for favourable ratings to keep the client, and it was identified as one of the causes of the 2008 crisis. It persists because the alternative, an investor-pays model, runs into the fact that a rating, once published, benefits everyone without anyone wanting to fund it. Reading a rating therefore means keeping in mind that its producer is paid by its subject, and cross-checking the rating with an independent signal, the market spread, which often moves faster and more honestly than the rating itself. Public versus private rating: the grey zone The most current fracture is not between agencies, but between public and private ratings. A public rating is disseminated, tracked, revised under the market's gaze. A private rating, or "private letter rating", is disclosed only to the subscriber, often to let it fit an asset into a favourable regulatory box. It is the mechanism of the rated feeder notes that bring private credit onto insurers' balance sheets, described in our guide on a life insurer's soundness. The private channel is not neutral. The research relayed by the specialist press shows that a security moving to a private rating is upgraded more than four times as often as it is downgraded, while the same move to a public rating produces as many upgrades as downgrades. The same drift reads elsewhere: AI-chip-backed debt reached investment grade on structures where the rating does much of the work. A favourable private rating is not false by nature, but its lack of public challenge makes it a signal to treat with caution. The reading grid Five reflexes sum up the guide. Bring every rating back to its perimeter: a probability of default, not a guarantee or a price. Locate the issuer relative to the BBB- / Baa3 frontier, where a single notch changes everything. Read the rating in light of the historical default rate of its category, keeping in mind the run-away at the bottom of the scale. Remember who pays for the rating, and cross-check it with the market spread, quicker to punish. And tell a public rating from a private one, weighting the latter downward. A credit rating is a valuable tool and a biased tool, both at once. It condenses a mass of analysis into a readable symbol, but it is produced by an actor paid by its subject, revised with a lag, and increasingly manufactured in a private channel that escapes scrutiny. To read it correctly is not to believe it or reject it, but to know exactly how far it carries, and to complete with the price what it does not say. Sources - Wolf Street, "Corporate Bond Credit Ratings Scales: Moody's, S&P, Fitch" (parallel scales, notch correspondence, investment grade / high yield frontier): https://wolfstreet.com/credit-rating-scales-by-moodys-sp-and-fitch/ - Fidelity, "Bond Ratings" (investment grade at BBB- / Baa3, definition of high yield): https://www.fidelity.com/learning-center/investment-products/fixed-income-bonds/bond-ratings - Wikipedia, "Nationally recognized statistical rating organization" (S&P, Moody's, Fitch ~95% of the market, SEC approval): https://en.wikipedia.org/wiki/Nationallyrecognizedstatisticalratingorganization - U.S. GAO, "Credit Rating Agencies: Alternative Compensation Models for NRSROs" (issuer-pays model and its conflict of interest): https://www.gao.gov/products/gao-12-240 - S&P Global / market summaries, default rates by rating (one-year probabilities: ~0% AAA, 0.2% BBB, 0.6% BB, 3% B, 26% CCC/C; cumulative 5-year < 2% in IG, > 20% in B): https://investmentgrade.com/bond-ratings/ - RapidRatings, "Rating Transition Matrix" (transition matrix, rating migration probabilities): https://help.rapidratings.com/hc/en-us/articles/360045797832-Rating-Transition-Matrix - Alternative Credit Investor, "Insurers and private credit: Ratings under the microscope" (asymmetry of revisions in private ratings): https://alternativecreditinvestor.com/2025/12/04/ratings-under-the-microscope/ This guide is educational analysis and does not constitute investment advice. Default rates are historical orders of magnitude, cited as of the date of their sources. ============================================================================ REFERENCE GUIDE: Reading the gas and LNG market: the three benchmarks and inter-basin arbitrage URL: https://l0g.fr/en/guides/read-gas-lng-market/ Canonical French source: https://l0g.fr/guides/lire-le-marche-du-gaz-et-du-gnl/ Date: 2026-07-16 (reviewed 2026-07-16) ---------------------------------------------------------------------------- Gas reads as the mirror image of oil. Oil has a world price, give or take a few quality differentials, because a barrel travels easily anywhere. Gas stays long a prisoner of its pipes: without a pipeline, it must be liquefied to be moved, a costly operation that fragments the market into three great regions with distinct prices. To understand gas is first to accept that it has no single price, but three, and that the essential plays out in the spread between them. This guide lays out the benchmarks, the LNG chain, the units, the arbitrage and the signals, with the 2026 Hormuz crisis as the thread, alongside our oil market guide. Three benchmarks, not one price The world gas market is organised around three references. Henry Hub is a physical delivery point in Louisiana, the benchmark for US gas and the base of most US LNG export contracts. TTF, the Title Transfer Facility, is a Dutch virtual trading point that has become Europe's reference, quoted in euros per megawatt-hour. JKM, the Japan-Korea Marker published by S&P Global Platts, is the benchmark for spot LNG delivered to Northeast Asia. These three prices do not align, and their hierarchy has a logic. Henry Hub is structurally the cheapest, because the United States produces abundant gas it long struggled to export. TTF and JKM trade higher, because they embed the cost of bringing liquefied gas to them. According to the LNG association, TTF and JKM sustain a premium of $5 to $10 per MMBtu over Henry Hub, precisely to cover liquefaction, shipping and the relative scarcity of supply in Asia and Europe. The LNG chain: what connects the basins What turns three siloed markets into a connected system is liquefied natural gas, LNG. The principle has three steps. The gas is cooled to about −162°C to liquefy it, which cuts its volume by more than six hundred and makes it transportable by carrier. It is shipped by sea toward the best-paying basin. It is regasified on arrival, at an import terminal, to inject it into the grid. According to the LNG association, world trade reached about 400 million tonnes in 2024. This chain has an energy cost, often forgotten. Liquefying, shipping and regasifying consume between 10 and 15% of the gas's initial energy content. It is this cost, added to freight and margin, that explains the premium of the import benchmarks over Henry Hub. It also explains why LNG does not redirect instantly: a carrier takes weeks to switch basins, and liquefaction plants are built over years. LNG's flexibility is real but slow, which lets price spreads persist longer than in oil. The units, a trap to avoid Comparing the three benchmarks requires handling their units, and that is where errors creep in. Henry Hub and JKM are quoted in dollars per MMBtu, the million British thermal units. TTF is quoted in euros per megawatt-hour. Yet a MWh is worth about 3.412 MMBtu, and the euro must still be converted to the dollar. A TTF at 40 euros a MWh therefore does not compare directly to a JKM at 13 dollars a MMBtu without going through these conversions. Many hasty comparisons conclude to an arbitrage that does not exist, for lack of aligning the units. The rule is simple: bring the three prices to dollars per MMBtu before any comparison. The arbitrage: who captures the cargo Once the prices are aligned, the mechanics become clear. A cargo of LNG not committed under a long-term contract goes to the basin that pays it best, net of costs. Concretely, if Asian JKM exceeds European TTF by more than the shipping cost and the boil-off en route, cargoes divert from Europe to Asia, and vice versa. The TTF-JKM spread is thus the market's compass: it says not only which basin is more expensive, but where the next ships will go. This arbitrage has a political consequence. In a shock, Europe and Asia compete for the same cargoes, and the marginal price rises for whoever has to outbid. That is what happened during the 2026 Hormuz crisis, when the suspension of part of Qatari production removed at a stroke nearly a fifth of world LNG supply, sending TTF above 60 euros a megawatt-hour in a single session. Record US export volumes only partly offset it, because LNG often sets the marginal price in Europe. Storage and chokepoints: the shock absorbers Two variables cushion or amplify these shocks. Storage, first, especially European: the fill levels of gas reserves before winter determine Europe's margin against a rupture. Full storage can absorb a cut for a few weeks; low storage turns the slightest incident into a spike. Following the fill rate of European storage is therefore one of the most predictive signals for TTF. Maritime chokepoints, next. A major share of world LNG depends on a few narrow passages, first among them the Strait of Hormuz, through which almost all Qatari exports transit. We documented in the Hormuz supply chain how a disruption of this corridor, aggravated by the absence of an alternative route for gas unlike oil, transmits directly to TTF and JKM. Gas is more vulnerable than oil to these chokepoints, because it has neither a bypass pipeline of the right scale nor a spot market as fluid. The signals to watch Five needles concentrate the information on this market. The three benchmarks brought to a single unit, dollars per MMBtu, to read the hierarchy and its distortion. The TTF-JKM spread, to anticipate the direction of cargoes. The fill rate of European storage, a barometer of resilience to a shock. US export capacity and the feedgas of liquefaction plants, which cap flexible supply. And the state of maritime chokepoints, Hormuz first, on which access to Gulf gas depends. Gas is a market of pipes and ships before it is a market of prices. Its fragmentation into three basins, the slowness of its arbitrage and its dependence on a few narrow passages make it an asset where geography counts as much as supply and demand. Reading gas means following not a figure but a spread, and remembering that what costs the most is not the gas, but the voyage it must make to reach where it is scarce. Sources - LNG Allies (liquefiednaturalgas.org), "LNG Pricing & Market Benchmarks: Henry Hub, TTF, JKM Explained" (role of the three benchmarks, $5-10/MMBtu TTF/JKM premium over Henry Hub): https://liquefiednaturalgas.org/market/pricing/ - LNG Allies, "LNG Market" (liquefaction-shipping-regasification chain, world trade ~400 Mt in 2024, energy cost of 10 to 15%): https://liquefiednaturalgas.org/market/ - ICE, "Natural gas benchmarks: a new landscape" (globalisation of gas, liquidity and hedging of the benchmarks): https://www.ice.com/insights/market-pulse/lng-trading-liquidity-hedging-a-new-landscape-for-natural-gas-benchmarks - LNGPriceIndex, JKM, TTF and Henry Hub quotes (indicative levels and units): https://lngpriceindex.com/lng-benchmark - Natural Gas Intelligence, "U.S. LNG Profits Exposed as Market Again Shifts, Global Natural Gas Prices Converge" (convergence and inter-basin arbitrage): https://naturalgasintel.com/news/us-lng-profits-exposed-as-market-again-shifts-global-natural-gas-prices-converge/ This guide is educational analysis and does not constitute investment advice. Price levels are indicative and cited as of the date of their sources. ============================================================================ REFERENCE GUIDE: Reading a life insurer's soundness: capital, reinsurance and opaque assets URL: https://l0g.fr/en/guides/read-life-insurer-health/ Canonical French source: https://l0g.fr/guides/lire-la-solidite-d-un-assureur-vie/ Date: 2026-07-16 (reviewed 2026-07-16) ---------------------------------------------------------------------------- A life insurer promises to pay in ten, twenty or forty years. Its soundness is therefore not judged by its quarterly profit, but by the certainty that the assets it holds today will still be worth, when the time comes, what its commitments are worth. That certainty has blurred. Since a growing share of American annuities is backed by private credit, housed in Bermuda-affiliated reinsurers and rated by private agencies, reading an insurer means looking beyond the headline capital ratio. This guide lays out the dials to watch, from RBC to the liability, with 777 Re as the thread, and extends our article on retirement savings in private credit. Regulatory capital: the RBC ratio and its levels The first measure is capital, and its yardstick in the United States is RBC, for Risk-Based Capital. It is the insurance equivalent of bank CET1: a minimum capital calculated not on the gross size of the balance sheet, but on the risk of the assets and commitments. A portfolio loaded with equities or speculative credit requires more capital than a portfolio of government bonds. The RBC ratio divides the insurer's total adjusted capital by the authorized control level. What matters is less the absolute figure than the threshold it crosses. According to the NAIC, the association of state insurance regulators, regulatory actions kick in by stages: above 250%, no intervention; between 200 and 250% with a failed trend test, or between 150 and 200%, the insurer must submit a recovery plan; between 100 and 150%, a corrective plan is required; between 70 and 100%, the regulator is authorized to place the insurer under control, up to liquidation. One caveat applies, which the NAIC underlines itself: RBC is a detection tool, not a league table. An insurer at 600% is not mechanically sounder than one at 400%. Above the thresholds, the ratio stops discriminating; it becomes informative only near the levels. Reading RBC therefore means watching the distance to the threshold, not admiring a big number. The statutory balance sheet: where the risk hides RBC is only worth as much as the quality of the assets it weights, and that is where reading becomes technical. The American insurer publishes statutory accounts, distinct from GAAP, of which two schedules concentrate most of the information. Schedule D lists bonds and equities, the classic core of the portfolio. Schedule BA, titled "other long-term invested assets", houses the rest: private credit funds, holdings, joint ventures, and above all affiliated and alternative assets. It is Schedule BA that should be opened first when looking for displaced risk. Two signals read there. The share of illiquid assets, first, harder to sell under stress. The IMF, in its April 2026 financial stability report, notes that insurers backed by private-equity firms hold nearly twice as many illiquid assets as the others. The share of affiliated assets, next: when an insurer invests in the own funds of the group that controls it, the valuation and soundness of those assets become circular. It is this concentration in affiliated assets that precipitated the fall of 777 Re. NAIC designations and private ratings Each security in the portfolio receives an NAIC designation, from 1 to 6, that sets its capital treatment: 1 and 2 correspond roughly to investment grade, 3 to 6 to rising risk up to default. The better the designation, the less capital the insurer must set aside. The stake is therefore to fit risky assets into the best boxes. That is the function of rated feeder notes: a feeder vehicle invests in a private credit fund and issues debt securities carrying a rating, most often a private one, disclosed only to the subscriber. The economic exposure is that of a fund interest; the capital treatment is that of a rated bond. The problem is the quality of those grades. The research relayed by the specialist press shows that, when a security moves from the in-house assessment of the NAIC's securities valuation office to a private rating, it is upgraded more than four times as often as it is downgraded, while the same move to a public rating produces as many upgrades as downgrades. The choice of rating channel then looks like an optimisation of capital, more than a measure of risk. To judge what a rating is worth, our guide dedicated to reading a credit rating details the difference between a public and a private grade. Reinsurance: reading the leverage and the destination The third dial is the most specific to insurance. An insurer can cede part of its commitments to a reinsurer, which takes them on in exchange for a commission and reinvests the corresponding assets. This asset-intensive reinsurance is legitimate in itself, but two parameters change its reach: the reinsurance leverage, the share of commitments ceded relative to capital, and the destination of the cession. The destination has become the subject. According to Bloomberg's investigation of the sector, US life insurers ceded $2.4 trillion of reserves in 2024, of which more than $1.1 trillion went to offshore jurisdictions, Bermuda first, where the prudential and accounting regime is more accommodating. When the Bermuda reinsurer belongs to the same group as the ceding insurer, the cession does not really transfer the risk: it moves it onto a less-regulated balance sheet, under the same shareholder's control. Reading an insurer therefore means checking not only how much it cedes, but to whom. The liability: the long-liability test The first four dials are on the asset side; the fifth is on the liability side, and it is the model's defence. An annuity is not a bank deposit: the liability is long, predictable, and early surrenders are curbed by contractual and tax penalties. A holder of twenty-year commitments is therefore, in principle, best placed to carry illiquid assets, far better than a semi-liquid fund open to quarterly redemptions, whose repeated gating we have documented. The strength of that argument depends on one condition: that the liability stays genuinely long. And it does so only as long as surrenders remain discouraged. A sharp rise in rates, which makes old annuities uncompetitive against new ones, can accelerate exits at the precise moment illiquid assets are hardest to sell. The IMF described this run scenario as early as its 2023 work on private equity and life insurers. Reading the liability therefore means estimating the sensitivity of surrenders to a rate shock, and comparing this potential liability liquidity with the real liquidity of the assets. 777 Re, the case study All these dials read together in a concrete case. 777 Re, the Bermuda reinsurer of the 777 Partners group, had accumulated on its balance sheet affiliated assets, invested in its shareholder's own businesses. On 8 October 2024, the Bermuda Monetary Authority cancelled its registration, finding an excess of affiliated assets, deficient governance and insufficient capital contributions. Upstream, the American insurer A-CAP, which had ceded $1.7 billion of reserves to it, saw its rating cut by AM Best in February 2024, the agency citing high reinsurance leverage and the deteriorating quality of its counterparties. The pattern condenses the guide's signals: reinsurance ceded offshore, to an affiliate, loaded with illiquid assets tied to the shareholder, discovered by the regulator from the end of the chain. The difference between 777 Re and the sector's large players is one of scale and asset quality, not of structure. That is why reading these dials one by one, on a sound insurer as on a fragile one, is the only way to tell a robust model from one that has simply been lucky. The reading grid Five questions sum up the guide. Is the RBC ratio at a comfortable distance from its thresholds, and how is it moving? Does Schedule BA reveal a high share of illiquid or affiliated assets? Are the ratings that carry the capital public or private, and does the private channel dominate? Is the reinsurance leverage high, and does the cession go to an offshore affiliate? Would the liability withstand a rate shock, or would surrenders accelerate? None of these dials is enough on its own, and none is visible in the financial-strength rating the insurer puts forward. A life insurer's soundness is not read in a single figure, but in the coherence between its capital, its assets, its ratings, its reinsurance and its liability. The day one of the five gives way, the other four reveal at once what they were hiding, and the story of 777 Re starts again, at another scale. Sources - NAIC, "Risk-Based Capital" (RBC definition, regulatory action levels, caveat on comparing high ratios): https://content.naic.org/insurance-topics/risk-based-capital - NAIC, Schedule BA and the bond-definition project (reclassification of assets not qualifying as bonds, share of affiliated assets in "other long-term invested assets"): https://content.naic.org/insurance-topics/private-credit - IMF, Global Financial Stability Report, April 2026 (private credit ~35% of North American insurers' portfolios; private-equity-backed insurers ~2x more illiquid assets): https://www.imf.org/en/publications/gfsr/issues/2026/04/14/global-financial-stability-report-april-2026 - Bloomberg, "Apollo and Wall Street Private Equity Firms Bet on America's Life Insurance" ($2.4tn of reserves ceded in 2024, of which more than $1.1tn offshore): https://www.bloomberg.com/graphics/2025-america-insurance-part-1/ - American Academy of Actuaries, "Asset-Intensive Reinsurance Ceded Offshore" (reinsurance leverage, affiliated assets, risks of offshore cession): https://actuary.org/wp-content/uploads/2024/02/risk-brief-bermuda-reinsurance0.pdf - Alternative Credit Investor, "Insurers and private credit: Ratings under the microscope" (asymmetry of revisions: more than four upgrades per downgrade on moving to a private rating), December 2025: https://alternativecreditinvestor.com/2025/12/04/ratings-under-the-microscope/ - Bermuda Monetary Authority, notice of cancellation of 777 Re Ltd's registration, 8 October 2024: https://www.bma.bm/viewPDF/documents/2024-10-08-12-44-33-Notice---Cancellation-of-Registration---777-Re-Ltd.pdf - Retirement Income Journal, "Double Trouble in the Bermuda Triangle" (A-CAP: $1.7bn of reserves ceded to 777 Re, rating cut by AM Best): https://retirementincomejournal.com/article/double-trouble-in-the-bermuda-triangle/ This guide is educational analysis and does not constitute investment advice. Regulatory data is cited as of the date of its sources. ============================================================================ REFERENCE GUIDE: Reading a bank stress test: DFAST, CCAR and the stress buffer URL: https://l0g.fr/en/guides/read-bank-stress-tests/ Canonical French source: https://l0g.fr/guides/lire-un-stress-test-bancaire/ Date: 2026-07-13 (reviewed 2026-07-13) ---------------------------------------------------------------------------- Once a year, the Federal Reserve inflicts an imaginary apocalypse on the largest U.S. banks, a stock-market collapse, a jobless spike, a property crash, then publishes the verdict: who survives, and with how much capital. The stress test is the closest thing to an X-ray of bank soundness, and it concretely shapes how much each bank can return to shareholders. It is also a contested ritual, at once illuminating and misleading, because it measures one danger well and almost entirely ignores the one that, in reality, brings banks down. This guide explains how to read it, what it says and what it leaves out. The 2026 test serves as the illustration. DFAST and CCAR: two names, one exercise The exercise carries two acronyms often confused, which in fact answer two distinct questions. DFAST, the Dodd-Frank Act Stress Test, is the quantitative part, born of the 2010 law: it asks about survival. Would the bank keep its capital above regulatory minimums if a severe recession struck? CCAR, the Comprehensive Capital Analysis and Review, is the capital-planning part: it asks about distribution. Can the bank pay dividends and buy back its shares while staying sound under stress? The first tests resilience, the second draws the consequences for capital policy. The test covers only large banks, those above $100 billion in assets, sorted into categories by size and complexity. In 2026, thirty-two firms were subject to it. Smaller banks, such as the regionals, are exempt, a point that is not trivial, as we will see, because it is precisely there that some recent flaws have lodged, described in our guide to reading a bank's soundness. The scenario: a tailor-made apocalypse The heart of the test is the "severely adverse" scenario, a chain of catastrophes the Fed designs itself each year and publishes in February. It does not claim to predict the next crisis, but to subject banks to a shock calibrated to hurt. The 2026 one illustrates the exercise well. A baseline scenario, milder, serves as a comparison, and the Fed sometimes adds an exploratory element to probe an emerging risk. But it is the severely adverse scenario that makes the headline, because it is the one that sets capital. The fact that banks know it in advance is both a strength, transparency, and a weakness, the incentive to optimize to it, to which we will return. The mechanics: losses, revenues, capital Once the scenario is set, the Fed simulates, quarter by quarter over nine quarters, what it would do to each bank. On one side, it projects losses: loan defaults, trading write-downs, operational losses. On the other, it projects the revenue the bank would keep earning despite the crisis, pre-provision net revenue or PPNR, the first cushion absorbing losses. The difference, applied to equity, traces the path of the CET1 ratio throughout the shock. The 2026 result is reassuring on paper, as it is almost every year: the thirty-two banks pass, the aggregate ratio cedes only 1.6 points and stays far above the floor. But reading a stress test does not stop at that "pass." The real lesson lies elsewhere, in the figure the test produces for each bank that will constrain it all year. The real point: the Stress Capital Buffer The test's most important output is not the binary verdict, it is the Stress Capital Buffer. Its calculation is simple in principle: take the starting CET1 ratio, subtract the lowest point reached during the nine quarters of stress, and add the dividends planned over a year, with a 2.5% floor. The result becomes a firm-specific capital requirement, added to the bank's minimums. The stakes are large, because this buffer sets the bank's room to manoeuvre. The more a bank suffers in the scenario, the higher its buffer, and the less capital it has left to distribute in dividends and buybacks. The stress test is therefore no consequence-free exam: it fixes, bank by bank, how much capital each must retain rather than return. It is the channel through which a poor result is paid for in fewer buybacks, which explains the attention investors give it. In 2026, notably, the Fed froze the buffer requirements while finalizing a revision of its framework, a point worth dwelling on. The limits of the exercise A careful reader never takes a stress test at face value, because its limits matter as much as its results. The first is that it models the last war. Its scenarios draw on past crises, a 2008-style recession, and nothing guarantees the next shock will resemble them. The second is that the scenario is known in advance: banks can shape their balance sheets to look good in it, which improves the score without necessarily strengthening real resilience. The third is that it is a point-in-time snapshot, quickly outdated as portfolios shift. The fourth limit is the gravest, and it goes to the heart of the matter. The test measures solvency, the ability to absorb losses, but barely liquidity, the ability to honour sudden withdrawals. Yet it was a liquidity run, not a solvency shortfall, that killed Silicon Valley Bank in 2023, a bank that was not even subject to the test. The most common way banks fail, a deposit flight at the speed of an app, is precisely the one the stress test captures least. Passing the test is therefore no blank cheque, as several voices in the sector recalled at the very moment when regulation, with the framework revision and the lightening of Basel III, was loosening the capital constraint. A clean bill of solvency says nothing about vulnerability to a run. Reading the stress test in practice In practice, a few reflexes draw the real signal from a stress test. The calendar first: scenarios appear in February, results in June, and the resulting buffer applies thereafter. The number to isolate is not the collective "pass or fail," but the CET1 drawdown bank by bank, since a large drawdown betrays a high exposure to the scenario's losses. The resulting stress buffer says how much capital the bank must retain, hence its capacity to buy back shares. The loss composition reveals where the balance sheet is fragile, commercial real estate, credit cards, trading. And pre-provision revenue measures the cushion that absorbs the shock before capital. Above all, read the test for what it is, a solvency exam under a known scenario, not a guarantee of survival. It complements the grid of our guide to bank soundness, which adds what the test neglects, liquidity, unrealized losses and deposit structure. The stress test says a bank would absorb a modeled recession. It does not say it would survive a panic of its own depositors, and it is that second question that history has shown to be the deadlier. Sources and further reading - Federal Reserve, Dodd-Frank Act Stress Test 2026 results, 24 June 2026: aggregate CET1 from 12.8% to 11.2%, about $625bn of loan losses. - Federal Reserve, 2026 stress test scenarios: equities -58%, VIX 72, unemployment 10%, commercial real estate -39%. - Bank Policy Institute, "The 2026 Federal Reserve Stress Test Results: A Framework in Transition": the framework revision and the frozen buffers. - Forbes (M. Rodriguez Valladares), "2026 Bank Stress-Test Results Are Not A Green Light For Lower Capital": the limits of the exercise. - l0g, Q2 bank earnings, reading the risk and Commercial real estate and the refinancing wall. - Related guides: Reading a bank's soundness and Reading CLOs and leveraged loans. ============================================================================ REFERENCE GUIDE: Reading the copper market: Dr. Copper, LME, tariffs and deficit URL: https://l0g.fr/en/guides/read-copper-market/ Canonical French source: https://l0g.fr/guides/lire-le-marche-du-cuivre/ Date: 2026-07-11 (reviewed 2026-07-11) ---------------------------------------------------------------------------- Copper is said to hold a PhD in economics. Present in almost everything industry makes, from electrical cable to motors, from buildings to smartphones, its price long served as a bulletin on the world's health. But the red metal has changed status: it has become the raw material of electrification, without which there is no grid, no electric car, no data center. In 2026 it trades around $13,000 a tonne, a record level, torn between a structural supply deficit and a US tariff that has split its price in two depending on whether you buy it in London or New York. This guide explains how to read this market that has turned strategic. Dr. Copper, barometer of the economy The nickname Dr. Copper comes from a simple intuition: because copper goes into almost every industrial chain, its demand tracks the economic cycle, and its price often anticipates turning points before official indicators. When global industry accelerates, copper rises; when it slows, copper falls. For a long time, following copper meant taking the pulse of growth, Chinese growth in particular. That reading still holds, but it is now blurred by a second, more structural force. The electrification of the economy is intensely copper-hungry: an electric vehicle contains about three to four times more than a combustion car, a wind or solar farm far more than a conventional plant, and the rise of data centers for artificial intelligence adds new demand for cabling and transformers. Copper is no longer just the thermometer of the cycle, it is also the bet on the energy transition. These two engines, cyclical and structural, can pull the same way or contradict each other, which complicates the reading. Where the price is set: London, New York, Shanghai The world price of copper is set first at the London Metal Exchange, the LME, the base-metals exchange. It quotes a cash price and a three-month price, and runs a network of approved warehouses whose stocks are watched as a tightness barometer: dwindling stocks signal a tight physical market. Alongside the LME, two other venues matter: New York's COMEX, the US benchmark, and Shanghai's SHFE, a mirror of Chinese demand. In normal times, these three prices move together, give or take gaps that reflect transport and taxes. This neat three-price balance shattered in 2025. The reason is not geological, it is a tariff, and it is worth pausing on, because it shows how a political decision can fracture a global market. The 2026 split: the tariff that cuts the price in two In 2025, Washington invoked Section 232 of its trade law, the one that allows tariffs on imports deemed a threat to national security, to target copper. The announcement of a coming tariff triggered a rush: traders front-loaded cargoes to the United States before it took effect, swelling COMEX stocks to a record of about 650,000 tonnes and pushing the New York price well above the London one. This split is instructive. It reminds us that copper is not a single price but a family of prices, and that a regulatory shock can misalign them durably. It also shows how a tariff, before it even fully exists, distorts physical flows: what is scarce elsewhere is stockpiled in the United States, creating apparent scarcity in London and abundance in New York. For the arbitrageur, the gap is an opportunity; for the US manufacturer, it is a surcharge. Reading copper in 2026 means first asking which price you are talking about. Supply: concentrated mines, long lead times, thinning ore Beneath the price turbulence, the underlying constraint is geological and slow. Copper mine production is concentrated in a few countries, Chile, Peru, the Democratic Republic of Congo, Indonesia, exposed to political, social and climatic hazards. Opening a new mine takes ten to twenty years from discovery to first production, which leaves supply unable to respond quickly to a demand shock. And the grade of the ore mined is trending down: more and more rock must be moved for the same amount of metal, raising costs and energy and water use. In 2026, these fragilities materialized as disruptions across several major producers, feeding the expected deficit. We described these tensions, worsened by cyclical factors such as the Strait of Hormuz and El Niño, in our article on the copper shortage. Copper supply cannot be decreed; it is built over a decade, and the world did not invest enough in the last one. The smelters' signal: TC/RCs There is a leading tightness indicator that insiders watch closely, invisible to the public: TC/RCs, treatment and refining charges. These are the sums smelters charge mines to turn copper concentrate, the crushed and enriched rock that leaves the mine, into usable refined metal. Their level tells the balance of power between mines and smelters. When concentrate is plentiful, smelters hold the upper hand and charge dearly: TC/RCs are high. When concentrate grows scarce, smelters fight to secure supply and cut their charges, sometimes to the point of working at a loss. In 2025-2026, TC/RCs fell into negative territory, an unheard-of sustained event: a sign of concentrate scarcity at the source, coupled with smelter overcapacity, notably Chinese. It is one of the most reliable signals that copper tightness truly comes from deep in the mine, not just from trading floors. Demand: China and electrification Copper demand long had one name: China, which alone absorbs more than half of the world's refined metal, mostly for its construction and infrastructure. That is why copper long tracked the Chinese property cycle. But two shifts are reshuffling the deck. On one side, China's property crisis weighs on traditional construction demand. On the other, China is investing heavily in its power grid and renewables, a green demand that partly offsets the weakness in building. Beyond China, global electrification sets a floor of structural demand. Grids to modernize, electric vehicles, wind and solar, data centers: all are voracious for copper. One analyst's caveat, so as not to slip into the tale of the inevitable super-cycle. A high price calls forth responses: substitution of copper by aluminium in some uses, more recycling, efficiency. And a severe global recession would sink cyclical demand far faster than electrification supports it. The deficit is likely, it is not guaranteed, and shortage forecasts have often been wrong in the past. Reading the copper market in practice Reading copper means combining several dials. The LME price and curve give the world reference, and the slope, contango or backwardation, says whether the market is abundant or tight in the near term. Warehouse stocks at the LME, COMEX and SHFE measure physical availability. The COMEX-LME gap, now, reveals the effect of the US tariff more than the fundamental balance, a trap to avoid. TC/RCs signal tightness at the source, at the concentrate level. And Chinese demand data, imports and activity, remain the leading cyclical driver. One last caution about the famous Dr. Copper signal. Copper remains a valuable barometer of industry, but its message is now distorted: by tariff-related flows, by electrification demand that partly disconnects it from the short cycle, and by financial positions on COMEX. Rising copper no longer necessarily means the economy is accelerating; it may reflect a mine shortage, tariff stockpiling or a bet on the transition. The doctor still examines the world economy, but its diagnosis must now be read with more caution than before. Sources and further reading - International Copper Study Group (ICSG): global data on copper production, consumption and market balance. - London Metal Exchange, copper price and stocks: the world reference for price and warehouses. - Goldman Sachs Research and ING, analyses of the US tariff (Section 232) impact on the COMEX-LME gap, 2026. - Morgan Stanley, J.P. Morgan and ICSG, estimates of the refined copper market deficit in 2026 (150,000 to 600,000 tonnes). - l0g, Copper: the shortage arriving via Hormuz and El Niño. - Related guides: Reading the gold market, Reading the uranium market and Reading the oil market. ============================================================================ REFERENCE GUIDE: Reading CLOs and leveraged loans: tranches, waterfall and real risks URL: https://l0g.fr/en/guides/read-clos-and-leveraged-loans/ Canonical French source: https://l0g.fr/guides/lire-les-clo-et-prets-a-effet-de-levier/ Date: 2026-07-11 (reviewed 2026-07-11) ---------------------------------------------------------------------------- There is a machine that turns risky loans to indebted companies into AAA-rated bonds, seemingly as safe as a government's debt. This machine, the CLO, buys two-thirds of a market of more than $1 trillion, and it is regularly accused of being the next 2008. The accusation is partly unfair, partly deserved. Unfair, because the CLO came through the 2008 crisis without damage, exactly where its cousin the subprime CDO collapsed. Deserved, because it concentrates a major share of today's corporate credit risk, in an increasingly opaque market. This guide separates the legend from the mechanics. The leveraged loan, the raw material It all starts with the leveraged loan. It is a credit granted to a company that is already indebted, whose credit rating is below investment grade. Three traits define it. It is syndicated: a bank arranges it, then sells shares to institutional investors. It is senior and secured: in a bankruptcy, its holders are repaid before other creditors, out of the collateral. And it is floating-rate, indexed to SOFR plus a margin, which protects the lender from a rise in rates but strains the borrower when rates climb. These loans finance the buyout economy: LBOs, acquisitions, refinancings. The U.S. market amounts to about $1.2 trillion. One momentous change has transformed it: the near-disappearance of covenants. A cov-lite loan no longer requires the borrower to meet financial ratios tested regularly. Now the norm, these loans let a company sink for longer before default is recognized, at the cost of a lower recovery for creditors when the day comes. The CLO, a tranche factory Alone, a leveraged loan is illiquid and risky. Pooled with hundreds of others in a CLO, it changes nature. The CLO, for Collateralized Loan Obligation, is a securitization vehicle: it buys a portfolio of 150 to 300 loans, and funds that purchase by issuing its own securities, split into tranches of rising risk. It is the leading buyer of the leveraged loan market, absorbing about two-thirds of it. The apparent magic of securitization lies in this hierarchy. Even if the portfolio is made of risky loans, diversification and the order of repayment mean the top tranche takes a loss only if a huge fraction of the loans default at once. That is what earns it a AAA rating, though none of the underlying loans would deserve one. The equity tranche, at the bottom, plays the reverse role: it takes the first losses and absorbs the shock, in exchange for the highest yield. The waterfall and the protection tests The heart of a CLO is its waterfall, the order in which money flows. Each quarter, the interest paid by the loans first covers fees and the senior tranche, then descends tranche by tranche, the equity receiving only the residual. Losses follow the reverse path, from the bottom up. This structure would be fragile without safeguards: the overcollateralization tests. These tests continuously check that the value of the loans sufficiently exceeds that of the tranches. If too many loans are downgraded or default, a test fails, and the waterfall reconfigures automatically: money that went to the equity is diverted to prepay the senior tranche, until the cushion is rebuilt. The CLO manager must also respect limits, notably a cap on CCC-rated loans, often around 7.5% of the portfolio. Beyond it, the overload of very risky assets also triggers protective mechanisms. The CLO is therefore an actively managed vehicle with an internal firewall, two traits that set it apart from a static basket. A CLO is not a CDO: the lesson of 2008 This is the most widespread confusion, and it must be dispelled. The CDO that blew up the system in 2008 shared the CLO's technique, tranched securitization, but not its raw material. The crisis CDOs were backed by subprime mortgages, heavily correlated with one another: when U.S. housing fell, all the loans deteriorated together, and diversification protected nothing. Worse, tranches of CDOs had been re-securitized into other CDOs, piling opacity on leverage. The CLO, by contrast, is backed by loans to companies spread across dozens of sectors, whose troubles are less synchronized. The concrete result: during the 2008 crisis, no AAA CLO tranche took a loss, while subprime CDOs collapsed. This historical robustness is real and worth recalling against hasty comparisons. It has a limit, though: the past does not guarantee the future, and a shock hitting a large number of companies at once, a severe recession or a technological disruption, would test the diversification on which all confidence in the senior tranches rests. The real risks Dismissing the new-subprime legend does not mean the CLO is danger-free. Its risks are simply different. The first is the floating rate: borrowing companies pay more when rates rise, eroding their ability to repay at the worst moment. The second is the erosion of covenants: the cov-lite norm delays the recognition of default and lowers recovery. The third is the most insidious, and it blurs the very reading of risk. This third risk is that of negotiated restructurings, or LMEs, for liability management exercises. Rather than defaulting openly, a struggling company swaps its debt, extends its maturities or grants collateral to some creditors at the expense of others. The traditional default rate thus stays deceptively low, around 1.2% at end-2025, while these manoeuvres now make up the bulk of distress activity. In other words, the most-watched statistic understates the real stress in leveraged credit. A CLO can show few defaults while holding loans in silent restructuring. The rivalry with private credit The leveraged loan market, known as syndicated or BSL, no longer lives alone. It now faces private credit, that is, direct loans made by funds without going through bank syndication. This competitor, long confined to mid-sized companies, has grown to rival the syndicated market in size, and CLOs backed by these direct loans (middle-market CLOs) are the fastest-growing segment. This rivalry has a perverse effect. To win the best borrowers, syndicated and private lenders wage a race to the bottom on flexibility: even lighter covenants, interest paid in kind rather than cash (PIK), more permissive documentation. The risk does not disappear, it shifts and hides. We explored the private side of this dynamic in our article on private credit's two prices and our private credit guide. The CLO and the private-debt fund are two slopes of the same mountain of leveraged corporate debt. Reading a CLO in practice To judge a CLO, several dials combine. The price of the underlying loans on the secondary market, tracked by indices such as the Morningstar LSTA, gives the general sentiment on leveraged credit. The cushion in the overcollateralization tests says the margin before a tranche is affected. The share of CCC-rated loans signals the portfolio's quality drift. The proportion of cov-lite loans measures the weakness of protections. And above all, the default rate must be read alongside the restructuring rate, because the former alone now lies by omission. The quality of the manager, finally, matters, since it is the manager who steers the portfolio under stress. A balanced conclusion is in order. The CLO is not the CDO of 2008, and saying so serves the truth. But it concentrates the credit risk of an unprecedented corporate-debt cycle, in structures that few end investors truly understand, and in a market where the sovereign statistic, the default rate, has become misleading. The past resilience of senior tranches is a fact; the growing fragility of the raw material is another. Reading a CLO means holding both at once. Sources and further reading - Moody's Ratings, 2026 leveraged finance and CLO outlooks: market sizes, issuance and default trends. - S&P Global Ratings, U.S. leveraged finance quarterly update: discipline in recent deals and risk in older vintages. - PineBridge Investments, "2026 Leveraged Finance Outlook": the place of restructurings (LMEs) in default activity. - Congressional Research Service, "Leveraged Loans and Collateralized Loan Obligations": mechanics and financial-stability issues. - l0g, Private credit: one asset, two prices and The silent contagion of private credit. - Related guides: Reading private credit risk and Reading credit spreads. ============================================================================ REFERENCE GUIDE: Reading money market funds: Rule 2a-7, stable NAV and run risk URL: https://l0g.fr/en/guides/read-money-market-funds/ Canonical French source: https://l0g.fr/guides/lire-les-fonds-monetaires/ Date: 2026-07-11 (reviewed 2026-07-11) ---------------------------------------------------------------------------- Where does money go when it is neither in a bank nor invested? A large part sits in money market funds, a reservoir of nearly $8 trillion in the United States that almost no one watches, until the day it empties. These funds finance a major share of the government's short-term debt, absorb the world's spare cash, and sit at the heart of every liquidity crisis, from 2008 to March 2020. Reputed to be dull, they are in fact one of the most sensitive nodes of financial plumbing. This guide explains how they work, and how to read them. A money market fund, in brief A money market fund is a pooled vehicle that invests only in very short-term, high-quality instruments: Treasury bills, repo, agency debt, commercial paper from solid issuers. Its promise is three words: safety, liquidity, yield. You park cash there as in an account, you can withdraw it any time, and it earns interest close to short-term market rates. It is the waiting vehicle par excellence, where corporate treasurers, managers and households park their cash between two uses. This convenience rests on a useful fiction: the stable one-dollar net asset value. Where an equity fund's price moves constantly, a money fund strives to keep each share at exactly one dollar, so the saver feels they hold cash, not a risky investment. The whole regulatory edifice aims to keep that fiction workable. And all the danger arises the day it no longer is. Government, prime, municipal: three families Money market funds are not all alike. Three families differ by what they hold, and hence by their risk-return trade-off. The distinction between government and prime matters most. A government fund can hardly default, since it lends only to the state and against state collateral. A prime fund holds commercial paper and certificates of deposit from companies and banks, which pay more but can lose value, or even become unsellable under stress. It is this prime family that broke the buck in 2008 and suffered the runs of 2020, while government funds saw money pour in seeking safety. Rule 2a-7: the regulatory corset For a fund to promise a stable value, the SEC imposes a precise corset, Rule 2a-7. It limits what it can hold and how. Credit quality must be high. The portfolio's weighted average maturity cannot exceed sixty days, bounding interest-rate risk. Above all, the fund must keep buffers of immediately mobilizable assets: since the 2023 reform, at least 25% in daily liquid assets and 50% in weekly ones. These buffers are its first line of defence against redemptions. The rule also distinguishes two types of valuation. Government and retail funds keep a stable one-dollar NAV, rounded. Institutional prime funds, since the 2016 reform, must show a floating NAV that varies at the fourth decimal with the portfolio's real market value. The idea was to wean large investors off the illusion of the fixed dollar, and to make panic withdrawals less mechanical. The result was mixed, as we will see. Breaking the buck: the dread A money fund's nightmare has a name, breaking the buck: seeing its NAV fall below one dollar, precisely below 0.995, which materializes a loss. It is exceedingly rare, but devastating, because the fund's very structure encourages flight. The first to leave is repaid at par, at one dollar; those who stay inherit the losses. This first-mover advantage turns the slightest doubt into a run, a dynamic identical to that which threatens a bank, but on a vehicle meant to be risk-free. History has two dates. In September 2008, the Reserve Primary Fund, the oldest U.S. money fund, held $785 million of Lehman Brothers commercial paper. The bank's collapse dropped its NAV to $0.97, triggering a general run on prime funds that only an exceptional federal guarantee stopped. In March 2020, the scenario nearly repeated: the dash for cash drained prime funds, and the Federal Reserve urgently created a support facility, the MMLF, to buy their assets and halt the panic. Twice in twelve years, a product sold as near-cash required the state's rescue. The 2023 reform: goodbye gates These episodes fed two waves of reform. The 2016 one had imposed the floating NAV on institutional prime funds and introduced redemption gates, along with liquidity fees, which funds could trigger under stress. The problem is that these gates, far from calming holders, encouraged them to flee even earlier, before the barrier fell. The remedy worsened the ill. The 2023 reform learned the lesson. It removed the gates, raised the liquidity buffers, and imposed mandatory liquidity fees on institutional prime funds as soon as one day's net redemptions exceed 5% of assets, to make leavers pay the cost they impose on others. The effect was radical: rather than comply, most managers preferred to close or convert their institutional prime funds, whose number fell from about twenty-five to nine. The regulator made prime funds safer mainly by making them nearly disappear, which the industry held against it. The systemic role: a tap of T-bills and repo Beyond their internal mechanics, money funds matter because they are huge and concentrated on a precise segment. With nearly $8 trillion in assets, a record reached in May 2026 amid a flight to safety, they are one of the leading buyers of Treasury bills and a pillar of the repo market. What they do with their cash irrigates the whole plumbing of short-term funding. Their behaviour carries two signals. First, the Fed's reverse repo facility, the RRP, where they parked spare cash, went from over $2.5 trillion in 2023 to near zero in 2025: that money went to seek yield elsewhere, notably in T-bills, absorbing part of the record issuance we described in our article on Treasury auctions. Second, the size of money funds is a barometer of risk aversion: when it swells fast, as in 2026, it often means investors prefer the paid safety of cash funds to the uncertainty of risk assets. Finally, a cousin was born of the same logic, payment stablecoins, whose short Treasury-bill reserves make them money funds that dare not speak their name, as our stablecoins guide shows. Reading money market funds in practice Reading a money fund means, first, looking at what it holds: a government fund and a prime fund do not carry the same risk, and the latter's extra yield is paid for in fragility at the worst moment. It means, next, checking its liquidity buffers, the share of assets mobilizable within a week, which says its ability to absorb redemptions without selling at a loss. It also means watching its weighted average maturity, a gauge of its rate sensitivity. At the aggregate level, two series deserve the eye: total assets, a thermometer of risk appetite, and RRP usage, which reveals where cash finds a home in the plumbing. One caveat, finally. A money fund's advertised stability is a contract of trust, not a law of nature. The one-dollar value holds as long as the assets are safe and the holders calm; it can break fast if either gives way. The 2023 reform strengthened the defences, but the original sin remains: promising immediate liquidity on assets that, in a general stress, are no longer quite liquid. That is why these funds, reputed the dullest in finance, remain among the most closely watched by those who track systemic risk. Sources and further reading - U.S. Securities and Exchange Commission, 2023 money market fund reform (fact sheet): mandatory liquidity fees, removal of gates, raised buffers. - Investment Company Institute, weekly money market fund statistics: assets by fund family. - U.S. SEC, money market fund statistics: portfolio composition and maturity. - Bloomberg, money market fund assets at a record $8.3 trillion in May 2026. - l0g, Record auctions: the weekly referendum on U.S. debt. - Related guides: Reading net liquidity: reserves, TGA, RRP, Reading the repo market and SOFR and Reading stablecoins and the GENIUS Act. ============================================================================ REFERENCE GUIDE: Reading interest rate swaps: IRS, OIS and swap spreads URL: https://l0g.fr/en/guides/read-interest-rate-swaps/ Canonical French source: https://l0g.fr/guides/lire-les-swaps-de-taux/ Date: 2026-07-11 (reviewed 2026-07-11) ---------------------------------------------------------------------------- It is the largest market in the world, and almost no one talks about it. Beneath every fixed-rate mortgage, every corporate loan, every pension promise, there is an interest rate swap that was used to turn one risk into another. Hundreds of trillions of dollars of notional circulate this way, invisible, until the day they break and take the sovereign bond market down with them, as in the United Kingdom in 2022. This guide takes apart the mechanics of these contracts, from the simplest exchange to the signals they send. An interest rate swap, in one transaction An interest rate swap is an agreement in which two parties exchange interest flows calculated on the same reference amount, the notional. One pays a fixed rate, set at the outset; the other pays a floating rate, recalculated each period on a market rate such as SOFR. The counter-intuitive point is that the notional principal is never exchanged: only the interest flows circulate, and most often only the net difference between the two legs is settled. What is it for? To transform an exposure. A company borrowing at a floating rate, fearing a rise, pays fixed and receives floating: it thereby locks in its funding cost. An investor expecting a fall does the reverse. The swap does not make interest-rate risk disappear, it transfers it to whoever is willing to carry it. Multiplied across the whole economy, this redistribution makes swaps the most widespread tool for managing rate risk, and the largest derivatives market on the planet. The swap rate and its curve How is the fixed rate of a swap set? At a level such that, at the outset, the contract is worth nothing to either party: this is the swap rate, the one that equalizes the expected value of the fixed and floating flows. By construction, it therefore embeds the path the market expects for the floating rate over the whole life of the contract. A ten-year swap rate sums up in a single number what the market thinks about short rates for the coming decade. Linking the swap rates of all maturities gives the swap curve, a benchmark parallel to that of government bonds. The two curves track each other closely, but the gap between them, discussed below, carries valuable information. For banks and corporates, the swap curve is often the most direct pricing reference, because it reflects the real cost of hedging rather than the yield of a particular security. OIS: the risk-free rate and bets on the central bank One family of swaps deserves special attention, the OIS, for Overnight Index Swap. Its floating leg is not a three-month rate but the overnight rate compounded over the period, SOFR in the United States. Because that overnight rate tracks the central bank's policy rate very closely, an OIS quote directly reveals what the market expects of monetary policy. Reading the OIS curve means reading the bets on coming rate hikes and cuts, central-bank meeting by meeting. The OIS plays a second role, more technical but fundamental: it serves as the "risk-free" rate for discounting the future flows of derivatives. Since the 2008 crisis, which showed that banks were not risk-free among themselves, the market has abandoned LIBOR in favour of OIS discounting. It is invisible plumbing, but it underpins the valuation of trillions of contracts. From LIBOR to SOFR: the great shift For decades, the floating leg of swaps was indexed to LIBOR, an interbank rate calculated from banks' submissions. The scandal over its manipulation, revealed after 2008, sealed its fate. For the dollar, LIBOR stopped being published at the end of June 2023, replaced by SOFR, a rate backed by actual repo transactions on Treasuries, hence far harder to rig. We detail how it is built in our repo and SOFR guide. The shift is not just a change of name. LIBOR embedded a bank-risk premium and existed for several maturities; SOFR is an overnight, secured rate with no credit premium. The whole edifice of swaps had to be rebuilt on this new foundation, and once-common instruments, such as forward rate agreements, have virtually disappeared in the post-LIBOR world. It is one of the largest market overhauls ever undertaken, and it happened almost without a hitch. The swap spread: swap versus Treasury Back to the gap between the swap curve and the government-bond curve. This swap spread is the difference between the swap rate and the Treasury yield of the same maturity. Long positive, it reflected a logical premium: a swap with a bank is deemed slightly riskier than a government bond. Since 2015, however, the long-maturity swap spread has turned negative, an apparent anomaly where the swap rate is below the government yield. This inversion is not a market error, it is a signal. Holding a physical Treasury consumes balance sheet and regulatory capital for a bank, whereas a swap, off balance sheet, consumes almost none. When the balance-sheet constraint tightens, investors prefer the synthetic exposure of the swap to the physical security, pushing the swap rate below the government yield. The swap spread has thus become a barometer of balance-sheet and bank-regulation pressures, as much as of credit risk. A sharp tightening or widening can also betray the unwind of a leveraged arbitrage, like the roughly $60 billion of positions unwound in April 2025, an episode we tied to the basis trade in our dedicated article. Clearing and margin: the safeguard that can bite After 2008, regulators sought to reduce the risk that a counterparty default on a swap would spread. The answer was central clearing: most standardized swaps now pass through a clearing house, a central counterparty that steps between the two parties and guarantees the flows. In exchange, it demands margin: an initial margin posted as collateral, and a variation margin marked to market every day. This setup makes the system safer in normal times, but it introduces a fragility channel. When rates move violently, variation-margin calls explode, and participants must find cash immediately to meet them. If they do not have it on hand, they sell assets, often the most liquid ones, which can amplify the initial shock. Margin protects against counterparty risk, but it transmits liquidity risk. This is exactly the mechanism that nearly took down the UK debt market. When swaps break: the UK LDI spiral In 2022, British pension funds heavily practised liability-driven investment, or LDI: to hedge their very long-term obligations, they held gilts and leveraged interest rate swaps. As long as rates rose slowly, all was well. On 23 September 2022, an unfunded budget sent gilt yields soaring at an unprecedented pace. The loop was under way: rising yields, margin calls on leveraged positions, gilt sales to find cash, further yield rises. Within days, a pension-hedging problem threatened the country's financial stability. The Bank of England had to intervene in emergency on 28 September 2022, buying gilts to break the spiral. This episode remains the textbook case of the risk that leveraged swaps pose to sovereign debt, a risk the UK regulator is still trying to defuse, as we described in our article on deleveraging the gilt market. Reading swaps in practice Interest rate swaps offer several layers of reading. The swap curve gives, better than any other instrument, the price at which the market exchanges rate risk at each maturity. The OIS curve, more precisely, reads as a permanent poll on the central bank's coming decisions. The swap spread says less about credit risk than about the state of bank balance-sheet constraints, and a sharp move can signal an arbitrage unwind. Finally, the margin mechanism is the channel through which a rate shock becomes a liquidity shock: watching who is leveraged on swaps, and with what collateral cushion, means watching the next possible accident. One last caution guards against a misleading figure. The size of the swap market is measured in notional, and that notional runs into the hundreds of trillions of dollars, enough to alarm. But the notional is never exchanged: the real risk bears only on the net flows and the market value of the contracts, a tiny fraction of that amount. As often in finance, the biggest number is not the most dangerous; the danger lies in leverage and margin, not in the notional on display. Sources and further reading - Bank for International Settlements, OTC derivatives statistics: the size of the interest-rate swap market in notional. - Bank for International Settlements, "Beyond LIBOR: a primer on the new benchmark rates": the transition to risk-free rates. - Federal Reserve Bank of New York, Secured Overnight Financing Rate (SOFR): the reference rate for dollar swaps. - Bank of England, gilt market operation, 28 September 2022: the response to the LDI crisis. - l0g, Gilts: deleveraging before the next accident and Basis trade: at its peak per the Fed, moribund per the market. - Related guides: Reading the repo market and SOFR and Reading the Treasury market. ============================================================================ REFERENCE GUIDE: Reading dealer gamma: when options hedging drives the market URL: https://l0g.fr/en/guides/read-dealer-gamma/ Canonical French source: https://l0g.fr/guides/lire-le-gamma-des-dealers/ Date: 2026-07-10 (reviewed 2026-07-10) ---------------------------------------------------------------------------- There is a force that moves markets every day with no news to explain it. It comes neither from fundamental analysis nor from a manager's conviction, but from plumbing: the obligation, for options sellers, to hedge their risk by buying or selling the underlying, mechanically, continuously. On days of heavy same-day options activity, this hedging becomes the primary driver of moves. Understanding dealer gamma means seeing the invisible hand that, depending on the case, pins the market in place or hastens its fall. This guide takes apart its workings. Options, delta and gamma: the bare minimum An option gives the right to buy (a call) or sell (a put) an asset at a set price up to an expiry. Two measures are enough to follow this guide. Delta is the sensitivity of the option's price to the underlying's: a delta of 0.5 means the option moves 0.50 when the asset moves 1. Gamma is the speed at which that delta changes as the underlying moves. Gamma is greatest when the price is near the strike and when expiry approaches, two conditions met by same-day options. Keep the image: delta says how much you are exposed, gamma says how fast that exposure changes. High gamma means an exposure that changes fast, hence a hedge to be readjusted constantly. It is from this constant readjustment that the flows that move the market are born. The dealer who must hedge When a retail trader or a fund buys an option, someone sells it to them. That seller is most often a market maker, a dealer, whose job is not to bet on market direction but to earn the spread between bid and ask. To stay neutral, it must cancel the delta its option sale left it with, by taking an offsetting position in the underlying. If it sold a call, it buys shares or futures to hedge; if it sold a put, it sells them. The crucial point is that this hedging is not a directional choice, it is a mechanical constraint. And because delta changes with price, through gamma, the dealer must adjust its hedge continuously, buying and selling the underlying as the market moves. Multiply this by millions of contracts, and you get a flow that genuinely weighs on price. The direction of that flow depends on one thing: is the dealer broadly long or short gamma. Long gamma, short gamma: dampen or amplify This is the heart of the matter. When dealers are collectively long gamma, their hedging pushes them to buy when the market falls and sell when it rises. This behaviour is counter-cyclical: it pulls the price back toward its starting point, dampens volatility and tends to pin the market around the large strikes. When dealers are short gamma, the logic reverses: they sell when the market falls and buy when it rises, a pro-cyclical behaviour that amplifies moves and accelerates declines. The price level where dealers flip from long to short gamma has a name, the gamma flip. Above it, the market tends to be calm and drawn toward the large strikes; below it, it turns nervous and prone to accelerations. Specialist data providers estimate this level from options open interest, under the name GEX, for Gamma Exposure. A positive GEX signals a dampened market, a negative GEX one at risk of a runaway move. The 0DTE regime This mechanism, long confined to monthly expiry days, has become daily with the explosion of same-day options, 0DTE. In 2026, they account for roughly 40 to 50% of total S&P 500 options volume, and up to nearly 60% on some days. Because their gamma is extreme in the final hours before expiry, the hedging they impose on dealers concentrates within the session and must adjust in real time. On indices, this shows up as a magnet effect: toward the close, the underlying tends to gravitate to the strikes where open interest is densest, because dealer hedging turns stabilizing there. The same effect, reversed, makes expiry days (OpEx) more volatile, when the disappearance of a large block of options abruptly changes dealer positioning. Vanna, charm and rallies with no news Dealer hedging does not respond to price alone. It also responds to two other variables, the source of moves that seem inexplicable. The first is volatility: when implied volatility falls, the hedging of dealers who sold puts pushes them to buy the underlying, a flow called vanna that can feed a "vanna rally," a slow, steady rise with no news at all. The second is the passage of time: as expiry approaches, the erosion of time value (charm) also generates directional hedging flows. These second-order flows have powerful explanatory value. They account for markets that grind higher for weeks in a deceptive calm, carried by hedging mechanics alone, then reverse violently as soon as volatility rebounds and flips the sign of all those flows. Calm is not the absence of risk; it is sometimes the product of a self-reinforcing hedge, until it breaks. When gamma breaks: Volmageddon and GameStop Two episodes show the same mechanism in both directions. On 5 February 2018, "Volmageddon," a rebound in volatility caught out a mass of short-volatility positions and short-gamma dealers: the VIX jumped from about 17 to over 37, the Dow lost 1,175 points in the session, a record at the time, and the inverse VIX product, XIV, collapsed some 96% before being wound down. Forced hedging fed its own decline. Conversely, in January 2021, GameStop stock showed the gamma squeeze in its upward version. Heavy call buying by retail traders forced dealers to buy the stock to hedge; that buying pushed the price, raised the delta to hedge, and compelled the dealers to buy still more, in a loop that carried the stock from under $20 to an intraday peak of $483 on 28 January. Down in 2018, up in 2021, but in both cases the same truth: gamma does not create the trend, it accelerates it. Reading gamma in practice Following dealer gamma means adding a layer of reading on top of volatility. Three reflexes help. First, locate the market relative to the gamma flip: above it, expect a dampened, pinned market; below it, amplified moves. Second, distrust the calm: very low volatility can reflect a stabilizing hedge that will reverse abruptly the day it breaks. Third, read expiry days and large strikes as potential tipping points. One caveat, finally, separates serious analysis from noise. Gross options volume does not reveal dealers' net exposure: what matters is not how many contracts trade, but the balance between customer buys and sells, which alone determines the volume dealers must hedge. GEX and gamma-flip estimates are valuable, but they rest on positioning assumptions no participant knows with certainty. Dealer gamma explains much of short-term moves; it replaces neither the fundamentals, nor caution before an indicator that remains, by nature, a reconstruction. Sources and further reading - CBOE, "Evaluating the Market Impact of SPX 0DTE Options": share of 0DTE and weight of dealer hedging. - CBOE, "0DTE Index Options and Market Volatility": research on the impact of 0DTE and gamma squeezes. - SpotGamma, gamma exposure (GEX) and gamma-flip methodology: estimating dealer positioning. - Related guides: Reading VIX and MOVE volatility and Reading the CFTC COT report. ============================================================================ REFERENCE GUIDE: Reading a bank's soundness: capital, liquidity and hidden losses URL: https://l0g.fr/en/guides/read-bank-health/ Canonical French source: https://l0g.fr/guides/lire-la-solidite-d-une-banque/ Date: 2026-07-10 (reviewed 2026-07-10) ---------------------------------------------------------------------------- A bank is not a company like the others. It can post record profits the quarter before it disappears, because its raw material is not a product but trust, and trust evaporates in hours, not years. In March 2023, Silicon Valley Bank went from "well capitalized" to placed under receivership in two days. Understanding how that happens means learning to read four things: capital, liquidity, the losses the balance sheet does not show, and the nature of the deposits. This guide takes them one at a time, with SVB as the thread. Solvency and liquidity, two ways to die The first distinction to master is between solvency and liquidity, because a bank can die of each, for opposite reasons. Solvency is a question of capital: does the bank have enough equity to absorb its losses before they swallow depositors' money? Liquidity is a question of cash: can it honour the withdrawals demanded of it, here and now, without dumping its assets? The two do not coincide. A bank that is solvent on paper, with assets exceeding liabilities, can be killed by a run if it cannot raise cash fast enough. This is a liquidity crisis, and it is almost always how banks die: not through an accounting hole found in the cold, but through a funding flight felt in the heat of the moment. Solvency wears a bank down slowly; liquidity kills it fast. Reading a bank therefore means watching both dials at once. Capital: how much loss a bank can absorb Capital is the loss-absorption cushion. The headline measure is the CET1 ratio, for Common Equity Tier 1: the hardest form of equity, essentially common shares and retained earnings, divided by risk-weighted assets. That weighting is crucial: a loan to a fragile company counts for more than a government bond deemed risk-free, so two banks with the same balance-sheet size can face very different capital needs. The Basel III framework stacks several requirements. The CET1 minimum is 4.5% of risk-weighted assets, to which a 2.5% conservation buffer is added, lifting the effective floor to 7%. Systemically important banks (G-SIBs) carry an additional surcharge of 1 to 3.5 points, and large banks hold a management buffer above that in practice. A second safeguard completes the setup, the leverage ratio. It divides equity by total exposures without weighting them by risk, which stops a bank from looking sound by holding only assets the models deem "risk-free." A bank that scores well on CET1 but poorly on leverage holds little capital relative to its gross size: the dual view is essential. Liquidity: surviving a run Capital is useless if the bank cannot pay its depositors tomorrow morning. That is the purpose of the liquidity ratios, also from Basel III. The LCR, or Liquidity Coverage Ratio, requires a bank to hold enough high-quality liquid assets, central-bank reserves and Treasuries first, to cover its net cash outflows in a 30-day stress scenario. It must stay above 100%. The NSFR extends the logic over a year: it checks that illiquid assets are backed by stable resources, not by wholesale funding that can evaporate. These ratios have a limit, exposed in 2023: they assume a 30-day stress and outflow rates calibrated before the age of banking apps. A digital run, where billions leave in a few hours at the swipe of a finger, moves faster than the LCR anticipated. Regulators drew the lessons, as we detailed in our article on the regional-bank liquidity reform. The trap of unrealized losses: HTM, AFS and AOCI Here is the subtlest and most dangerous mechanism. When a bank buys bonds and interest rates rise, the market value of those bonds falls. But whether that fall appears in its accounts depends on an accounting choice. Securities classified "available-for-sale" (AFS) are marked to market, and their losses flow through an equity line called AOCI. Securities classified "held-to-maturity" (HTM) stay recorded at purchase cost: their unrealized loss appears nowhere on the balance sheet, as long as the bank does not sell them. This treatment creates a time bomb. A bank can display a comfortable CET1 while carrying unrealized losses that, once realized, would wipe out much of its equity. A U.S. exemption also let mid-sized banks exclude AOCI from their CET1, inflating their apparent solvency. The "Basel III endgame" reform, re-proposed in March 2026, corrects exactly this by requiring more banks to include AOCI, with a five-year transition. When you read a bank, the question is not only "what is its CET1," but "what would its CET1 be if all its securities were marked to market." The liability side: where the money comes from, and how fast it leaves People often look at a bank's assets, its loans and securities. Its vulnerability, though, sits on the right of the balance sheet, in its liabilities, that is, in the nature of its funding. Not all deposits are equal. A deposit covered by federal insurance, under the $250,000 cap, is stable: its holder has no reason to flee, since they are protected even in a failure. An uninsured deposit, above that cap, is volatile: at the first doubt, its holder has every incentive to be first out. The share of uninsured deposits is therefore a first-order fragility indicator, as is their concentration. A bank whose depositors are few, connected and alike, as SVB's start-ups and venture funds were, faces a herd-run risk: they hear the same rumour, draw the same conclusion and flee at the same instant. A retail bank with millions of small insured depositors, by contrast, has far stickier funding. Reading the liability side means measuring not only how much the bank owes, but how fast that money can leave. The textbook case: Silicon Valley Bank Silicon Valley Bank combined every fragility at once, which makes it the perfect case study. On paper, it was "well capitalized" in the regulatory sense. In reality, three flaws stacked up: a massive bond portfolio carrying about $15 billion of unrealized HTM losses, close to its equity; a liability base roughly 94% uninsured; and a depositor base ultra-concentrated in the tech ecosystem, connected in real time. The sequence was blinding. To meet the first withdrawals, SVB had to sell securities, realizing its unrealized losses and confirming the doubt about its solvency. The news spread within hours in a hyper-connected community, and on 9 March 2023 customers sought to withdraw more than $40 billion in a single day, with some $100 billion more expected the next day. No bank survives that. SVB fell on 10 March, a victim not of a credit default but of a liquidity crisis triggered by a solvency flaw its balance sheet masked. The Fed responded with the BTFP, a facility lending against securities valued at par to neutralize precisely those unrealized losses. Reading a bank in practice Judging a bank's soundness therefore means cross-checking several dials rather than trusting one. CET1 says how much loss it can absorb, but it must be read net of the unrealized losses hidden in HTM, not just as reported. The leverage ratio corrects the illusion of a balance sheet padded with supposedly risk-free assets. The LCR and NSFR say whether it can withstand a funding shock, keeping in mind that a digital run moves faster than their assumptions. And the structure of the liabilities, the share of uninsured deposits and the concentration of the client base, says how fast the money can flee. None of these numbers is enough on its own. A bank can be solvent and illiquid, well capitalized in the regulatory sense yet fragile because its losses are off balance sheet and its depositors nervous. A bank's soundness is not a number, it is the coherence between these dials. And the last one, the hardest to quantify, remains trust: it appears in no ratio, but it is always the first to leave. Sources and further reading - Federal Reserve, "Review of the Federal Reserve's Supervision and Regulation of Silicon Valley Bank," 28 April 2023: the Barr report, uninsured-deposit share (~94%) and withdrawal figures. - Basel Committee on Banking Supervision, the Basel III framework: capital minimums, LCR and NSFR. - Basel Committee, report on the 2023 banking turmoil: lessons from the 2023 failures. - Federal Reserve, OCC and FDIC, "Basel III endgame" re-proposal, March 2026: broader inclusion of AOCI in CET1. - l0g, U.S. regional banks: from the 2023 panic to the liquidity reform. - Related guides: Reading net liquidity: reserves, TGA, RRP and Reading the Fed's H.4.1 balance sheet. ============================================================================ REFERENCE GUIDE: Reading the carry trade: borrow low, invest high, and manage the unwind URL: https://l0g.fr/en/guides/read-the-carry-trade/ Canonical French source: https://l0g.fr/guides/lire-le-carry-trade/ Date: 2026-07-10 (reviewed 2026-07-10) ---------------------------------------------------------------------------- Some strategies pay a little, often, then a lot at once, but the wrong way. The carry trade is one of them. The idea is plain: borrow where money is cheap, invest where it pays, and live off the gap. It is one of the most discreet and powerful engines of global markets, able to keep a lid on a currency for years and then trigger a crash in hours. This guide takes apart its mechanics, from principle to unwind. The yen, the funding currency par excellence, is the guiding thread. The principle: capturing a rate gap The carry trade rests on an interest-rate asymmetry between two currencies. An investor borrows in a low-yielding currency, the funding currency, and converts the proceeds to invest in a higher-yielding asset, often denominated in another currency. As long as the exchange rate holds still, they pocket the difference between the two yields, what markets call the carry, or the pickup. Take a deliberately simple example. Borrowing in yen costs about 1% a year; investing the proceeds in U.S. bonds earns close to 4%. The gap, about 3 points, is the gross income of the carry, earned without committing much of your own capital. Repeated across trillions of dollars and amplified by leverage, this mechanism drives a large share of global capital flows. It explains why a low-rate currency can stay weak for a long time: as long as the carry works, everyone is selling it. Funding and target currencies Not all currencies are equal in this game. A good funding currency combines two traits: a low interest rate and a reputation for stability. The Japanese yen has ticked both boxes for twenty years, the Bank of Japan having kept rates on the floor well after everyone else. The Swiss franc plays a similar role, and the dollar itself served as a funding currency during the zero-rate years that followed 2008. At the other end, target currencies offer a higher yield: the dollar when the Fed holds rates high, but above all high-yielding emerging currencies, the Mexican peso, Brazilian real, South African rand, Indian rupee, whose policy rates often exceed 8 to 14%. The choice of pair depends on risk appetite. Carry between developed currencies, yen versus dollar for instance, offers a more modest gap but contained volatility. Carry into emerging markets promises a much bigger pickup, at the cost of markedly higher currency and default risk. In both cases, the implicit bet is the same: that the funding currency will not strengthen abruptly. A bet on calm: the carry as a short-volatility trade This is the most important point, and the most misunderstood. The carry trade is not only a bet on a rate gap, it is a bet on stability. Its payoff profile resembles that of an insurance seller: it collects a steady premium as long as nothing happens, and suffers a heavy, sudden loss when the accident strikes. In market language, the carry amounts to selling volatility. This nature explains two characteristic behaviours. First, the carry loves quiet: the lower the volatility, the safer it looks and the more capital flows in, which compresses volatility further, in a self-reinforcing loop. Second, it hates surprises: a volatility shock, even one unrelated to the currency, can be enough to send capital fleeing and reverse the move. The VIX and currency-volatility indices are, on that score, leading barometers of the carry's health. Leverage, the accelerator both ways A 3-point gap makes no one rich if it applies to little capital. The carry trade only becomes significant with leverage, obtained in several ways: in the FX forward market, where you take a position far larger than your stake; via repo, pledging the securities bought to borrow again; or through derivatives that replicate the exposure without tying up the notional. Ten-times leverage multiplies the carry's return by ten, but also the loss if it reverses. Leverage introduces a second danger, more insidious than a simple loss: the margin call. When the position turns, the lender demands more collateral. To provide it, the investor must sell assets, often the most liquid in the portfolio, including those with nothing to do with the carry. This is the channel through which an accident confined to the yen contaminates equities, credit or crypto on the other side of the world. The unwind: when it all comes undone at once The dreaded scenario has a name, the unwind. It is triggered when the funding currency strengthens abruptly, which typically happens when the central bank that issues it raises rates, or when a market shock sparks a rush to safety. The first losses trigger margin calls, which force sales, which strengthen the funding currency further as it is bought back in a panic, which deepen the losses of the other holders: the spiral is under way. The reference episode is 5 August 2024. A rate hike by the Bank of Japan, combined with weak U.S. data, triggered a lightning rebound in the yen. The unwind that followed sank the Nikkei 12.4% in a single session, its worst since the 1987 crash, drove the VIX volatility index above 65, and dragged down assets as distant as bitcoin. The Bank for International Settlements devoted a bulletin to the shock, whose lesson fits in one sentence: a carry unwind never stays confined to its home market. We extended this analysis in our article on the carry trade caught out by Japanese long rates and the one on the dollar-yen unwind risk. The dials to watch The carry gives no warning, but it leaves traces. Four dials help gauge its tension. The first is the rate gap: it is what pays the trade, and its compression, when the funding central bank tightens or the others ease, reduces its appeal. The second is FX implied volatility: low, it lulls; a spike wakes it and triggers sales. The third is speculative positioning, published weekly by the CFTC in its COT report: an extreme short position on the yen betrays a crowded trade, hence vulnerable to the slightest reversal. The fourth is the exchange rate itself relative to the thresholds where an FX intervention becomes likely, since that intervention can be the spark of the unwind. The l0g Yen Carry Monitor aggregates precisely these signals. Reading the carry in practice The carry trade is not a trading-desk curiosity, it is a structuring force. Read well, it illuminates three things. It explains why a currency can stay weak for a long time with no visible reason: because it funds the planet. It is a reminder that a calm market is not a safe market, and that the lowest volatility often precedes the most violent shock, because it draws in the leverage that will make it brutal. And it links markets one believes are separate: the day the yen turns, Japanese equities, emerging markets and crypto can fall together, not through fundamental contagion but through the need for liquidity. Reading it therefore means looking beyond the rate gap that feeds it, toward the conditions that make it sustainable or explosive: volatility, leverage, positioning. The carry thrives on boredom and dies in panic. The whole difficulty, for the observer, is to recognise the moment when boredom turns to complacency. Sources and further reading - Bank for International Settlements, Bulletin no. 90, "The market turbulence and carry trade unwind of August 2024": anatomy of the 5 August 2024 unwind. - Bank of Japan, monetary policy decisions: the path of Japanese rates, the bedrock of the yen carry. - CFTC, Commitments of Traders: speculative positioning on the yen and currencies. - l0g, Yen carry trade: the fuse in Japanese bonds and Dollar-yen: the unwind risk. - Related guides: Reading VIX and MOVE volatility, Reading the CFTC COT report and Reading the dollar: DXY and cross-currency basis. ============================================================================ REFERENCE GUIDE: How to Read CBO Forecasts: the Budget Baseline and Its Traps URL: https://l0g.fr/en/guides/read-cbo-budget-outlook/ Canonical French source: https://l0g.fr/guides/lire-les-previsions-du-cbo/ Date: 2026-07-08 (reviewed 2026-07-08) ---------------------------------------------------------------------------- import InfographicBars from '../../components/infographic/InfographicBars.astro'; Every projection of U.S. debt, every debate about a tax bill, every headline on the deficit leans on the same numbers: those of the Congressional Budget Office. Yet those projections are among the most misunderstood in macro, because people read them as forecasts when they are benchmarks built under strict rules. This guide explains what the CBO is, how to read a baseline, and why its conventions change everything. An independent budget referee The Congressional Budget Office was created in 1974 to give Congress an independent budget-analysis capacity separate from the executive branch. It is nonpartisan by design. It recommends no policy; it estimates the consequences of policy. That neutrality is why its numbers become the common reference for both parties, even when they attack the assumptions. Understanding the CBO starts with this: it does not say what will happen. It says what would happen under specified conditions. What the CBO publishes The CBO produces several documents that should not be confused. The most followed is the Budget and Economic Outlook, the ten-year baseline generally published twice a year, projecting revenues, spending, deficits and major economic variables. The Long-Term Budget Outlook extends the exercise over thirty years, where demographics and interest costs dominate. Cost estimates, or scores, evaluate the budget effect of a specific bill. Add the Monthly Budget Review, which tracks execution during the year, and thematic studies. Knowing which document a number comes from is the first reflex. Reading a baseline: the numbers that matter A baseline is read through a few magnitudes, preferably as a share of GDP so they are comparable over time. First, the annual deficit: for fiscal year 2026, the CBO expects $1.9 trillion, or 5.8% of GDP, unusually high outside recession or war. Second, debt held by the public, the relevant measure of the burden, projected to rise from 101% of GDP in 2026 to 120% in 2036. Third, net interest, now the fastest-growing line: $1.0 trillion in 2026, rising to $2.1 trillion in 2036, already above the defense budget. One number deserves special attention: the primary deficit, meaning the deficit excluding interest. It separates what the state spends beyond current revenue from what it pays for past debt. When the deficit widens mostly because of interest, the problem is the stock of debt. When it comes from the primary balance, current fiscal policy is the issue. The misunderstood rule: the current-law baseline This is the core issue and the source of most errors. The baseline is not a forecast of what will happen. It is a projection of what would happen if current law remained unchanged. That convention follows statutory rules, notably section 257 of the Balanced Budget and Emergency Deficit Control Act, and those rules produce counterintuitive effects. Three examples are enough. Discretionary spending is assumed to start from the latest enacted level and grow with inflation, not stay frozen. Tax or spending provisions scheduled to expire are assumed to expire, even when everyone expects Congress to extend them: a temporary tax cut therefore disappears from the baseline at its sunset date, so an extension, however likely, shows up as an added cost relative to the benchmark. Finally, benefits under a program are assumed to be paid in full even if its trust fund is exhausted and no law authorizes full payment. The baseline is therefore a legal object as much as an economic one. Reading it without knowing those conventions means mistaking an accounting artifact for a real trajectory. Static or dynamic: reading a score When the CBO evaluates a bill, the method matters as much as the result. A conventional score is static: it measures direct effects on revenues and outlays without assuming the policy changes growth, and often without including the debt-service cost it creates. A dynamic score, rarer and more contested, includes macroeconomic feedbacks, for example the idea that a tax cut could stimulate activity and therefore revenues. Neither method is neutral. The methodological choice can make a reform look costly or self-financing on paper. You need to know which kind of score you are reading before judging a fiscal claim. Fragile assumptions The longer the horizon, the more the baseline depends on economic assumptions whose small changes create huge gaps. The CBO projects only 1.7% average growth over thirty years and the weakest demographic growth in U.S. history. Under current law, debt held by the public would rise from around 100% of GDP in 2025 to 156% in 2055, or roughly 172% in projections incorporating laws enacted since. But those figures are only the center of a fan. If the average interest rate on the debt diverged from the assumption by only 5 basis points per year, debt in 2055 would be 204% of GDP in one case and 121% in the other. Likewise, productivity higher or lower by 0.5 point per year would bring debt to 113% or push it to 203%. In other words, the gap between a manageable and explosive path rests on tiny assumptions. A single thirty-year debt number should always be read as the middle of a range, not destiny. Reading the CBO in practice Used well, the CBO is extremely valuable, provided it is taken for what it is: a direction, a common benchmark and an order of magnitude, not prophecy. A few habits help. Separate the primary deficit from the total deficit to know where deterioration comes from. Read shares of GDP rather than dollar amounts. Check the economic assumptions · growth, rates, inflation · behind the baseline, and remember the sensitivity. When reading a score, ask whether it is static or dynamic, and whether the supposedly expiring provisions will really expire. This framework applies to concrete debates. The debt and interest path documented by the CBO is the starting point for our analysis of the return of term premium, and the supply of securities it implies is read through our guide to the Treasury market. The sensitivity of debt to productivity also connects to the debate over AI productivity gains. The glossary defines the acronyms. --- Main sources: Congressional Budget Office, “The Budget and Economic Outlook: 2026 to 2036”; CBO, “The Long-Term Budget Outlook: 2025 to 2055”; CBO, “The Long-Term Budget Outlook Under Alternative Scenarios for the Economy and the Budget”; CBO, “CBO Explains How It Develops the Budget Baseline”; CBO, “CBO Explains the Statutory Foundations of Its Budget Baseline”; Committee for a Responsible Federal Budget on CBO’s long-term outlook. ============================================================================ REFERENCE GUIDE: How to Read the CFTC COT Report: Who Is Positioned, and How to Know URL: https://l0g.fr/en/guides/read-cftc-cot-report/ Canonical French source: https://l0g.fr/guides/lire-cot-cftc/ Date: 2026-07-08 (reviewed 2026-07-08) ---------------------------------------------------------------------------- In futures markets, everyone eventually shows their cards once a week. The Commitments of Traders report, or COT, is that moment of transparency imposed by the U.S. regulator. Each Friday, the CFTC publishes open positions in futures contracts, broken down by trader type. It is one of the few public windows into the real positioning of funds, banks and industrial hedgers. But you need to read the right report, understand who sits in which bucket, and know the time lag that limits its use. This guide takes the mechanism apart. The COT does not say where the market is going. It says who is positioned how, at a given date. It is structural data, not a buy or sell signal. Its value lies in revelation: the split between those hedging physical risk and those speculating on direction. Read methodically, it gives faces to open interest, the mass of outstanding contracts. Read carelessly, it gives false certainty. What COT measures The report decomposes, for each covered contract, total open interest: the number of contracts opened and not yet closed. This mass is split across trader categories, each shown in long positions, short positions and, when relevant, spreads, meaning offsetting long and short positions held by the same participant. By construction, total longs equal total shorts: every buyer has a seller. The common reading is to calculate a category’s net position, longs minus shorts. A positive number signals a long bias, a negative number a short bias. A market is included only if 20 or more traders hold positions above the CFTC reporting thresholds. Below that, trader anonymity would not be guaranteed. Positions below the reporting threshold are grouped in a residual nonreportable category. The calendar, and the three-day lag COT follows a strict rhythm. Positions are recorded as of Tuesday evening. Reporting firms send them to the CFTC on Wednesday morning. The CFTC checks and publishes them on Friday at 3:30 p.m. New York time. There is therefore a three-day lag between the snapshot and publication. The CFTC publishes no intra-week update. That lag has a direct consequence: in a fast market, real positioning may already have changed when the report comes out. COT is a background-reading tool, suited to swing and position horizons, never an intraday trigger. Confusing it with real-time data is the first mistake. The four reports, and which one to read The CFTC publishes four families of reports, each in futures-only and futures-and-options-combined versions. The historical report, Legacy, dates back to 1986. It distinguishes only two classes of reportable traders, commercial and non-commercial, plus nonreportables. It is the oldest and most consulted, but also the most misleading, because market-making banks can appear among commercials, blurring the line between real hedging and speculative activity. The Disaggregated report, available from 2006 and published from September 2009, fixes that problem for physical markets: agriculture, energy and metals. It splits participants into four clearer buckets. Traders in Financial Futures, or TFF, does the same for financial markets: rates, currencies and equity indexes. Finally, the Supplemental CIT report covers 13 agricultural contracts and isolates index funds. The practical rule is simple. For a commodity, read the Disaggregated report and the Managed Money category. For a currency or rate contract, read TFF and the Leveraged Funds category. These two buckets isolate speculative money, the one that carries the directional signal. Using Legacy for these markets means reading an outdated map. The categories that carry signal In the Disaggregated report, there are four main boxes. Producers, merchants, processors and users are physical actors · miners, refiners, merchants · who hedge. Swap dealers are banks facilitating client trades, often for hedging on behalf of others. Managed Money covers hedge funds, commodity trading advisors and third-party managers. Other Reportables gathers other large participants, such as some pension funds or corporate treasuries. In TFF, the same logic applies to financial markets. Dealers and intermediaries are market makers, usually banks. Asset managers and institutional investors cover pension funds, insurers and mutual funds. Leveraged Funds are hedge funds and leveraged managers. Other Reportables are the rest of the large-reporting universe. If you want to follow directional speculative money, Leveraged Funds are the bucket. How classification is done One decisive point is often ignored. A trader’s category is not inferred from each trade, but from its predominant business activity, self-declared on CFTC Form 40 and monitored by the Commission. Classification is therefore at the entity level, based on declared purpose, not at the intention level of each position. That nuance matters. An industrial company classified as commercial can also take speculative positions, and a bank classified as swap dealer can carry directional bets alongside hedges. COT categories are institutional approximations, not certainty about the motive behind every contract. That is exactly the kind of caveat this site insists on before any interpretation. Reading a positioning extreme The most common analytical use is to watch extremes. The idea is that hedgers, or commercials, tend to lean against the move when the market drifts far from fundamental value, while speculative money · Managed Money or Leveraged Funds · crowds into trends until saturation. A speculative net position at a historical extreme is then read as a crowded market, vulnerable to reversal once buyers or sellers are exhausted. This framework has real contextual value, but it is not a timing signal. An extreme can stay extreme and become even more extreme before it unwinds. Serious reading looks not at the raw level alone but at the weekly change and at the position within a long historical range, through percentiles or z-scores. COT locates pressure. It does not give the hour of reversal. It is a positioning compass, not a stopwatch. Blind spots Several limits frame the tool. The three-day lag prevents reactive use. Weekly frequency provides no finer granularity. Classification by predominant activity can misplace hybrid participants. Scope is also limited: COT covers listed futures and options, and ignores OTC markets, swaps and the underlying physical market. The 20-trader threshold excludes some thin markets. There is also a current-policy angle. In May 2026, the CFTC opened a public consultation on reforming its COT program, including whether to publish more recent data. The format described here may therefore evolve, and official CFTC announcements must be watched. Reading COT from the primary source The raw data is free and public. It is available on the CFTC website under Market Reports, in text and spreadsheet files, with Legacy history back to 1986. The CFTC also provides a Public Reporting Environment to filter by contract and date and export series. Exchanges such as CME publish charting tools built from the same data. For rigorous use, return to the CFTC source file rather than an aggregator, and make sure you are reading the right report for the asset class. COT sits in the same family as the other regulatory disclosures decoded on this site, such as SEC Form 13F. It usefully complements futures-market readings in our analyses of the Treasury basis trade and gold positioning. Methodology This guide describes the public functioning of the Commitments of Traders program using official CFTC explanatory notes and release schedules. Categories, thresholds and dates come from Commission documents and institutional sources. No investment strategy is recommended. COT is presented as a positioning-reading tool, with explicit limits. --- Main sources: Commodity Futures Trading Commission, Commitments of Traders page and explanatory notes; CFTC release schedule, Friday at 3:30 p.m. Eastern time using Tuesday data; CFTC FAQ and histories, with Legacy since 1986 and Disaggregated, TFF and CIT series from 2006; Federal Register, “Review of the Commitments of Traders Reporting Program,” May 5, 2026; Office of Financial Research and CME Group for financial-category definitions. Dates, thresholds and categories checked against CFTC documents. ============================================================================ REFERENCE GUIDE: How to Read Credit Spreads: OAS, Quality Scale and Stress Signal URL: https://l0g.fr/en/guides/read-credit-spreads-oas/ Canonical French source: https://l0g.fr/guides/lire-les-spreads-de-credit/ Date: 2026-07-08 (reviewed 2026-07-08) ---------------------------------------------------------------------------- import InfographicBars from '../../components/infographic/InfographicBars.astro'; A credit spread is the price, in basis points, of lending to a company rather than the state. It is one of the rare indicators that speaks before defaults arrive: when it widens, the market has already decided risk is rising. This guide covers spreads end to end, from the definition of OAS to the credit-quality scale and the stress signal. The year 2026, with spreads at their tightest since 2007, serves as both illustration and warning. What is a credit spread? A corporate bond yields more than a Treasury of similar maturity, and that yield gap is the credit spread. It compensates for everything that distinguishes a private issuer from the U.S. government: the risk of default, the difficulty of reselling the security, and uncertainty about its future. Lending for ten years to a BBB company at 5.15% while the ten-year Treasury yields 4.38% means demanding 77 basis points for the extra risk. That number, and its change over time, is the information. A spread tightens when risk appetite dominates: investors accept a small premium, confident few issuers will default. It widens when fear returns: investors demand more, or refuse to lend. A spread is therefore not only a measure of default risk. It is also a measure of sentiment. That is why it is a leading indicator: it moves before balance sheets, before defaults and often before equities. OAS, the reference measure Comparing raw bonds would be misleading because many contain options. A callable bond lets the issuer repay early if it can refinance more cheaply. That option has value and distorts the displayed yield. The reference measure adjusts for it: OAS, option-adjusted spread, removes the value of embedded options to isolate the pure credit spread across the Treasury curve, not just against a single point. Bonds with different structures become comparable. ICE BofA indexes, published daily and mirrored by the St. Louis Fed, are the most followed source. They show OAS for broad indexes such as U.S. high yield or for rating buckets. This is the common language of trading desks when they speak about “spreads.” The quality scale: from investment grade to CCC Credit risk is ordered by rating, and spreads follow that ladder in a highly nonlinear way. At the top, investment grade includes issuers rated from AAA to BBB-, judged relatively solid; BBB alone represents almost half the segment. Below begins high yield, from BB to CCC, speculative debt. The lower you go, the wider the spread, but not proportionally: it explodes at the bottom of the scale because default probability rises much faster than the rating step suggests. The gap between the top and bottom of this scale is itself a signal. When it widens, the market discriminates: it makes the weakest risk pay dearly while leaving solid issuers alone, a sign of targeted concern. When it compresses, almost everyone funds near the same price, a sign of indiscriminate risk appetite, often the prelude to unpleasant surprises. Decomposing a spread A spread is not one block. It has three parts. The first is expected loss: probability of default multiplied by loss given default, after recovery. This is the actuarial part a credit model calculates. The second is the risk premium, the extra compensation for bearing uncertainty around that average, because defaults do not happen predictably. The third is the liquidity premium, compensation for the difficulty of selling the bond when needed, higher when the market is thin. This decomposition matters when reading a move. A widening caused by higher expected loss points to real fundamental deterioration. A widening caused by exploding liquidity premium, with no balance-sheet change, points to market panic, often brutal and sometimes reversible. They look similar on a chart but say different things. CDS and CDX indexes Beside cash bonds sits a faster synthetic market. A CDS, or credit default swap, is insurance against issuer default: its premium, in basis points, rises when default looks more likely. Grouped into baskets, these contracts form CDX indexes: CDX.NA.IG for investment grade, CDX.NA.HY for high yield. Continuously traded and highly liquid, they often move before the cash market, because selling synthetic protection is easier than moving a bond portfolio. That is why the difference between CDX and cash spreads, the basis, matters. When synthetic credit decouples from cash, one of the two markets is ahead of the other, usually CDX. Watching both gives you a better chance of seeing stress arrive. Reading spreads as signal A spread is not judged by level alone, but through four crossed readings. Level first, placed in history: a spread in its tightest decile says the market is expensive, not when it will turn. Speed matters more: a spread doubling in weeks is a much stronger stress signal than a high but stable spread. The quality scale shows whether CCC is separating from BB, meaning discrimination is returning. Finally, the distress ratio, the share of high yield trading above 1,000 bp, measures the part of the market already in trouble and often precedes default waves. A low spread with a rising distress ratio is a divergence not to ignore. 2026: tightest since 2007 In early 2026, spreads are exceptionally compressed. Broad high yield trades near 285 basis points, around its June 2007 level on the eve of the financial crisis. Investment grade touched 71 basis points in January, the tightest since 1998 and the LTCM episode. To understand that compression, remember where high yield went in recent shocks: roughly 1,100 bp in March 2020, near 2,000 bp in 2008. Is this complacency? Partly, but not only, and the honest reading includes the counterargument. Index composition has improved: today’s high yield is of higher average quality than in 2007, with more BB and fewer CCC names, so a structurally lower spread can be justified by lower average risk. Still, even after correcting for this, the cushion is thin. At these levels, spreads poorly compensate for reversal risk, and the main possible move is upward. A tight spread does not announce a crisis. It reduces the warning time. The private-credit mirror This framework also illuminates a blind spot. The public bond market has an observable price, updated daily; private credit is valued by expert judgment, without a market spread. That is why the public spread becomes the implicit reference price against which smoother private valuations are judged. When public spreads widen quickly while private marks stay stable, the gap does not measure opportunity. It measures delayed loss recognition, a risk we follow in our work on private-credit contagion and in the guide to private credit analysis. Reading spreads in practice Read properly, credit spread is not one thermometer but a system. OAS gives the comparable measure of risk. The quality scale says whether the market is discriminating or applying a uniform premium. Decomposition between expected loss and premia says where a move comes from, real deterioration or liquidity panic. CDX shows what the synthetic market thinks, often earlier. The distress ratio shows how much of the market is already in trouble. In 2026, all these gauges say the same thing: risk is cheap, in a market with little room for error. The glossary defines the acronyms. Credit-related dates should be checked with issuers and regulators listed on our reference sites page. This guide complements the guide to the Treasury market, of which credit spread is the private-sector mirror. --- Main sources: ICE BofA U.S. High Yield Index Option-Adjusted Spread, FRED BAMLH0A0HYM2; ICE BofA U.S. Corporate Index OAS, FRED BAMLC0A0CM; ICE BofA CCC & Lower U.S. High Yield Index OAS, FRED BAMLH0A3HYC; Morningstar, “Corporate Credit Spreads Trading in Tightest Decile of Historical Range”; T. Rowe Price, “Are structural spread changes concealing value in credit?”; ICE, Markit CDX.NA.HY. ============================================================================ REFERENCE GUIDE: How to Read the Dollar: DXY, Cross-Currency Basis and Offshore Dollar Debt URL: https://l0g.fr/en/guides/read-dollar-dxy-cross-currency-basis/ Canonical French source: https://l0g.fr/guides/lire-le-dollar-dxy-cross-currency-basis/ Date: 2026-07-08 (reviewed 2026-07-08) ---------------------------------------------------------------------------- import InfographicBars from '../../components/infographic/InfographicBars.astro'; People talk about the dollar as a spot price, usually DXY, when it is first a funding system. Trillions of dollars are borrowed outside the United States by actors that do not earn dollars, through instruments that show up on no balance sheet. This guide reads the dollar end to end: what DXY measures and hides, why covered interest parity broke, how to read cross-currency basis, and where the real dollar debt is hidden. The 2026 regime, where a soft spot dollar coexists with tight funding, is the illustration. DXY, a biased thermometer DXY, the U.S. Dollar Index, is the most quoted gauge of dollar “strength.” It is a basket of six currencies whose weights have not changed since the euro was launched in 1999, with a 1973 = 100 base. The problem is composition: the euro alone weighs almost 57.6%, the yen 13.6%, sterling 11.9%, and the rest is split between the Canadian dollar, Swedish krona and Swiss franc. DXY is therefore not the dollar against the world. It is mostly the dollar against Europe. What it ignores matters as much as what it includes. There is no Chinese yuan, Mexican peso or Korean won, even though those economies are major U.S. trading partners. For a less distorted reading, the Federal Reserve publishes a broad trade-weighted dollar index including China, Mexico and roughly two dozen partners. It is common for DXY and the broad index to tell different stories: the dollar can fall against the euro and yen while remaining firm against emerging-market currencies. Reading DXY alone means looking at the dollar through the wrong end of the telescope. Covered interest parity, and its break Below the exchange rate lies a deeper mechanism: funding. In theory, obtaining dollars directly or synthetically · borrowing in another currency, then swapping into dollars · should cost exactly the same. That is covered interest parity, an arbitrage that should eliminate any gap. If a Japanese investor hedges a dollar investment, the hedge should cost exactly the interest-rate differential between yen and dollars, no more. Since 2008, that parity has not held. The actual cost of hedging deviates from the rate differential, and the residual, which should not exist, is the cross-currency basis. Two forces explain the break. Post-crisis regulation made it costly for banks to use balance sheet for arbitrage. At the same time, structural demand for dollars · from Japanese and European investors buying U.S. assets and hedging FX risk · pushes in one direction. Arbitrage no longer closes the gap. The gap persists. Cross-currency basis, the hidden dollar price Basis has become the best thermometer of dollar funding stress. It has been negative and persistent since 2008, meaning obtaining synthetic dollars carries a premium: the dollar is structurally expensive to fund outside the United States. When that premium widens, the dash for dollars intensifies. When it narrows, pressure eases. As with onshore repo, basis deteriorates around balance-sheet dates, quarter-end and especially year-end, when banks cut back their supply of dollars. The underlying market has exploded. In the first quarter of 2026, cross-currency swap volumes hit a $3.6 trillion notional record, up 36% year over year, with March alone at $1.56 trillion, the highest ever measured. This is not a niche market. It is the artery through which the world funds itself in dollars, and the basis says whether the blood is flowing or clotting. The FX swap, an off-balance-sheet repo To understand the risk, see what an FX swap really is: a spot currency exchange paired with the reverse exchange forward. Economically, it is a collateralized loan, the equivalent of a repo with a currency as collateral. But there is a decisive accounting difference: repo appears on the balance sheet as debt; an FX swap does not. The obligation to repay dollars at maturity is real, but it sits off balance sheet, in notes, invisible to standard debt statistics. That invisibility is the danger. An actor funding itself in dollars through FX swaps must roll the funding at maturity, often very short. While the market works, the roll is smooth. When dollar liquidity becomes scarce, as in March 2020, everyone must roll at the same time a debt no balance sheet showed, and the dash for dollars feeds on itself. The “missing” dollar debt The Bank for International Settlements has measured the hole. Obligations to pay dollars embedded in FX swaps, forwards and currency swaps exceed $80 trillion in notional value, a sum larger than combined dollar Treasury bills, repo and commercial paper. Most is very short term, multiplying rollover needs. The BIS estimates that non-U.S. non-banks alone carry nearly $25 trillion of this hidden debt, up from $17 trillion in 2016, with non-U.S. banks carrying even more. Fed swap lines, the offshore backstop Against this risk, the Federal Reserve has a backstop, symmetrical to the standing repo facility that caps onshore stress: central-bank swap lines. The Fed lends dollars to a network of foreign central banks · the ECB, Bank of Japan, Bank of England and others · which then lend them to their own banks. Dollars reach the offshore system without the Fed dealing directly with foreign private counterparties. These lines were activated massively in 2008 and again in March 2020, when usage exceeded $400 billion and broke the dash for cash. Their usage, visible weekly on the asset side of the Fed balance sheet, is a last-resort signal: when it rises, offshore funding has seized up badly enough to require the issuer central bank. Watching it means watching the moment when the plumbing breaks. Reserve or funding: reading de-dollarization Everything above leads to the most misunderstood distinction in the dollar debate. The dollar is both a reserve currency, the one central banks hold, and a funding currency, the one the world borrows and invoices in. Those two roles do not decline at the same speed. The dollar’s share of FX reserves, measured by the IMF’s COFER, is slowly eroding, around 58% in 2025 versus close to 70% in 2000. But its dominance as a funding currency is barely moving: the dollar remains on one side of nearly 88% of global FX transactions. That is why de-dollarization is real but partial. Central banks diversify reserves, including toward gold, but as long as global debt is denominated in dollars, the demand for dollar funding remains, and with it the vulnerability to dashes measured by basis. 2026: soft spot dollar, tight funding The 2026 regime shows why multiple gauges are necessary. In spot terms, the dollar is soft: DXY trades around 101, near a three-week low after its largest weekly fall since April. An observer looking only at that number sees a dollar in retreat. Yet at the same time, dollar-funding volumes are hitting records and basis pressure is widening, a sign that demand for synthetic dollars remains strong. A dollar can fall in price while remaining expensive to fund. Level and stress are separate dimensions. Confusing them is how you miss the point. The same decoupling appears in the yen carry trade, where a weak yen and low volatility support yen borrowing until the reversal. Once again, the calm surface hides tense plumbing. Reading the dollar in practice Read properly, the dollar is not a price but a multi-layered system. DXY gives market mood against Europe, with major biases. The Fed’s broad index gives the dollar against the real trading world. Cross-currency basis gives the hidden price of dollar funding and its stress level. BIS-estimated off-balance-sheet dollar debt gives the size of the rollover risk. Fed swap lines show when the backstop is deployed. And the distinction between reserve share and funding share tells what de-dollarization changes, and what it does not. This framework is the offshore counterpart to the onshore plumbing described in our guides to repo and SOFR and system liquidity. It also illuminates the international dollar layer discussed in our pieces on eurodollars and the hidden price of the dollar. The glossary defines the acronyms. Release dates should be checked directly with the institutions listed on our reference sites page. --- Main sources: BIS, “Dollar debt in FX swaps and forwards: huge, missing and growing”; ClarusFT, Q1 2026 cross-currency swap volumes; ICE, U.S. Dollar Index; Federal Reserve, trade-weighted dollar indexes; Federal Reserve, central-bank dollar liquidity swap lines; IMF COFER database; BIS triennial FX survey; TradingEconomics, U.S. Dollar Index. ============================================================================ REFERENCE GUIDE: How to Read the Dot Plot and the SEP: the Fed’s Projections, Without the Misread URL: https://l0g.fr/en/guides/read-dot-plot-sep/ Canonical French source: https://l0g.fr/guides/lire-le-dot-plot-sep/ Date: 2026-07-08 (reviewed 2026-07-08) ---------------------------------------------------------------------------- Four times a year, a chart made of dots moves global markets in a second. The Fed dot plot attracts enormous attention and generates almost as many misunderstandings as any object in macro. People read it as a rate plan, although it binds nobody; as a forecast, although it records views; as a collective decision, although it aggregates individual and anonymous opinions. The June 2026 SEP, the first under Kevin Warsh’s chairmanship, offered a live demonstration: in one quarter, the median moved from an implied cut to an implied hike. This guide explains what the dots say, and what they absolutely do not say. What the SEP is, and where the dot plot sits The Summary of Economic Projections is published by the Fed four times a year, in March, June, September and December, alongside the monetary-policy decision. It gathers each FOMC participant’s projections for four variables: real GDP growth, unemployment, headline PCE inflation and core PCE inflation, plus the policy rate each participant judges appropriate. The projections cover the current year, the following two years and the longer run. The dot plot is the most watched part of the document. Each dot represents, for a given year-end, the federal funds rate a participant judges appropriate, rounded to the nearest eighth of a point. One governance detail changes the whole reading: nineteen participants project · seven governors and twelve Reserve Bank presidents · while only twelve vote in a given year. All dots therefore carry equal weight in the median, whether they come from a voter or not, and none is named. You cannot know which dot belongs to the chair. The Fed publishes three readings for each variable. The median is the middle point once projections are ranked. The central tendency excludes the three highest and three lowest values. The range keeps everything. Markets obsess over the median, but the distance between those three readings shows whether the committee is united or divided. June 2026: a one-quarter flip On June 17, 2026, the Fed kept its policy-rate target range at 3.50% to 3.75%, by unanimous vote, unchanged since the last cut in December 2025. It was the first meeting chaired by Kevin Warsh, sworn in on May 22. Notably, Warsh submitted no dot: the June plot reflected eighteen of nineteen participants, the new chair being an open critic of the tool and having launched a task force to review it. The content surprised markets. The median policy rate for end-2026 came in at 3.8%, versus 3.4% in March. In one quarter, the message inverted: the March median still implied a cut by year-end, while the June median implied a hike relative to the current rate. Of the eighteen 2026 dots, one saw a cut, eight saw no change, and nine saw at least one hike. The reason was visible in the other variables: the median 2026 PCE inflation projection jumped from 2.7% to 3.6%, core PCE from 2.7% to 3.3%, while growth was lowered from 2.4% to 2.2% and unemployment to 4.3%. Seventeen of the eighteen participants saw inflation risks tilted upward. Reading the path, neutral rate and market gap Beyond the current year, the dot plot draws a path. In June 2026, the median peaked at 3.8% at end-2026, then slipped to 3.6% at end-2027 and 3.4% at end-2028: the committee saw a near-term tightening partly unwound later. The far-right point, the longer-run projection, deserves special attention. It represents the level to which each variable would converge under appropriate policy and in the absence of new shocks. For the policy rate, it is the estimate of the neutral rate, the famous r-star, neither accommodative nor restrictive. In June 2026 it sat around 3%, well below the current rate: the Fed judged itself restrictive. The most useful reflex is not to read the median alone, but to compare it with the market. Fed funds futures, summarized by tools such as FedWatch, show the path markets actually price. Before the June meeting, that market expected no 2026 cut and leaned toward a hike by year-end. The median dots converged toward that reading. When dots and market align, the signal strengthens. When they diverge, one of them is wrong, and that gap becomes the thing to watch. Positioning in rate futures, visible through the CFTC COT report, usefully completes the read. Reading traps The first trap is the biggest: taking the dot plot for a plan. The Fed says it plainly: these projections are not forecasts of the most likely policy-rate outcome, but each participant’s assessment of appropriate policy. No commitment is made, and the committee does not vote on the dots. A dot plot can show two hikes and the Fed can deliver none if the data change. The second trap is overinterpreting the median while ignoring dispersion. In June 2026, the 2026 dots ranged from 3.4% to 4.4%, a sign of a deeply divided committee. A stable median can hide widening disagreement, often a precursor to noisier communication. Central tendency and range exist precisely to measure that disagreement. The third trap is the tool’s instability. The move from March to June 2026, with the median jumping from 3.4% to 3.8%, reminds us that every dot plot is a revisable snapshot, conditional on each participant’s forecast. A higher dot does not necessarily mean a more hawkish Fed; it may simply reflect a higher inflation forecast. The same rate path can mean two opposite things: sticky inflation forcing restraint, or a robust economy permitting it. The path alone does not decide. Finally, the tool itself is not sacred: Warsh’s task force could modify its form, or even its principle, by late 2026. Reading the SEP in practice Everything is public and free on the Fed’s website on meeting day at 2:00 p.m. Eastern time, with the statement followed by the press conference thirty minutes later; dates should be checked directly with the Fed, listed on our reference sites page. The efficient reading sequence is simple. Compare the new median with the previous one, never in isolation. Read dispersion as much as the median. Read the other variables, because the rate path follows inflation and growth projections. Compare the rate path with the neutral-rate estimate to judge restrictiveness. Then confront the dots with market pricing. The dot plot should not be read alone. It depends on inflation data, covered in detail in the guide on CPI and U.S. inflation, while remembering that the Fed’s target is PCE. It also drives the other major central-bank lever, the balance sheet, explained in the guide to H.4.1, especially as Warsh also opened a review of the ample-reserves regime tied to system liquidity. For the context of this first meeting under the new chair, see our analysis of Warsh’s first FOMC. Read properly, the dot plot remains useful: it reveals the committee’s reaction function, dispersion and direction of revision. Read badly, it becomes a false promise, mistaken for a decision calendar when it is only a photograph of opinions at a point in time. The June 2026 SEP captures the difference: the Fed did not decide to hike. It said that a majority of participants would judge a hike appropriate if inflation behaved as they projected. Between those two statements lies the full distance between a projection and a commitment. --- Main sources: Federal Reserve, FOMC projections, June 17, 2026, accessible version; Federal Reserve, June 2026 SEP projection tables; FRED Blog, “FOMC Summary of Economic Projections, June 2026”; CNBC, “Fed holds rates steady,” June 17, 2026; Bondsavvy, “June 2026 Fed Dot Plot”; StockTitan, “Fed Holds Rates June 2026; Dot Plot Flips to a Hike.” ============================================================================ REFERENCE GUIDE: How to Read Liquidity: Reserves, TGA, RRP and the Net-Liquidity Proxy URL: https://l0g.fr/en/guides/read-net-liquidity-tga-rrp/ Canonical French source: https://l0g.fr/guides/liquidite-tresor-dts-tga-rrp/ Date: 2026-07-08 (reviewed 2026-07-08) ---------------------------------------------------------------------------- Market liquidity is not a mood. It is plumbing. Three accounts, sitting around the Federal Reserve and the U.S. Treasury, push cash into and out of the system: bank reserves, the Treasury General Account and reverse repos. A popular formula combines them into a “net liquidity” indicator that thousands of traders watch as a risk-asset barometer. This guide explains the real mechanics, shows how to read the primary sources, and marks exactly where the proxy misleads. The formula comes first because it is the one everyone quotes. Net liquidity is most often calculated as the Federal Reserve balance sheet minus the balance of the Treasury General Account and the amounts placed in the overnight reverse repo facility. When the Treasury account or reverse repos fall, cash returns to the system and net liquidity rises. When they swell, the reverse happens. Everything else follows from understanding these three taps. The three taps The first tap is the stock of bank reserves: deposits that banks hold at the Federal Reserve. They are the ultimate settlement liquidity, the fuel banks use to settle with each other. When the Fed buys securities, it credits reserves and injects liquidity; when it lets securities mature without reinvestment, it withdraws it. That is the mechanism behind QE and QT. The second tap is the Treasury General Account, or TGA: the federal government’s checking account at the Fed. Its mechanics are counterintuitive but decisive. When Treasury issues debt and builds cash, that cash leaves the banking system and sits in the TGA, draining reserves. When Treasury spends, it reinjects cash. A TGA rebuild after a debt-ceiling standoff is therefore a powerful liquidity drain, often underestimated. The third tap is the overnight reverse repo facility, or RRP, where money-market funds place excess cash with the Fed against securities. While it was full, RRP acted as a buffer: it absorbed excess cash and protected bank reserves from shocks. Once drained, it no longer plays that role, and every stress lands more directly on reserves. The net-liquidity formula, and its limits The attraction of the formula is real. Over long periods, the net-liquidity proxy has often moved in the same direction as risk assets, making it a popular macro lens among money-market analysts. When liquidity flows in, it tends to look for return. When liquidity flows out, the most liquidity-sensitive assets usually suffer first. Honesty requires stating the limits, because this is where the tool becomes a trap. Net liquidity is a proxy, not a law. Its correlation with markets is loose, unstable and regime-dependent. The formula itself varies by author: some start from the total balance sheet, others from reserves only, changing the level. It ignores money velocity, private credit, foreign flows and off-balance-sheet leverage, channels that can dominate in the short run. Above all, it can fail at turning points, when markets anticipate the plumbing rather than passively absorb it. Net liquidity illuminates a background regime. It is not an entry signal. Reading the plumbing from primary sources The good news is that all of this is public, free and daily or weekly. Three sources are enough to rebuild net liquidity yourself, without an intermediary. The Treasury account is visible day by day in the Daily Treasury Statement, published by the Bureau of the Fiscal Service. It details federal cash receipts, outlays and the operating cash balance. The Fed balance sheet, reserves and RRP are in the weekly H.4.1 release, “Factors Affecting Reserve Balances.” Policy and money-market rates, including the reverse repo rate, effective fed funds, SOFR and other repo rates, are published daily by the New York Fed. If you want ready-to-use series, the St. Louis Fed’s FRED database aggregates them. Rebuilding the indicator by hand remains the best way to understand it and avoid becoming its prisoner. The 2026 regime: QT over, RRP drained The backdrop changed at the end of 2025, and any liquidity reading must integrate it. After shrinking its balance sheet from a peak of roughly $8.9 trillion in 2022 to about $6.5 trillion, the Federal Reserve ended quantitative tightening on December 1, 2025, earlier than expected, having unwound only about half of the pandemic-era expansion. At the same time, RRP, which peaked above $2.5 trillion in 2023, fell close to zero. The consequence is structural: the shock absorber is gone, and reserves, near $2.8 trillion in late 2025, a more than four-year low, now absorb shocks alone. The Fed targets “ample” reserves, roughly 10% to 11% of GDP, versus around 13% today. When the plumbing clogs: stress signals When reserves approach scarcity, even a small shock moves funding rates, and that is the signal to watch. The reference episode remains September 2019, when repo rates jumped from about 2% to nearly 10% overnight after tax payments and Treasury settlements drained reserves, forcing the Fed to intervene. More recently, on October 31, 2025, the Fed injected about $29.4 billion through its standing repo facility, the largest one-day intervention since the early 2000s, during month-end stress with reserves at multi-year lows. The leading indicators are well known: an effective fed funds rate moving toward, or above, the rate paid on reserves; repo rates widening; quarter-end pressure. A January 2026 Fed note on the “balance-sheet trilemma” formalizes it: as reserves fall relative to the stock of public debt, repo rates become more sensitive to TGA moves, Treasury issuance and quarter-end dates. The standing repo facility is now the ceiling designed to prevent another 2019. Why markets care The issue is not technical trivia. A liquidity drain, whether from a TGA rebuild or an issuance shock, first transmits through funding markets, then through risk assets, via the cost and availability of collateral. This is also where public debt and private money meet: stablecoin issuers, now large buyers of Treasury bills, add a new source of demand at the short end, while a massive TGA rebuild can drain the system at the worst moment. Following liquidity is not about finding a buy signal. It is about knowing the calendar of major cash movements · debt ceiling, tax dates, quarter-ends · so you do not mistake a plumbing stress for a monetary-policy turn. Reading the risk, step by step A few habits are enough. Follow the TGA daily through the Daily Treasury Statement and anticipate rebuilds after debt-ceiling deals. Read RRP and reserves in the weekly H.4.1 to locate the remaining buffer. Watch the gap between repo rates and the reserve rate, as well as effective fed funds, as scarcity thermometers. Mark quarter-ends and major tax dates, classic stress windows. And keep the net-liquidity formula for what it is: a trend compass, never a trading trigger. The FOMC steers rates, but the plumbing is read in these three accounts. Methodology This guide relies on primary sources: Federal Reserve releases and H.4.1, the Fed’s January 2026 note on the balance-sheet trilemma, the Congressional Research Service’s work on the Fed balance sheet, reference rates published by the New York Fed, and the Daily Treasury Statement from the Bureau of the Fiscal Service. Balance-sheet, reserve and RRP levels are dated late 2025 and early 2026; they change, hence the revision date. Practical liquidity work on l0g.fr rebuilds the indicator from these series rather than trusting a prepackaged number, and separates plumbing stress from a true monetary-policy pivot. --- Main sources: Federal Reserve, H.4.1 “Factors Affecting Reserve Balances” and monetary-policy releases; Federal Reserve, FEDS Note “The Central Bank Balance-Sheet Trilemma,” January 14, 2026; Congressional Research Service, “The Federal Reserve’s Balance Sheet”; Federal Reserve Bank of New York reference rates, including EFFR, ON RRP, SOFR and TGCR, plus standing repo facility operations; U.S. Treasury, Bureau of the Fiscal Service, Daily Treasury Statement; FRED, Federal Reserve Bank of St. Louis, for aggregated series. Figures are dated late 2025 and early 2026. ============================================================================ REFERENCE GUIDE: How to Read the Repo Market: SOFR, the Rate Corridor and Reserve Scarcity URL: https://l0g.fr/en/guides/read-repo-market-sofr/ Canonical French source: https://l0g.fr/guides/lire-le-marche-du-repo-sofr/ Date: 2026-07-08 (reviewed 2026-07-08) ---------------------------------------------------------------------------- import InfographicBars from '../../components/infographic/InfographicBars.astro'; Repo is the most important market nobody talks about. This is where banks, hedge funds and money-market funds finance themselves overnight against collateral, for trillions of dollars a day. Its median gives SOFR, the rate that replaced Libor and anchors hundreds of trillions of contracts. This guide follows the plumbing end to end: the mechanics of a repo, how SOFR is built, the corridor through which the Fed keeps it in range, and how to read rising stress. September 2019 and year-end 2025 are the case studies. Repo, the basic block of market funding A repurchase agreement, or repo, is deceptively simple: one party sells a security, almost always a Treasury, while agreeing to buy it back the next day at a slightly higher price. Economically, it is a cash loan secured by collateral, and the difference between the two prices is the interest rate. The cash lender who receives the security is doing a “reverse repo”; the security lender who receives cash is doing a “repo.” Same contract, two views. Two settings govern the risk. The first is the haircut: the lender does not advance 100% of the security’s value but slightly less, a margin that protects against collateral price moves. A low haircut allows high leverage; a rising haircut forces the borrower to find cash or sell. The second is the type of collateral. When the cash lender does not care which exact security it receives, as long as it is high quality, the trade is “general collateral” and the rate reflects the price of cash. When a specific security is in high demand, it becomes “special” and finances at a lower rate because owning that security has value. This plumbing manufactures liquidity every day. How SOFR is built SOFR, the Secured Overnight Financing Rate, is not a number quoted by a bank. It is a statistic. Every business morning around 8 a.m., the New York Fed publishes the volume-weighted median of the previous day’s Treasury repo transactions. A median, not an average, prevents a handful of extreme trades from distorting the number. That is what makes SOFR robust, based on real transactions, unlike Libor, which rested on submitted estimates and collapsed under manipulation scandals. Its strength comes from its base. SOFR rests on more than $3 trillion of daily transactions across three segments: tri-party repo, where an agent such as BNY manages collateral; cleared bilateral repo, or DVP, through FICC; and GCF. No reference rate has ever rested on such a broad base, making it almost impossible to manipulate. The short-rate corridor The Federal Reserve does not set SOFR by hand. It builds a corridor in which it should trade, and the market does the rest. The central reference is IORB, the rate the Fed pays on reserves banks hold with it. A bank has little reason to lend cash below what the Fed pays risk-free, making IORB a magnet for short rates. The corridor has two boundaries. At the bottom is the overnight reverse repo rate, ON RRP, where money-market funds can place cash with the Fed; nobody should lend cheaper elsewhere. At the top is the standing repo facility, SRF, which lends cash against Treasuries to eligible counterparties at a fixed rate; nobody should need to borrow more expensively while the window is open. Between those boundaries sit SOFR and EFFR, the effective federal funds rate, the true operational target of monetary policy. In July 2026, with a fed funds target range of 3.50% to 3.75%, SOFR was around 3.64%, just below IORB at 3.65%, well inside the corridor. The thermometer: SOFR minus IORB Within this system, one number concentrates the information: the spread between SOFR and IORB. In an ample reserves regime, cash is abundant, lenders compete, and SOFR sits at IORB or a basis point below it, as in early July 2026, when the gap was about -1 bp. When the gap narrows and turns positive, reserves are becoming scarce: borrowers are willing to pay more than what the Fed offers to get cash. A SOFR rate persistently above IORB is the first symptom of a system short of liquidity. The same mechanism appears one layer away in regional-bank liquidity ratios and in repo-funded gilt stress in the United Kingdom. SOFR-IORB is to secured funding what a thermometer is to fever: it does not tell you the cause, but it tells you there is one. When the mechanics break: September 2019 The reference episode remains September 17, 2019. In two days, two drains combined: settlement of roughly $54 billion of Treasuries and quarterly corporate tax payments, which drained about $120 billion of reserves. The system was already tight after years of Fed balance-sheet reduction. Result: SOFR jumped from 2.43% the previous day to 5.25% on September 17, with intraday trades near 10%, roughly twice the corridor ceiling. The lesson of 2019 is not merely that an accident happened. It happened without anyone seeing it coming, because the boundary between “ample” and “scarce” reserves is not visible in advance. The Fed created the SRF in 2021 to avoid depending on ad hoc interventions and to cap rates before a blow-up. Calendar shocks Since then, stress has not disappeared; it has migrated to balance-sheet dates. At quarter-end, and especially year-end, banks shrink balance sheets to manage regulatory ratios and lend less cash to the market, lifting rates for a few days. Year-end 2025 was clear: SOFR jumped to 3.87% on December 31, with trades at 4.0%, well above IORB at 3.65%. Counterparties drew $75 billion from the SRF that day, before the stress dissolved in early January and usage fell back to zero. A smaller warning appeared in mid-September 2025, when SRF usage reached about $18.5 billion in one day, its largest draw since creation. Those jumps are not alarming in themselves. They are seasonal and usually fade. What matters is their amplitude and frequency, which reveal how thin the reserve cushion has become. If SRF usage becomes ordinary rather than exceptional, the system has moved from overflowing cash to cash that must be fetched at the window. The 2026 regime That shift is the central issue in 2026. The Fed ended quantitative tightening on December 1, 2025, stopping balance-sheet run-off so it would no longer drain reserves. At the same time, ON RRP, long a buffer above $2 trillion, fell close to zero: the cushion that absorbed shocks disappeared, and every drain now hits bank reserves more directly. In parallel, the Fed strengthened the SRF, removing the aggregate cap and moving to full allotment so it can function more clearly as a backstop. The 2026 regime is therefore a system with fewer margins. Reserves are no longer superabundant, merely “sufficient,” a more fragile state in which SOFR-IORB and SRF usage become the indicators to watch closely. This is the same thread followed in our guides to Treasury liquidity and the Fed balance sheet, of which repo is the daily extension. Reading the market in practice Read properly, repo is not a rate but an early-warning system. Repo mechanics and haircuts say how much leverage is available. SOFR and volume show the price and depth of secured funding. The IORB-RRP-SRF corridor shows where the Fed wants to hold short rates. SOFR-IORB, together with SRF usage, says when the mechanics are tightening, often before stress becomes visible elsewhere. In 2019 and again at year-end 2025, repo spoke first. This framework matters when leverage accumulates in Treasuries through the basis trade, funded in repo, and it mirrors the Treasury market itself. The glossary defines the acronyms. Quarter-end dates should be checked with the infrastructures and authorities listed on our reference sites page. --- Main sources: Federal Reserve Bank of New York, SOFR data and methodology; Federal Reserve policy rates and target range; MacroMicro, SOFR-IORB spread; Federal Reserve FEDS Notes, “What Happened in Money Markets in September 2019?”; Office of Financial Research, “Anatomy of the Repo Rate Spikes in September 2019”; Federal Reserve, end of balance-sheet run-off on December 1, 2025; Wolf Street, year-end 2025 SRF usage. ============================================================================ REFERENCE GUIDE: How to Read the Treasury Market: Auctions, Curve and Term Premium URL: https://l0g.fr/en/guides/read-us-treasuries-market/ Canonical French source: https://l0g.fr/guides/lire-le-marche-des-treasuries/ Date: 2026-07-08 (reviewed 2026-07-08) ---------------------------------------------------------------------------- import InfographicBars from '../../components/infographic/InfographicBars.astro'; The U.S. Treasury security is the most important asset in global finance, and one of the most poorly read. People speak about “the ten-year yield” as if it were a thermometer, without always knowing what it measures or how it is formed. This guide walks through the Treasury market from the anatomy of the debt to auction reading and the decomposition of long yields. The year 2026, marked by the return of term premium, provides the illustration. The anatomy of the debt: bills, notes, bonds Marketable U.S. federal debt exceeds $28 trillion, making it by far the largest bond market in the world. It is made of several instruments that differ by maturity. Bills mature in one year or less, are issued at a discount and pay no coupon; they represent roughly 22% of the market. Notes, from two to ten years, are the largest block, around 52%. Bonds, at twenty to thirty years, account for roughly 17%. Add more specialized instruments: TIPS, indexed to inflation, and two-year floating-rate notes, introduced in 2014. This structure has a direct consequence: the federal government must refinance constantly. In fiscal year 2026 alone, Treasury must replace roughly $9.7 trillion of maturing securities, on top of financing the current deficit. The average interest rate paid on the debt was about 3.4% in spring 2026, rising slowly as old low-rate debt is replaced by more expensive new debt. How the government issues: refunding and auctions Treasury does not issue randomly. It follows a calendar. Each quarter, it publishes its Quarterly Refunding Announcement, which sets issuance sizes by maturity. A rising share of bills, for example, signals that the government is financing more through the short end, a choice with heavy consequences for the curve. Issuance happens through a single-price auction, often called a Dutch auction. Bidders submit yields, and Treasury accepts bids from the lowest yield upward until the offered amount is filled. All winners then receive the same yield, the stop-out yield. Three bidder categories split the auction. Primary dealers, a group of banks designated by the New York Fed, must participate in each auction and act as market makers. Direct bidders are institutions bidding directly. Indirect bidders bid through an intermediary, and this bucket is often used as a rough proxy for foreign and central-bank demand. Before the auction, the security already trades in the “when-issued” market, which gives the reference price. Reading an auction: three signals An auction is read through three numbers, and knowing how to interpret them separates the observer from the tourist. The first is bid-to-cover, the ratio of total bids received to the amount sold. It is a volume signal: the higher it is, the broader the demand. Because primary dealers must bid, the key question is whether it sits above 2, below which appetite looks weak. The second is the tail, the gap between the auction yield and the when-issued yield just before the auction. It is a price signal: a zero or negative tail means strong absorption; a 1 · 2 bp tail means soft demand; above 3 bp on a long maturity, the auction starts to worry people. The third is the dealer takedown, the share absorbed by primary dealers. When dealers take more than 20 · 25% of the issue, end investors · direct and indirect bidders · have stepped back, leaving market makers with inventory to distribute. A strong auction combines a solid bid-to-cover, a near-zero tail and a low dealer share. The yield curve, a leading signal Link the yields of different maturities and you get the yield curve, whose shape carries dense macro information. In normal times it slopes upward: lending for longer earns more. When it flattens and then inverts, with short rates above long rates, it is a warning signal: the market expects a slowdown and future rate cuts. The New York Fed notes that the gap between the ten-year and the two-year has preceded each of the last eight U.S. recessions. In 2026, the curve exited a long inversion and steepened again. At the end of June 2026, the ten-year yield stood near 4.38% and the two-year near 4.07%, putting the curve back in positive territory. A steepening driven by long yields rising faster than short yields has a name: bear steepening. It often points to concerns about debt supply and inflation rather than a simple easing cycle. The direction of the slope matters as much as its level. Decomposing a long yield: term premium A long yield is not a single object. It has two parts. The first is the average expected path of short rates over the life of the bond, meaning what the market thinks the Fed will do, a path connected to the dot plot. The second is the term premium, the extra compensation demanded for locking up money for a long time and bearing the risk that rates, inflation or debt supply move against the investor. The benchmark model is Adrian, Crump and Moench at the New York Fed. The reading changed in 2026. Negative or near zero for a decade, the ten-year term premium turned positive again, around 0.73 percentage point in April 2026, while remaining below its long-term median near 1.41 percentage point over sixty-five years. Two lessons follow. First, a positive but still moderate premium can rise further if debt supply expands. Second, in this regime, a Fed rate cut does not automatically pull long yields down, because term premium can rise even as expected short rates fall. Who owns U.S. debt Demand matters as much as supply, and its composition says a lot about risk. The Federal Reserve, long a major buyer through quantitative easing, has stepped back after the end of balance-sheet reduction, covered in our guide to the Fed balance sheet. Foreign holders own roughly $8.5 trillion of Treasuries, or 28 · 30% of marketable debt, led by Japan, the United Kingdom and China. But their share is falling as debt grows faster than their holdings, a dynamic followed through TIC data. The rest is split among domestic actors with different stability profiles. Money-market funds and stablecoin issuers mostly buy short bills. Hedge funds hold enormous leveraged positions through the basis trade, an arbitrage between cash bonds and futures whose disorderly unwind can destabilize the market, as in March 2020. Demand increasingly carried by leveraged players is less reliable under stress. Stress thermometers A few instruments concentrate the information. Auction tails and bid-to-cover ratios show primary-market demand, issue by issue. The thirty-year yield captures pressure at the long end. The MOVE index measures implied rate volatility, the bond-market equivalent of the VIX: it rises when the market tightens. The U.S. sovereign CDS, insurance against technical default, moves during debt-ceiling episodes. Finally, a TGA rebuild after a debt-ceiling increase creates liquidity shocks, a mechanism detailed in our guide to Treasury liquidity. Reading the market in practice Read properly, the Treasury market is not one thermometer but a system. Debt composition and the refunding calendar show supply. Bid-to-cover, tail and dealer takedown show demand at each auction. The curve shows macro expectations. The split between rate expectations and term premium explains where a long-yield move comes from, and therefore whether it is about monetary policy or debt confidence. Stress gauges, from MOVE to CDS, say when the mechanics are seizing up. That framework matters when the balance tightens, as in our analysis of the return of term premium, where supply rises while demand becomes more fragile. The glossary defines the acronyms. Auction and refunding dates should be checked directly with the U.S. Treasury, listed on our reference sites page. --- Main sources: U.S. Treasury, Monthly Statement of the Public Debt; U.S. Treasury quarterly borrowing estimates and refunding announcement; SIFMA, U.S. Treasury Securities Statistics; Charles Schwab, “How Do Treasury Auctions Work?”; Loomis Sayles, “The Anatomy of a Treasury Auction”; Federal Reserve Bank of New York, primary dealers; New York Fed yield-curve recession FAQ; New York Fed Treasury term premia, Adrian-Crump-Moench model; U.S. Treasury TIC system for foreign Treasury holdings. ============================================================================ REFERENCE GUIDE: How to Analyze SEC Form 13F Filings URL: https://l0g.fr/en/guides/how-to-analyze-sec-13f-filings/ Canonical French source: https://l0g.fr/guides/analyser-13f-sec/ Date: 2026-07-08 (reviewed 2026-07-08) ---------------------------------------------------------------------------- Form 13F is the most cited window into the portfolios of major U.S. managers, and one of the most misread. It is a rear-view mirror: a quarterly snapshot of long positions, published up to 45 days after quarter-end, with no shorts, no hedges, no off-balance-sheet exposure. This guide explains how the filing works, how to read it on EDGAR, and how to avoid the misreadings that turn public data into a trap. In one sentence, to fix the vocabulary: Form 13F is a quarterly filing the SEC requires from institutional investment managers exercising investment discretion over more than $100 million in eligible U.S. securities. It lists their long positions within 45 days after quarter-end. Everything else, both its uses and its traps, follows from that definition. Where 13F comes from, and what it is for The form comes from Section 13(f) of the Securities Exchange Act of 1934, added by Congress in 1975. The stated goal was to increase transparency around the holdings of large institutional investors and, by doing so, strengthen confidence in the integrity of U.S. markets. The underlying idea: if individuals and regulators can see which securities the biggest players hold, the market is less opaque. Fifty years later, 13F has become raw material for an entire industry: “superinvestor” tracking websites, strategies that replicate star managers, academic research on flows. That popularity creates an illusion of precision that must be defused immediately. 13F was never designed as a real-time investment signal. It was designed as an instrument of retrospective transparency. The distinction changes everything. Who must file, and on what calendar The trigger is discretion, not wealth. Any “institutional investment manager” exercising investment discretion over at least $100 million of eligible securities, measured on the last trading day of any month during the calendar year, must file. The category is broad: investment advisers, banks, insurers, brokers, pension funds, family offices and even some companies. It covers both U.S. and foreign managers, as long as they use the means of U.S. interstate commerce. The $100 million threshold was set in the late 1970s and has never been raised. A 2020 SEC proposal to lift it to $3.5 billion did not go through. The practical consequence is that inflation and higher market values have mechanically widened the net, and 13F now captures a much larger universe of filers than the 1975 legislator likely imagined. The calendar is strict. Once the threshold is crossed, the manager files within 45 days after the end of the calendar year, then within 45 days after each of the first three quarters of the following year. In practice, deadlines fall around February 14 for Q4, May 15, August 14 and November 14. The obligation continues as long as the threshold is met, and an error found in a past filing requires an immediate amendment. What a 13F contains A filing has three blocks: a cover page, a summary page, and most importantly an information table in XML, the usable core. Each row describes a position: issuer name, security class, CUSIP, market value in dollars at quarter-end, number of shares or principal amount, nature of investment discretion and voting authority, and a decisive field, putCall, which tells you whether the row is a PUT or CALL option. The universe of “13(f)” securities is defined by an official list the SEC updates each quarter. It includes U.S.-listed equities, some options and warrants, shares of certain ETFs, and some convertible debt securities. A commonly missed point: open-end fund shares, traditional mutual funds, are not 13(f) securities and never appear. The official list is the only arbiter of eligibility, and it changes from quarter to quarter. How to read it on EDGAR, step by step Everything is public and free on EDGAR, the SEC database. The process is simple. First, identify the filer. Search for the manager by name or CIK in EDGAR, then filter filings by type. The holding report carries the code 13F-HR; a 13F-NT is only a notice indicating that another manager reports the securities on its behalf; a /A suffix marks an amendment. Then open the information table. Read positions line by line. The useful work starts there: calculate the weight of each row by dividing its value by the total reported portfolio value. That reveals real conviction, invisible in a simple alphabetical list. Finally, compare with the previous quarter. This is the most informative reading: by matching two successive filings, you identify new, exited, increased or reduced positions. A manager doubling an already large line sends a stronger signal than another opening a symbolic position. The dynamic matters more than the isolated snapshot. One rigorous reflex: always check the putCall field. A row can be a put option, and therefore bearish exposure, while appearing in the value column like any holding. Counting downside protection as a bullish bet is the most common and most expensive reading mistake. The limits, without mercy This is where most analysis goes wrong, and where an honest reading adds value. 13F is useful precisely to the extent that you know its blind spots. First, the lag. A position held on March 31 may not be published until May 15. The manager has had six weeks to exit. 13F tells you where the money was, not where it is. It is a rear-view mirror, never a windshield. Second, the absence of short sales. 13F shows only long positions. Shorts and most hedges are not visible, making it impossible to infer a fund’s net exposure. A manager can show a large long position while being neutral or short overall through instruments you cannot see. In October 2023, the SEC adopted a short-position reporting regime, Rule 13f-2 and Form SHO, precisely to fill that blind spot. But after a remand by the Fifth Circuit Court of Appeals in August 2025, the SEC, on December 3, 2025, pushed first filings back to 2028. In practice, for now, the short side remains invisible. Third, the quarterly snapshot. Any round trip inside the quarter is undetectable: a manager can buy and sell a position between two snapshots without leaving a trace. 13F also ignores cash, non-convertible debt, stocks listed outside the United States, commodities and private positions. It offers a partial view centered on long U.S. equity exposure. Finally, confidential treatment. A manager can ask the SEC to delay publication of some positions, for example during an accumulation phase. Those lines are temporarily missing. Add anti-duplication rules, which mean a same position may be reported by only one manager in a group, so the nominal filer is not always the true economic holder. And quarter-end window dressing can polish the snapshot for show. Using it well: the logic of confluence A rear-view mirror is still useful when crossed with other mirrors. The right practice is never to read a 13F alone, but to compare it with signals of a different nature. Form 4 is its natural complement. Insider transactions are reported within two business days, making them much fresher and more directional than 13F: an executive buying their own shares commits personal capital almost in real time. Where 13F is slow and long-only, Form 4 is fast and signed. 13D and 13G beneficial ownership filings, triggered from 5% of capital, add an intent layer, with 13D often signaling activist purpose. Confluence between managers matters just as much. A position simultaneously increased by several respected firms weighs more than an isolated bet, however large. That is exactly the logic used in 13FLOW , which cross-checks 13F filings and Form 4 insider filings to surface convergence between slow institutional flows and fast insider signals. None of these signals is sufficient on its own. Their overlap reduces noise. Methodology This guide relies exclusively on primary sources: the text of Section 13(f), Rule 13f-1, the form itself and the SEC Division of Investment Management’s FAQ, as well as releases and orders relating to Rule 13f-2 and Form SHO. Figures and deadlines were checked one by one. Any practical reading of 13F data on l0g.fr starts from the raw information table filed on EDGAR, recalculates weights, flags option rows, and systematically reminds readers of the 45-day lag. Because rules evolve, this page carries a last-reviewed date: a regime like Form SHO, delayed to 2028, can still be amended before then. --- Main sources: SEC Division of Investment Management, Frequently Asked Questions About Form 13F; SEC, Section 13(f) of the Securities Exchange Act of 1934 and Rule 13f-1; Investor.gov, Form 13F; SEC, Short Position and Short Activity Reporting by Institutional Investment Managers, Rule 13f-2 and Form SHO, adopted October 13, 2023; SEC exemptive order of February 7, 2025 and order of December 3, 2025 delaying first Form SHO filings to 2028; United States Court of Appeals for the Fifth Circuit, remand of August 25, 2025. The eligible securities list is the SEC’s quarterly Official List of Section 13(f) Securities. ============================================================================ REFERENCE GUIDE: How to Read a 10-K Without Drowning: the SEC Annual Report URL: https://l0g.fr/en/guides/how-to-read-10-k-sec/ Canonical French source: https://l0g.fr/guides/lire-le-10-k-sec/ Date: 2026-07-08 (reviewed 2026-07-08) ---------------------------------------------------------------------------- The 10-K is the most complete document available on a U.S.-listed company, and the most intimidating one. Dozens or hundreds of standardized pages, much of them legal boilerplate. The skill is not to read everything. It is to know where the signal hides: risk factors that changed year over year, management’s discussion, and the footnotes to the accounts. This guide gives the map and the reading order. Start with the definition. Form 10-K is the annual report a listed company files with the SEC under the Securities Exchange Act. In a standardized format, it brings together the business description, risks, audited financial statements and management’s analysis. Its standardization is its strength: it lets you compare a company with itself over time, and with its peers. When it lands, and for whom The 10-K is filed on EDGAR within a deadline that depends on company size, measured by public float. Large accelerated filers, with a public float of at least $700 million, have 60 days after fiscal year-end. Accelerated filers, between $75 million and $700 million, have 75 days. Others have 90 days. That calendar determines when information arrives, and therefore when to read it. The map of the document A 10-K is organized into four parts. Part I describes the business: business overview, Item 1; risk factors, Item 1A; unresolved SEC staff comments, Item 1B; and, more recently, cybersecurity, Item 1C, along with properties and legal proceedings. Part II is the financial core: market for the securities, Item 5; management’s discussion and analysis, Item 7; market risk, Item 7A; financial statements and notes, Item 8; and internal controls, Item 9A. Part III covers governance and executive compensation, often incorporated by reference from the proxy statement. Part IV lists exhibits. On that skeleton, three zones contain most of the signal. Where to start Do not read from page one to the end. The effective order starts with management’s discussion and analysis, Item 7, where management explains results, liquidity and known trends in its own words. Then move to risk factors, Item 1A, looking for what changed. Finish with the footnotes to the financial statements, where the details nobody highlights are buried. That order moves from narrative to evidence, not the other way around. Risk factors: new information versus boilerplate Item 1A is a trap for naive readers, because much of it is boilerplate designed for legal protection. The technique is to compare the section year over year: a newly added, removed or rewritten risk says more than the entire list. Be especially wary of a risk described as purely hypothetical when it has already materialized. The SEC has sanctioned companies for presenting known problems as conditional risks, from Mylan, hit with a $30 million penalty, to Yahoo and SolarWinds in cybersecurity. The signal is not the length of the list. It is what is specific and what moved. MD&A: the company seen from above Item 7, Management’s Discussion and Analysis, is where you hear management’s voice. Regulation requires companies to discuss known trends and uncertainties, liquidity and the analysis of results. It is valuable twice over: for what is said, the framing management chooses, and for what is not said, the topics it minimizes or skirts. A serious reading confronts the narrative with the numbers in the accounts. The financial statement notes: where things are hidden The real analytical work lives in the notes to the financial statements, in Item 8. That is where you inspect revenue recognition, segment data that reveal which activity really carries earnings, litigation and off-balance-sheet commitments, related-party transactions, the debt maturity schedule, and any mention of going concern. You also check the reconciliation between adjusted figures and GAAP figures, because the gap between flattering non-GAAP earnings and accounting earnings is often the most honest story in the filing. Critical accounting estimates show where management judgment weighs most heavily on the accounts. Controls and audit Two final points deserve attention. Item 9A covers internal control over financial reporting: a disclosed material weakness is a serious red flag. The auditor’s report, finally, highlights critical audit matters, the areas that required the most judgment, which are useful indications of accounting fragility. How to read it, step by step The method is a handful of moves. Retrieve the 10-K on EDGAR and the previous year’s 10-K. Read MD&A first. Then compare risk factors year over year to isolate what is new. Dive into the notes: segments, related parties, debt, litigation, going concern. Rebuild the bridge between adjusted profit and accounting profit. Finally, check controls and critical audit matters. A 10-K is not read like a novel. It is probed like a mine. Methodology This guide relies on primary sources: Form 10-K and SEC Regulations S-K and S-X, including Item 106 on cybersecurity, integrated into Item 1C for fiscal years ending on or after December 15, 2023, and Item 303 for management’s discussion. Public float thresholds and filing deadlines come from SEC definitions. Any practical 10-K reading on l0g.fr starts from the raw EDGAR filing, compares two fiscal years, and prioritizes financial statement notes over promotional text. Because rules evolve, this page carries a last-reviewed date. --- Main sources: SEC, Form 10-K and instructions; SEC Regulation S-K, including Item 303 for management’s discussion and Item 106 for cybersecurity, and Regulation S-X; definitions of filer categories, large accelerated, accelerated and non-accelerated, under Exchange Act Rule 12b-2; SEC actions relating to misleading risk factors, including Mylan, Yahoo and SolarWinds. The deadlines and thresholds cited above come from those sources. ============================================================================ REFERENCE GUIDE: How to Read CPI: U.S. Inflation, Measure by Measure URL: https://l0g.fr/en/guides/read-cpi-inflation-us/ Canonical French source: https://l0g.fr/guides/lire-le-cpi-inflation-us/ Date: 2026-07-08 (reviewed 2026-07-08) ---------------------------------------------------------------------------- No macro number moves markets faster than the U.S. CPI, and few are read so badly. A headline says “inflation is at X,” futures jump, and the actual information sits three layers deeper, in a component almost nobody checked. This guide explains what the Bureau of Labor Statistics actually measures, how the index is built, and how to move from the headline to the signal. The May 2026 release is the running example: it bundles almost every reading trap into one print. What CPI measures, and what it does not measure The Consumer Price Index is a price index published every month by the Bureau of Labor Statistics. It measures the average change over time in the prices of a fixed basket of goods and services bought by households. Prices are collected monthly in 75 urban areas, from roughly 6,000 housing units for rent measures and 22,000 retail establishments for the rest, through visits, phone calls, websites and apps. The reference base is 1982 · 1984 = 100: an index level of 335 means the basket costs 3.35 times what it did in that base period. Three versions coexist, and confusing them distorts the reading. CPI-U covers all urban consumers, more than 90% of the U.S. population; this is the number markets and the press usually cite. CPI-W covers only urban wage earners and clerical workers, about 30% of the population, but it is legally important because it is used to index Social Security benefits through the COLA. Finally, C-CPI-U, the chained index, incorporates consumer substitution across categories when relative prices change; it is first published as a preliminary estimate, then revised quarterly, and its reference base is December 1999 = 100. One methodological point matters a lot. CPI relies on a modified Laspeyres formula: a basket whose quantities remain fixed until weights are updated. That construction tends to slightly overstate the cost of living, because it does not capture in real time the fact that households shift away from products whose prices rise fastest. The chained index corrects part of that bias, and it is one reason the Fed prefers a different index, which we return to below. The basket and its weights CPI aggregates hundreds of components, but four blocks explain most of it. Services less energy services alone account for 60.3% of the basket, commodities less food and energy 19.0%, food 13.6%, and energy 7.1%. Food and energy are the most volatile blocks, and they are also the two blocks removed to calculate core inflation. Inside services, one line dominates: shelter. The shelter component represents 35.3% of the total basket, including 25.9% for owners’ equivalent rent and 7.7% for actual rent. No other component comes close. Mechanically, one tenth of a point on shelter moves the index more than a full point on apparel, which weighs 2.5%. That is why the way the BLS measures shelter, explained below, shapes almost the entire CPI reading. A clarification on energy: it is not a spending category in itself but a cross-cutting aggregate. The BLS reconstructs it from motor fuel, which sits inside transportation, and electricity and gas, which sit inside housing. Its weight can also move faster than other blocks. The base weight set in December 2025 was 6.3%, but relative importance rose to 7.1% in spring 2026 simply because energy prices rose and energy’s share of spending followed. Headline, core and supercore Three aggregates structure almost every U.S. inflation discussion. Headline, or all-items CPI, is the number in the headline: it includes everything, including food and energy. Core inflation removes those two volatile components to isolate the underlying trend. The logic is not to pretend households do not eat or heat their homes, but to strip out noise that can hide the durable price dynamic. This is the measure economists and the Fed watch to judge inflation persistence. Beyond core, central bankers also track a narrower aggregate: supercore, meaning services excluding energy and shelter. The idea is to capture the part of inflation most tied to wages and most sticky, after removing energy and the specific behavior of shelter. The BLS publishes the closest building block under “services less rent of shelter.” That indicator became central during the 2022 · 2023 tightening cycle, when the question was no longer whether goods inflation was cooling, but whether services inflation would hold firm. In May 2026, the gap between these measures was telling. Headline CPI came in at 4.2% year over year, its highest since April 2023, while core CPI stood at 2.9%. Reading only the headline that month means concluding that inflation broadly reaccelerated; reading core shows underlying inflation still near 3%, elevated but not runaway. Both readings are true. They simply do not say the same thing. Shelter, or why CPI looks in the rear-view mirror The BLS does not measure home purchase prices. In the cost-of-living framework behind CPI, an owner-occupied home is an investment asset, distinct from the shelter service it provides. What enters CPI is that service, and for owners its price is the rent they would have to pay to live in an equivalent home. That is OER, owners’ equivalent rent, used since 1987. Purchase prices, mortgage interest, property taxes and broker fees are excluded because they belong to capital, not consumption. This construction explains the best-known lag in CPI. Market rents for new leases react quickly to housing conditions. CPI shelter, by contrast, measures an average of existing rents, most of which are renegotiated only when the lease renews, and each housing unit in the sample is repriced only about every six months. When market rents accelerate or slow, CPI shelter takes many months to follow. To anticipate the turning point, analysts follow leading measures such as the New Tenant Rent Index, built by the BLS and the Cleveland Fed using only new tenants. A recent episode shows how much measurement mechanics can matter. During the October 2025 federal government shutdown, the BLS could not collect the scheduled rent sample for that month. In the absence of new observations, previous values were carried forward, temporarily understating shelter inflation. The effect reversed in April 2026, when the affected units were finally surveyed again: the accumulated increases entered all at once, lifting the monthly shelter change. In May 2026, once that catch-up had cleared, shelter returned to 0.3% on the month and 3.4% year over year, consistent with the slowdown since the 2023 peak. The lesson is simple: before reacting to one shelter month, check what the sample did that month. Reading traps The first trap is monthly change versus year-over-year change. A single month at 0.5% annualizes to about 6%, which sounds alarming, but one month does not make a trend and sampling error is real. The twelve-month rate smooths noise, but at the cost of a lag: it still carries shocks from the previous year, the base effects. A strong month dropping out of the twelve-month window can lower the year-over-year rate even as prices continue rising, and vice versa. The second trap is raw versus seasonally adjusted data. Markets react to the monthly seasonally adjusted number, which removes recurring seasonal patterns, while the year-over-year rate is published on a not seasonally adjusted basis. The BLS recalculates seasonal factors every year with January data, using the X-13ARIMA-SEATS method, and revises the previous five years. A same series can therefore change after the fact even though nothing real changed. The third trap is precision. CPI is an estimate from a sample, not a census of every price. The standard error of a monthly change in the all-items index is about 0.03 percentage point. For a published 0.2% monthly print, the 95% confidence interval is roughly 0.14% to 0.26%. A 0.1-point miss against consensus, though it moves markets, is statistically thin. Add to that the annual weight revision, based on the Consumer Expenditure Survey with roughly a two-year lag: 2026 weights reflect 2024 spending. CPI versus PCE: why the Fed watches another index This is the costliest misunderstanding for anyone following monetary policy. The Federal Reserve’s 2% inflation target is not CPI. It is PCE, the price index for personal consumption expenditures calculated by the Bureau of Economic Analysis. The Fed signaled its preference for PCE in 2000 and embedded the 2% PCE target in its formal framework in 2012, reaffirmed every year since. Three differences explain the gap. First, the formula: PCE is a chained Fisher index that incorporates consumer substitution, whereas CPI fixes its basket. Second, weights: shelter is roughly twice as heavy in CPI as in PCE, while health care is much heavier in PCE, notably because PCE counts spending made on behalf of households, such as employer-paid health insurance and Medicare or Medicaid programs. Third, scope: CPI covers only direct spending by urban households, while PCE covers a broader, urban and rural universe. As a result, since 2000 CPI has run about 0.39 percentage point above PCE on average. If the Fed hits its 2% PCE target, CPI would likely sit closer to 2.4%. The practical consequence: CPI at 3% is not necessarily a Fed failure, and CPI at 2% would already be below target once translated into PCE terms. CPI remains the number markets trade because it is more visible and released about two weeks before PCE, but PCE is what shapes the decision. Confusing the two means misreading the reaction function. How to read CPI in practice Everything is public and free. The release comes around the middle of the following month, at 8:30 a.m. Eastern time, on the BLS website; the June 2026 number is due on July 14, 2026. The efficient reading sequence is straightforward. Start with the monthly seasonally adjusted number, headline and core, to judge recent momentum. Check where the move came from by reading Table 1 line by line, not just the summary. Read shelter separately, because its weight and lag often distort the headline. Then place the whole thing inside the year-over-year rate, while keeping upcoming base effects in mind. Release dates should be checked directly with the BLS, listed on our reference sites page, and the glossary explains the acronyms used here. May 2026 bundles the lesson neatly. Headline jumped, but the component reading shows one cause: energy, whose spike came from the Middle East conflict and tensions around the Strait of Hormuz. This is exactly the transmission described in our analysis of the energy bill from the Hormuz shock and in the note on the Washington · Tehran memorandum: an oil supply shock moves into consumer prices through fuel, lifts headline, and leaves core relatively untouched. The right reflex is not to conclude that broad inflation has returned, but to separate the energy supply shock from the underlying trend. Inflation also has to be read with the rest of the macro plumbing. It cross-checks with money supply, whose growth preceded the 2021 wave, discussed in the guide on M2 money supply, and with the path of rates the Fed sets around its PCE target, visible through the Fed balance sheet and system liquidity. Read properly, CPI stops being a monthly verdict and becomes what it is: a careful but imperfect estimate of a reality nobody directly observes. The headline gives the market mood, Table 1 gives the cause, PCE gives the policy decision. Hold those three levels together and you stop being surprised by a number that “goes up” while underlying inflation did not move. --- Main sources: - BLS, “Consumer Price Index, May 2026”, June 10, 2026: headline +0.5% month over month and +4.2% year over year, core +2.9%, energy +23.5%, gasoline +40.5%, shelter +0.3% and +3.4%, energy’s contribution to the monthly increase. - BLS, CPI release Table 1, by expenditure category: relative importance of components: shelter 35.3%, OER 25.9%, energy 7.1%, food 13.6%, core 79.4%. - BLS, “Relative Importance and Weight Information for the CPI”: weights, annual update with January data, Laspeyres formula bias. - BLS, “Measuring Price Change in the CPI: Rent and Rental Equivalence”: OER definition, owner-occupied housing treated as an investment asset, exclusions. - BLS, CPI release schedule: release dates and time, next release on July 14, 2026. - Cleveland Fed, “The CPI Versus the PCE Price Index”: average gap of about 0.39 point in favor of CPI since 2000, formula, scope and weight differences. - Atlanta Fed, “What Is PCE? Explaining the Fed’s Preferred Inflation Measure”, May 20, 2026: adoption of PCE as target in 2000 and 2012, annual reaffirmation of the 2% target. - Property and Environment Research Center, Texas A&M, “One-Third of CPI: How Shelter Shapes Inflation Trends”, June 12, 2026: shelter weight, slowdown from the 2023 peak, measurement episode tied to the October 2025 government shutdown and April 2026 catch-up. ============================================================================ REFERENCE GUIDE: How to Read H.4.1: the Fed Balance Sheet, Line by Line URL: https://l0g.fr/en/guides/read-h41-fed-balance-sheet/ Canonical French source: https://l0g.fr/guides/lire-h41-bilan-fed/ Date: 2026-07-08 (reviewed 2026-07-08) ---------------------------------------------------------------------------- The U.S. central bank balance sheet is public, and it can be read once a week. The document is called H.4.1, a dry name for the most important accounting statement in the dollar system. It says which assets the Federal Reserve holds, which liabilities it carries, and how much liquidity actually circulates in the banking system. Read correctly, it lets you track quantitative tightening, anticipate funding stress, and understand why reserves rise or fall even when the Fed has done nothing. Here is how to decode it, line by line. This guide extends our reading of net liquidity. The official title of the release is “Factors Affecting Reserve Balances of Depository Institutions and Condition Statement of Federal Reserve Banks.” It is published every Thursday, normally at 4:30 p.m. New York time, with publication moved to the next business day when Thursday is a holiday. The data are as of Wednesday. The source is dual: the Federal Reserve Banks and the U.S. Treasury. What the Fed publishes every Thursday H.4.1 is not one table but a series. Table 1, the most closely watched, presents the factors affecting reserve balances: it separates what provides liquidity from what drains it. Table 2 gives the maturity distribution of securities and loans held. Table 5 is the consolidated condition statement of all Reserve Banks, the classic balance sheet where assets equal liabilities plus capital. Other tables detail each regional Reserve Bank and, during crises, special lending facilities. For regular reading, Table 1 and Table 5 are enough. The big idea: reserves are a residual This is the point most readers miss, and it is the key to the whole document. Bank reserves, meaning the deposits banks hold at the Fed, are not directly steered line by line. They are what remains after all other liability items are subtracted from the Fed’s assets. In plain English, the Fed’s holdings provide liquidity, while currency in circulation, the Treasury account and reverse repos absorb it. Reserves are the residual. That mechanics has a major consequence. The level of reserves can move sharply without any monetary policy decision, simply because the Treasury General Account fills or empties, or because reverse repo volumes change. Following reserves without following those two items is a recipe for misreading funding stress. This is exactly the mechanism explained in our guide to net liquidity. Assets: what the Fed holds Most of the asset side sits in one item, securities held outright: U.S. Treasury bills and bonds, and agency-guaranteed mortgage-backed securities. Add repurchase agreements, or repo operations through which the Fed lends against collateral, notably through the standing repo facility. Then come discount-window loans, primary, secondary and seasonal credit, and central-bank swap lines, which appear when the Fed provides dollars to foreign counterparts. A technical item, unamortized premiums and discounts, adjusts the difference between purchase price and face value of securities. Liabilities: where liquidity goes On the liability side, four items absorb most liquidity. Federal Reserve notes, the currency in circulation, form the largest share and grow slowly with the economy. Reverse repurchase agreements, including the overnight facility and foreign official accounts, temporarily drain reserves. The Treasury General Account is the federal government’s checking account at the Fed, whose movements mechanically shift reserves. Finally, deposits of depository institutions are bank reserves themselves. Reserve Bank capital completes the liability side. Reading quantitative tightening in the table H.4.1 is the best place to track quantitative tightening week after week. When the Fed lets securities mature without reinvesting them, within monthly caps, the securities-held-outright item falls and the balance sheet contracts. After peaking around $8.9 trillion in 2022, the balance sheet shrank through this run-off until the reduction stopped on December 1, 2025. Comparing Table 1 week over week, or following the series over a longer period, shows the actual pace of that movement far better than commentary. Reading traps Several precautions are required. The release is a weekly snapshot as of Wednesday and does not capture intra-week movements. Securities are shown at face value and amortized cost, not market value, so unrealized losses on the Fed’s bond portfolio do not appear in these lines. Another major subtlety: since late 2022, interest paid by the Fed has exceeded its income, creating operating losses. Rather than reducing capital, the Fed books them as a deferred asset, visible through remittances due to the Treasury, which turned negative. Until that deferred asset is worked off, the Fed sends nothing to the Treasury. Finally, accounting reclassifications and exceptional facilities can complicate period comparisons, so the table footnotes matter. The European counterpart of this release, the Eurosystem's weekly statement with its Target2 balances and loss-making central banks, is covered in our guide on the ECB balance sheet. Reading the primary source H.4.1 is free and public. It is available on the Federal Reserve website under statistical releases, in current form and weekly archives stretching far back, in HTML and PDF. For time-series analysis, the St. Louis Fed’s FRED database exposes hundreds of series derived from the release, including total balance-sheet size, reserves, reverse repos and the Treasury account, all downloadable and traceable. Since late 2025, the Fed’s interactive charts run through FRED, after the old visualization tool was retired. For rigorous use, return to the source release rather than second-hand summaries, and remember that reserves must be read alongside the Treasury account and reverse repos. This release also sheds light on the swap lines discussed in our piece on eurodollars. Methodology This guide describes the public structure of the H.4.1 release using official Federal Reserve documents and schedules. Line items, release timing and the functioning of reserves as a residual come from the release itself and Fed explanations. The end date of quantitative tightening and the balance-sheet peak are dated and sourced. No investment strategy is recommended. H.4.1 is presented as a tool for reading the central bank balance sheet, with explicit limits. --- Main sources: Federal Reserve, weekly H.4.1 statistical release “Factors Affecting Reserve Balances of Depository Institutions and Condition Statement of Federal Reserve Banks,” current release, archives and publication schedule, Thursday at 4:30 p.m. Eastern time with data as of Wednesday; Federal Reserve explanatory notes for the release, including securities held outright, repos, swap lines, premiums and discounts, the Treasury General Account, reverse repos and deposits of depository institutions; Federal Reserve Bank of St. Louis, FRED, H.4.1-derived series; Federal Reserve and FOMC sources for the end of quantitative tightening on December 1, 2025 and the balance-sheet peak around $8.9 trillion in 2022. Items, dates and calendar checked against Federal Reserve documents. ============================================================================ REFERENCE GUIDE: How to Read the U.S. Jobs Report: NFP and Its Traps URL: https://l0g.fr/en/guides/read-us-employment-report-nfp/ Canonical French source: https://l0g.fr/guides/lire-le-rapport-emploi-nfp/ Date: 2026-07-08 (reviewed 2026-07-08) ---------------------------------------------------------------------------- On the first Friday of the month, at 8:30 a.m. Eastern time, one number can move rates, the dollar and equities in a second: NFP, U.S. nonfarm payroll jobs created. It is the most traded macro data point, and one of the most poorly read. People treat it as a hard fact, when it comes from an incomplete survey, carries a margin of error almost nobody cites, and is revised the following month. Worse, two surveys coexist inside the same report and often contradict each other. This guide shows how to move from the Friday headline to a reading that holds up. Two surveys, two stories The jobs report is not one number but two surveys the BLS publishes together. The establishment survey, known as CES, surveys roughly 119,000 businesses and government agencies each month, covering about 622,000 worksites and around 26% of total payroll employment. It produces NFP, hours worked and wages. It counts jobs: one person holding two jobs is counted twice, while farmers and self-employed workers are excluded. The household survey, known as CPS, surveys about 60,000 households and classifies every person aged 16 or older as employed, unemployed or not in the labor force. It produces the unemployment rate, participation rate and broader underemployment measures. It counts people, not jobs. These two employment measures often diverge, and that is normal: one counts jobs, the other individuals. In May 2026, they fortunately pointed in the same direction, with 172,000 jobs created on the establishment side and 149,000 more employed people on the household side, while unemployment stayed at 4.3%. But in a divergent month, knowing which measure carries the relevant signal avoids confusion: for hiring momentum, use NFP; for unemployment, use the household survey. Reading NFP, and its margin of error NFP measures the net monthly change in nonfarm payroll employment. In May 2026, it came in at 172,000, far above consensus near 80,000, and was read as a robust labor market that pushed rate-cut expectations further away. The number includes a statistical adjustment, the birth-death model, which estimates jobs created or destroyed by firms that are born or die before appearing in the sample. The central trap is ignored by almost every comment: NFP is an estimate, not a count. The BLS states that the 90% confidence interval for the monthly change is roughly plus or minus 122,000. In other words, the May print at 172,000 is, with 90% confidence, really somewhere between about 50,000 and 294,000. Any print below 122,000 is not statistically distinguishable from zero. On top of that, the first estimate is based on only about 60% of expected responses, versus 70% before the pandemic. The number delivered as fact on the first Friday is therefore a first approximation, based on a fraction of a survey that covers only a quarter of employment. Revisions, or why the first number often lies The report is corrected in two ways, and this is where much of the signal sits. First, each month, the previous two months are revised as late responses come in. In the May report, March was revised up from 29,000 to 214,000, and April from 64,000 to 179,000, adding 93,000 more jobs than initially reported. Revisions can go both ways, and they often move more than the current month’s surprise. Second, once a year, the establishment survey is benchmarked to near-exhaustive unemployment-insurance records. The gap between the March survey estimate and the actual count is used as an approximate measure of total error. That correction has been huge in recent years. The preliminary benchmark revision for March 2025 subtracted 911,000 jobs, or 0.6% of the total. Once the process was finalized in early 2026, 2025 job creation was cut from 584,000 to 181,000, a reduction of more than 400,000 jobs: the actual pace was only about 15,000 per month, versus 48,000 shown in real time. For much of 2025, the reported labor market was almost three times stronger than the reality later confirmed. Reading NFP always means remembering that the first print can be erased. Unemployment: U3, U6 and the participation trap The headline unemployment rate is U3: the number of unemployed people divided by the labor force. In May 2026, it was 4.3%. But there is a broader measure, U6, which adds discouraged workers and people forced into part-time work because they cannot find full-time jobs; it stood at 8.1%. The gap between the two says something about labor-market quality, not just volume. The most counterintuitive trap concerns the participation rate, the labor force as a share of the total population, at 61.8% in May. The unemployment rate can fall for two opposite reasons: because unemployed people find work, which is good news, or because discouraged people stop looking and disappear from the statistics, which is bad news. A falling unemployment rate driven by weaker participation is not a sign of health. Always read unemployment together with participation: one without the other is easy to misinterpret. Wages and hours remain, often neglected. In May, average hourly earnings rose 0.3% on the month and 3.4% year over year. But against 4.2% inflation, that gain was negative in real terms: hourly purchasing power fell. The average workweek, stable at 34.3 hours, is a useful leading indicator because employers often adjust hours before headcount. Reading the report in practice Everything is public and free on the BLS website, on the first Friday of the month at 8:30 a.m. Eastern time; the June 2026 report is released on July 2. The BLS even publishes a table showing which household-survey changes are statistically significant, a healthy reflex before reacting. The reading method rests on a few principles. Never read NFP without its margin of error or the revisions to the previous two months. Cross-check the two surveys instead of isolating one. Read unemployment with participation. Above all, follow the trend over several months rather than the Friday surprise, because the trend, not the surprise, survives revisions. Two cautions on reliability. The October 2025 household survey could not be collected because of the partial federal government shutdown, a gap still visible in the series. And the institutional context has become tense: in summer 2025, after disappointing numbers and large downward revisions, BLS leadership was reshuffled, feeding debate about the independence of public statistics. None of this invalidates the data, but all of it argues for reading it as what it is: an imperfect estimate, not a verdict. The jobs report should not be read alone. It cross-checks with inflation: the guide on CPI and U.S. inflation shows why wages at 3.4% under inflation at 4.2% erode purchasing power. It weighs directly on the Fed, whose labor-market reading feeds the dot plot and SEP: the resilience of employment helped shift the median rate path toward a hike in June 2026. And positioning before the release can be read through the CFTC COT report. For the broader process, see the site’s methodology; release dates should be checked directly with the BLS, listed on our reference sites page. Read properly, the jobs report remains the best monthly X-ray of the U.S. labor market. Read badly, it becomes a weather vane: people overreact to a number whose margin of error makes it uncertain, whose next-month revision moves it, and whose benchmark revision can erase it. The signal is not in the Friday headline. It is in the trend, revisions and consistency between the two surveys. The rest is noise that markets trade anyway. --- Main sources: - BLS, “The Employment Situation, May 2026”: NFP +172,000 in May, March and April revised to +214,000 and +179,000, average hourly earnings +0.3% month over month and +3.4% year over year, workweek at 34.3 hours, next release on July 2, 2026. - BLS, Employment Situation technical note: two surveys, CES and CPS, birth-death model, benchmark revision using unemployment-insurance records, 90% confidence interval for monthly NFP change of plus or minus 122,000. - BLS, Current Employment Statistics program: preliminary March 2025 benchmark revision of −911,000, or −0.6%, real average hourly earnings down 0.1% in May, publication schedule. - BLS, household survey, Employment Situation A tables: U3 unemployment at 4.3%, broad U6 at 8.1%, participation rate at 61.8%, long-term unemployed and discouraged workers. - Real Investment Advice, “BLS Jobs Report Is Broken?”: 2025 cut from +584,000 to +181,000 after benchmark, actual pace around 15,000 per month versus 48,000 reported, response rate falling toward 60%, sample covering 26% of employment. - CNBC, “Jobs report May 2026”: print well above consensus, U6 slightly lower at 8.1%, household employment up 149,000, context around BLS leadership change in summer 2025. ============================================================================ REFERENCE GUIDE: Reading the gold market: London, Comex, paper versus physical URL: https://l0g.fr/en/guides/read-gold-market/ Canonical French source: https://l0g.fr/guides/lire-le-marche-de-l-or/ Date: 2026-07-08 (reviewed 2026-07-08) ---------------------------------------------------------------------------- Gold is the oldest monetary asset and one of the most misread. It is treated as a commodity, a currency, a relic or a hedge, depending on the story of the day. It is all of those, but not in the same way. This guide explains where gold's price is formed, how paper and physical markets connect, what really drives the price, and why central banks have become decisive buyers. The 2026 cycle, a record followed by a brutal correction, is the case study. A monetary asset, not a normal commodity Gold is mined, refined and stored, but it differs from ordinary commodities in the most important respect: it is barely consumed. Most of the gold ever mined, around 210,000 tonnes, still exists in jewellery, bars, coins or reserves. Its price is therefore not mainly a production-consumption balance like oil. It is the price of desired stock. Gold is an asset of hoarding, not an industrial flow. Its second feature is that it yields nothing. No coupon, no dividend. Holding gold has an opportunity cost: the return you give up by not holding bonds. That is why gold was long read as the inverse of real rates. When real rates rise, gold should fall. That model worked for decades, then weakened after 2022, when gold rose despite high real rates. Another driver had taken over. Where price is formed: London and Comex Gold price formation is split between two major markets. The first is London, the over-the-counter market coordinated by the LBMA. Banks, refiners, miners and central banks trade physical “loco London” gold deliverable in London vaults. This is the anchor of the physical market, with daily turnover around $180 billion. Twice a day, at 10:30 and 15:00 London time, an electronic auction administered by ICE sets the LBMA Gold Price, a global benchmark used to value reserves, ETFs and contracts. The second is the Comex in New York, a futures market where promises of metal trade with leverage. Comex does not anchor the physical market; it amplifies expectations and positioning. Shanghai now matters too, as Asian demand and pricing power rise. London anchors physical metal. Comex sets the tempo. Shanghai's weight is rising. Paper versus physical The link between Comex futures and London physical is the key plumbing. Moving from a Comex futures position to physical gold in London uses an exchange-for-physical transaction, or EFP. The spread normally stays narrow. When demand for deliverable metal rises sharply, the EFP can widen, signalling stress between paper promises and physical availability. The gold lease rate is another pressure gauge. Physical gold can be lent out. When metal becomes scarce, lease rates rise, which can pull futures below spot and drain deliverable liquidity. Watching EFP spreads and lease rates tells you where the stress is: in the displayed price or in the physical metal behind it. What really drives gold Four forces drive gold, and their hierarchy changes. Real rates remain the classic opportunity-cost variable. The dollar also matters: a weaker dollar makes dollar-priced gold cheaper for the rest of the world, a mechanism covered in the dollar guide. Safe-haven demand appears during geopolitical shocks. But since 2022, the dominant driver has been debasement risk: distrust of paper currencies amid exploding public debt, sanctions risk and reserve diversification. That is why gold rose even when real rates were high. If gold rises without real rates or the dollar explaining it, look for official-sector demand and debasement fear. Central banks: the marginal buyer Since 2022, central banks have become the marginal force. Their purchases roughly doubled compared with the previous decade, averaging about 225 tonnes per quarter from 2021 to 2025. The logic is simple: diversify reserves away from the dollar and hold an asset that cannot be frozen by a foreign sovereign. That engine slowed in early 2026. Declared net purchases fell to only 16 tonnes in the first quarter, with some central banks selling, including a large Turkish sale in March. But not all official purchases are reported to the IMF, so real demand can be partly hidden. Reading central-bank gold demand requires crossing declared reserve data with import and physical-flow estimates. Gold as the second reserve asset Official gold holdings have reached nearly 39,000 tonnes, worth about $5 trillion, making gold the second-largest reserve asset behind the dollar and ahead of the euro. This is one visible face of dedollarisation: not an abandonment of the dollar, but gradual diversification toward a neutral asset. The move must be measured carefully. Gold's reserve share rises both because central banks buy tonnes and because the price of gold rises. Volume effect and valuation effect are different. 2026: record, then correction The 2026 cycle shows the volatility of a sentiment-driven, non-yielding asset. After a near-vertical rise, gold touched a record close to $5,600 per ounce in January 2026 before correcting toward $4,150 by early July. That is a drop of more than 25% from the peak. The correction coincided with slower reported central-bank buying and a dollar that remained central as a funding currency. The debasement story did not vanish, but the official-demand engine paused, and price reflected it. Reading gold in practice Gold is not a relic; it is a monetary barometer. London versus Comex tells you where price and physical stress are formed. EFP spreads and lease rates reveal paper-physical tension. Real rates, the dollar, safe-haven flows and debasement risk explain the movement. Central-bank demand tells you whether the marginal buyer is still there. Gold predicts nothing. It measures how much confidence the world gives, or withdraws from, paper money. This guide complements the oil market guide for commodity-market method and the dollar guide for the monetary background. --- Main sources: World Gold Council, Gold Market Primer: Market Size and Structure; World Gold Council, Gold Demand Trends; LBMA Precious Metals Market Report; LBMA OTC guide on futures markets and exchange-traded products; TradingEconomics gold price data for the 2026 peak and correction. ============================================================================ REFERENCE GUIDE: How to Read JOLTS: under the hood of the U.S. labor market URL: https://l0g.fr/en/guides/read-jolts-report/ Canonical French source: https://l0g.fr/guides/lire-le-rapport-jolts/ Date: 2026-07-08 (reviewed 2026-07-08) ---------------------------------------------------------------------------- import InfographicBars from '../../components/infographic/InfographicBars.astro'; The payroll report gives the net balance of jobs, a single number that hides everything underneath. JOLTS opens the hood: how many jobs are open, how many workers were hired, how many quit voluntarily, and how many were laid off. It was one of Jerome Powell's favored gauges of labor-market tightness. The frozen 2026 labor market makes it especially useful. What JOLTS measures that payrolls do not The Employment Situation gives a snapshot: net jobs created or lost during the month. JOLTS gives the movie behind that snapshot. Behind a net gain of 150,000 payrolls sit millions of hires and separations, of which the NFP shows only the difference. Published by the BLS roughly a month after the payroll report, JOLTS is older data, but it reveals what the net figure cannot: labor demand, worker confidence and the nature of separations. A labor market can show a stable payroll balance while activity underneath freezes: fewer hires, fewer quits, fewer layoffs. Payrolls would miss the rotation; JOLTS shows it. The five key numbers JOLTS is read through five series. Job openings measure unsatisfied labor demand: positions open, actively recruited for, and available within 30 days. Hires count workers added during the month. Total separations include all exits, split into quits, layoffs and discharges, and other separations. In May 2026 the U.S. economy had 7.6 million openings, 5.2 million hires, 3.1 million quits, 1.7 million layoffs and discharges, and 5.1 million total separations. The quits rate: worker confidence The richest signal is the quits rate, quits divided by total employment. Workers usually quit voluntarily only when they are confident they can find something better. The quits rate is therefore a gauge of worker confidence and, indirectly, wage pressure: a market where workers can quit easily forces employers to pay to retain. At the peak of the Great Resignation in 2021-2022, the quits rate reached roughly 3%. In 2026 it was back near 2.0%, below the pre-pandemic norm of about 2.3%. Workers had become cautious. That confirms a cooling labor market before the unemployment rate necessarily moves. Job openings per unemployed worker The Fed's favored JOLTS gauge is the number of job openings divided by the number of unemployed workers. It measures labor-market tightness: how many open jobs are available for each job seeker. Before the pandemic it was near 1.2. The 2022 overheating pushed it close to 2.0. By 2026 the ratio had fallen to roughly 0.95, below even its pre-pandemic norm. The labor-market tightness had been fully drained. That is delicate: falling from 2.0 to 1.2 meant fewer openings without much unemployment. Falling further below 1.0 risks shifting from painless cooling to job losses. The Beveridge curve and the soft landing The Beveridge curve links the job openings rate and the unemployment rate. In normal slowdowns, openings fall and unemployment rises. The post-2022 soft-landing bet was that the economy could slide down the steep part of the curve: fewer vacancies, little unemployment damage. By 2026 the economy was near the point where this becomes harder. Once the vacancy ratio is close to normal, another drop in labor demand is more likely to show up as unemployment rather than just fewer job postings. That is why JOLTS matters for the Fed. Measurement traps JOLTS has all the weaknesses of a survey, and then some. Its sample is smaller than payrolls, roughly 21,000 establishments, and response rates have fallen. One month should never be over-read; the three- to six-month trend carries the signal. A job opening is also declarative: a position must be open, actively recruited for and available within 30 days, but firms can leave online postings stale or keep “ghost jobs” visible. Finally, JOLTS is lagged by a month relative to payrolls. It is a confirmation and decomposition tool, not a seconds-after-release trading signal. 2026: a frozen labor market The 2026 regime is best described as frozen. Openings around 7.6 million, hires around 5.2 million, quits low and layoffs low: fewer firms are hiring aggressively, but few are cutting yet, and workers are staying put. That configuration can turn quickly. If firms move from hiring freezes to headcount cuts, unemployment rises fast. JOLTS therefore complements the dot plot and the payroll report: it shows whether the labor market is cooling smoothly or approaching the point where cooling becomes damage. How to use it Read JOLTS as the indispensable counterpoint to payrolls. Payrolls give the net balance; JOLTS gives the flows. The quits rate tells worker confidence. The openings-to-unemployed ratio tells labor-market tightness. The Beveridge curve tells whether the economy is still shedding excess vacancies without creating unemployment. Separations, split between quits and layoffs, tell what kind of slowdown is underway. --- Main sources: BLS, Job Openings and Labor Turnover Summary, May 2026; BLS JOLTS methodology and publication calendar; FRED series for job openings and quits; MacroMicro JOLTS dashboard; Economic Policy Institute monthly JOLTS analysis. ============================================================================ REFERENCE GUIDE: How to Read PCE: the Inflation Gauge the Fed Actually Targets URL: https://l0g.fr/en/guides/read-pce-inflation-fed/ Canonical French source: https://l0g.fr/guides/lire-le-pce-inflation-fed/ Date: 2026-07-08 (reviewed 2026-07-08) ---------------------------------------------------------------------------- import InfographicBars from '../../components/infographic/InfographicBars.astro'; Markets trade CPI, but the Fed decides on PCE. Two weeks after the inflation headline, a second number is released with far less noise, and that is the one the central bank uses to calibrate rates. This guide walks through PCE end to end: what the Bureau of Economic Analysis measures, why its construction differs from CPI, and how to read it without using the wrong thermometer. May 2026 is the running example, with an apparent anomaly: core PCE above core CPI. What PCE measures, and why it is broader The Personal Consumption Expenditures Price Index is published every month by the Bureau of Economic Analysis (BEA), inside the “Personal Income and Outlays” report. It measures price changes across all goods and services consumed by US households, and the word all is what separates it from CPI. CPI only captures spending paid directly by urban households. PCE covers the whole consumption universe, urban and rural, including spending made on behalf of households by third parties. That “on behalf of households” principle is the keystone. A large share of US health care is paid by employers through health insurance, or by the government through Medicare and Medicaid. CPI only sees the patient's out-of-pocket cost; PCE counts the full expense, regardless of who pays. As a result, health care weighs roughly twice as much in PCE as in CPI, while housing, by contrast, weighs roughly half as much. That weighting gap alone is enough to make the two indexes diverge, as the May 2026 example shows. The formula: a chained index that follows the consumer The second major difference is the formula. CPI fixes the quantities in its basket until the next weight update, which tends to slightly overstate the cost of living. PCE is a chained Fisher-type index: its weights are updated continuously, so it incorporates household substitutions as consumers move away from products whose prices rise fastest and toward alternatives. When beef prices jump and households buy more chicken, PCE captures that shift almost in real time. CPI only catches up when its weights are revised. This construction has a mechanical consequence: over long periods, PCE tends to grow a little more slowly than CPI. Since 2000, the average gap has been about 0.4 percentage point in favor of CPI. That is one reason why a 2% PCE target would roughly correspond to CPI inflation around 2.4%, a gap detailed in the French CPI guide. Where the numbers come from One often overlooked point: PCE is not built from scratch. For many components, the BEA reuses prices collected by the BLS for CPI, then reweights them with its own expenditure weights and complements them with business and administrative sources, notably for health care and financial services. PCE is therefore partly reprocessed CPI, with a broader scope and a different formula. This is why markets can often anticipate PCE after the CPI and PPI for the same month are known: Cleveland Fed nowcast models reconstruct PCE with reasonable precision before the official release. Headline PCE, core PCE and market-based PCE As with CPI, the first split is between headline PCE, which includes everything, and core PCE, which excludes food and energy to isolate the underlying trend. The Fed watches core PCE closely when judging inflation persistence, even though its formal 2% target is defined on headline PCE. PCE adds its own distinction: market-based PCE. Some PCE prices are not observed in a market but imputed, meaning indirectly estimated, such as financial services measured through bank margins or some imputed rents. Market-based PCE removes those non-market components and keeps prices that are actually transacted. The Fed follows it as a cleaner inflation read, less exposed to measurement artifacts. When headline PCE and market-based PCE diverge, the driver is often an imputed item rather than a true market price. Robust measures: trimmed mean and median Removing food and energy is a blunt way to reduce noise. Two regional Federal Reserve banks publish finer filters specific to PCE. The Dallas Fed calculates trimmed mean PCE, removing each month's most extreme components on both sides of the distribution instead of always excluding the same two categories. The Cleveland Fed publishes median PCE, the price change of the component at the center of the distribution. These measures better capture the durable trend because they are less polluted by a single component that suddenly jumps. Their value is clearest when they diverge from core. In May 2026, core PCE printed at 3.4% year over year, its highest since October 2023, but the Dallas Fed trimmed mean held at 2.35% and the Cleveland Fed median at 2.83%. In other words, part of the height of core PCE came from atypical components, while the central trend was more moderate than the headline number suggested. May 2026: when core PCE moved above core CPI May 2026 condenses the lesson. Headline PCE rose to 4.1% year over year, almost level with headline CPI at 4.2%, with both driven by the same energy spike. So far, nothing unusual. But on core, the order flipped: core PCE reached 3.4% while core CPI stayed at 2.9%. PCE, usually running below CPI, moved above it. The explanation is entirely about weights. Housing, which represents roughly a third of CPI but about half as much in PCE, had been cooling since the 2023 peak. Its slowdown therefore pulled core CPI down heavily, but core PCE much less. Conversely, health care, roughly twice as heavy in PCE, started heating up and pushed core PCE higher. Two opposite forces, two opposite weights, and a crossing between the two measures. Anyone who only reads the CPI headline sees reassuring underlying disinflation; anyone who watches core PCE, the measure that matters for the Fed, sees stickier inflation. That is exactly why reading both matters for understanding the reaction function. Reading PCE in practice The “Personal Income and Outlays” report is released near the end of the month, at 8:30 a.m. Eastern Time, roughly two to three weeks after the CPI for the same month. The May 2026 PCE report was published on June 26, 2026. An efficient reading sequence is simple. Start with core PCE, the indicator the Fed targets in practice, rather than headline PCE, which is noisier because of energy. Cross-check it against robust measures, trimmed mean and median, to see whether the core number is driven by the trend or by a few atypical components. Look at market-based PCE to strip out imputation artifacts. Then compare it with CPI for the same month, because the gap between the two, especially on core, reveals which components are carrying inflation. The number only makes sense when linked to the rest of the monetary machinery. The Fed calibrates rates on its trajectory, a reaction function that can be read through the dot plot and whose effects flow through the Fed balance sheet and system liquidity. In May 2026, the jump came from an energy supply shock tied to the Middle East conflict and tensions around the Strait of Hormuz, lifting headline inflation without necessarily distorting the underlying trend, which is judged on core and robust measures. Read properly, PCE is not a duplicate of CPI but its decisive complement. CPI gives the market mood and arrives first; PCE, broader and better weighted, gives the measure on which the decision is made. Release dates should be checked directly with the BEA, listed on our reference sites page, and the glossary covers the acronyms used here. --- Main sources: - BEA, “Personal Income and Outlays, May 2026”: headline PCE +4.1% year over year, core PCE +3.4%, report structure. - BEA, Personal Consumption Expenditures Price Index: definition, scope, chained Fisher-type formula, headline/core/market-based distinction. - Federal Reserve Bank of Dallas, Trimmed Mean PCE Inflation Rate: robust measure excluding extreme components, 2.35% in May 2026, trimming methodology. - Federal Reserve Bank of Cleveland, Median PCE Inflation: median PCE at 2.83% in May 2026, central-trend reading. - Federal Reserve Bank of Cleveland, “The CPI Versus the PCE Price Index”: formula, scope and weight differences, housing and health care comparison, average gap of about 0.4 percentage point since 2000. - Federal Reserve Bank of Atlanta, “What Is PCE? Explaining the Fed's Preferred Inflation Measure”: adoption of PCE as target in 2000 and 2012, 2% target on headline PCE. - BLS, “Consumer Price Index, May 2026”: headline CPI +4.2% and core CPI +2.9% year over year, comparison points with PCE for the same month. ============================================================================ REFERENCE GUIDE: How to Read Volatility: VIX and MOVE, the market fear gauges URL: https://l0g.fr/en/guides/read-vix-move-volatility/ Canonical French source: https://l0g.fr/guides/lire-la-volatilite-vix-move/ Date: 2026-07-08 (reviewed 2026-07-08) ---------------------------------------------------------------------------- import InfographicBars from '../../components/infographic/InfographicBars.astro'; When commentators say fear is rising in markets, they usually mean the VIX, without saying what it measures. They also often forget its bond-market cousin, MOVE, which can be more informative. These indices do not say what markets have done; they say what investors are paying to insure against what may happen next. Implied volatility: looking forward Realized volatility measures past price movement. Implied volatility is extracted from option prices: it is the amount of movement the market prices for the future. Because an option is insurance, its price rises when investors demand protection. VIX and MOVE are therefore forward-looking gauges of protection demand, not backward-looking descriptions. A rising VIX means the market is buying umbrellas; it does not necessarily mean it is already raining. VIX: equity volatility VIX measures expected 30-day implied volatility on the S&P 500. Cboe calculates it from a wide strip of S&P options, without relying on a single pricing model, and expresses it as an annualized percentage. A rough translation helps: divide the VIX by about 3.5 to estimate the one-month move the market is pricing. A VIX of 16 corresponds to roughly a 4.5% expected monthly move, up or down. The VIX is asymmetric. It jumps when equities fall and protection demand spikes, then usually drifts lower during rallies. Its long-term average is around 19-20. Below 15 suggests complacency; above 30 signals stress; the largest shocks have sent it far higher, near 82 during March 2020. MOVE: bond volatility MOVE is the VIX-like gauge for U.S. Treasuries. It measures one-month implied volatility in Treasury options, with maturities along the curve and a heavy weight on the ten-year sector. Unlike VIX, it is expressed in basis points of rate volatility. Its reading grid is simple. Below 80, the bond market is calm. Between 80 and 120, volatility is moderate. Above 120, stress is significant. MOVE is also a rough gauge of uncertainty around the Fed, inflation and debt supply; when it rises, the term premium often matters more. Term structure: the signal inside the signal Level is not enough. The volatility term structure matters. VIX futures are normally in contango: three-month volatility is above spot volatility, because uncertainty grows with horizon. That is the normal calm state. When spot volatility moves above longer-dated volatility, the curve flips into backwardation: the market prices immediate stress that it expects to fade. VVIX adds another layer: the volatility of VIX options. A high VVIX with a low VIX can reveal nervousness under a calm surface, as investors buy protection against a volatility shock before spot volatility has moved. When VIX and MOVE diverge The two gauges do not always move together. In 2022-2023, aggressive Fed tightening pushed MOVE sharply above normal while VIX stayed more contained. The stress was in rates, not equities. Conversely, a stock-specific earnings shock can lift VIX without much movement in MOVE. Read together, they locate stress. MOVE rising alone points to Fed, inflation, duration or debt-supply risk. VIX rising alone points to equity risk. Both rising together, as in March 2020, points to systemic stress. Traps The first trap is confusing low volatility with low risk. Volatility is mean-reverting and short-horizon. It can stay low right before a shock. Calm markets often encourage leverage, making the system more fragile precisely because measured volatility is low. The second trap is thinking the VIX is directly investable. You cannot buy spot VIX; you can trade futures and products built on them. When the futures curve is in contango, roll costs erode long-volatility products. Short-volatility strategies earn small gains until one shock wipes them out, as in February 2018's “Volmageddon.” 2026: a surface that is too calm In mid-2026 both gauges are low: VIX around 15.8 and MOVE around 65.8. Read alone, that says serenity. Read with the rest of the dashboard, it is more ambiguous: credit spreads are at their tightest since 2007 and dollar-funding plumbing still needs watching through the cross-currency basis. The calm is therefore not a green light. It is a statement about the current price of protection. If protection is cheap while macro, credit and funding risks persist, the message is not “risk is gone”; it is “risk is not being paid much.” How to use them Read VIX and MOVE as maps of perceived risk. The level tells how much protection costs versus history. The term structure says whether stress is immediate or deferred. The divergence between the two locates the stress in equities or rates. VVIX reveals nervousness below the surface. The golden rule: read volatility with spreads, funding, liquidity and positioning. Low volatility plus tight spreads and a strained basis is not a clean all-clear. It is often a warning that the market is underpricing the cost of being wrong. --- Main sources: Cboe VIX methodology; FRED VIX series; ICE MOVE Index documentation; MacroMicro MOVE Index data; Cboe VVIX materials; historical volatility-market episodes including March 2020 and February 2018. ============================================================================ REFERENCE GUIDE: Reading the uranium market: fuel cycle, prices and bottlenecks URL: https://l0g.fr/en/guides/read-uranium-market/ Canonical French source: https://l0g.fr/guides/lire-le-marche-de-l-uranium/ Date: 2026-07-07 (reviewed 2026-07-07) ---------------------------------------------------------------------------- Uranium fascinates and confuses. It carries all the major stories of the moment: energy transition, nuclear revival, AI electricity demand. But its market works like no other: no open exchange, negotiated prices, and a bottleneck that is not always where investors look. This guide follows uranium from mine to reactor, so you know what to read and what to monitor. The fuel cycle, from ore to reactor Understanding uranium begins with the fuel cycle. Each stage has its own market and vulnerabilities. At the mine, ore is crushed and leached, often with sulphuric acid, then precipitated into uranium oxide concentrate, or yellowcake, the market's reference product quoted in dollars per pound of U3O8. Then comes conversion, which turns the oxide into uranium hexafluoride, UF6, the gaseous form required for enrichment plants. Enrichment raises the share of uranium-235, the fissile isotope, measured in separative work units, or SWU. Fabrication turns enriched uranium into fuel assemblies for reactors. The consequence is crucial: the uranium price does not tell the whole story. A country can have uranium and still depend on others for conversion and enrichment. How uranium is priced Unlike oil, uranium does not trade on a deep exchange with a continuous transparent price. Utilities and producers negotiate over-the-counter contracts. Reference prices are published by specialist firms such as UxC and TradeTech: spot, mid-term, long-term, conversion and enrichment indicators. The spot/long-term distinction is the first thing to learn. Spot covers near-term delivery, smaller volumes and more volatile flows, often amplified by financial buyers. Long-term contracts cover delivery three years or more ahead, with indexation clauses. Utilities care about the long-term price because they need secure fuel supply. In 2026, spot uranium moved above $100 per pound, the first such move in two years, but the more meaningful signal was the long-term price: around $93 at the end of March according to TradeTech, its highest level in more than eighteen years. Demand: reactors and front-loaded fuel Uranium demand follows reactor capacity, and that path is rising. The World Nuclear Association projects installed nuclear capacity rising from 398 GWe in 2025 to 746 GWe by 2040. Seventy-eight reactors were under construction globally, with China leading at around thirty-eight units. Uranium requirements would rise from about 68,900 tonnes in 2025 to more than 150,000 tonnes in 2040. New reactors also need an initial fuel load before regular refuelling begins, which pulls demand forward. AI data-centre electricity demand adds another layer, pushing large technology companies toward nuclear power and strengthening the demand narrative. Supply: concentrated, rigid, and padded by secondary supply Supply has two weaknesses. First, it is concentrated. Kazakhstan alone accounts for roughly 40% of global mined production, led by Kazatomprom. Canada, with Cameco, is another key producer. Second, supply is slow. Opening or restarting a mine takes years and capital. Production responds slowly to price. Kazatomprom cut its 2026 ambitions to 27,500 · 29,000 tonnes, below capacity, partly because of sulphuric-acid constraints. The gap between mine production, around 60,000 tonnes, and reactor needs, around 69,000 tonnes, is filled by secondary supply: inventories, reprocessing and stock adjustments. That cushion delays the imbalance; it does not eliminate it. The upstream bottleneck: conversion and enrichment This is the point the market often underestimates. Even with enough mined uranium, it must still be converted and enriched. Enrichment is an oligopoly, and Russia controls around 44% of global enrichment capacity. It historically supplied roughly 35% of U.S. enriched-uranium imports. The issue becomes sharper for advanced reactors and small modular reactors, which require HALEU, high-assay low-enriched uranium. Until 2024, Russia was the only commercial supplier. The United States is rebuilding domestic capacity, with Centrus as the only domestic HALEU producer at this stage, while the Russian import ban reaches full effect in 2028. The nuclear revival may therefore hit an enrichment bottleneck before it hits a mining bottleneck. Signals to monitor A uranium watchlist should focus on the long-term TradeTech price, not only spot. Utility contracting pace tells you when inventories are being rebuilt. Kazatomprom and Cameco production guidance shows the supply pulse. Conversion and enrichment prices reveal the real upstream stress. HALEU milestones and non-Russian enrichment capacity determine whether advanced-reactor promises become physical supply. Finally, AI electricity spending remains the demand wild card. Reading the market in practice The uranium market is not one price; it is a layered system. The fuel cycle shows where value is created and where bottlenecks appear. The long-term price shows utility conviction. Reactor demand shows the path. Concentrated and rigid supply shows fragility. Conversion and enrichment show where the system can seize up before mine supply does. This framework pairs with l0g's analysis of uranium deficits and with the oil market guide as a more conventional commodity comparison. --- Main sources: World Nuclear Association on uranium markets and the fuel cycle; World Nuclear News guide to the nuclear fuel cycle; U.S. Energy Information Administration on the nuclear fuel cycle; UxC fuel price indicators; TradeTech uranium price data; World Nuclear News on Cameco and Kazatomprom guidance; U.S. Nuclear Regulatory Commission and U.S. Department of Energy material on the Russian uranium import ban, enrichment and HALEU. ============================================================================ REFERENCE GUIDE: Reading the oil market: price, curve, supply and data URL: https://l0g.fr/en/guides/read-oil-market/ Canonical French source: https://l0g.fr/guides/lire-le-marche-petrolier/ Date: 2026-07-04 (reviewed 2026-07-04) ---------------------------------------------------------------------------- Oil has a price everyone quotes and almost nobody reads properly. A headline says “the barrel” is at a given level. Which barrel? Brent or WTI? Spot or a future contract? The useful information is often not the level, but the spread, the curve and the direction of inventories. This guide explains oil-market mechanics from price formation to the data calendar. The 2026 Hormuz shock is the case study. Two prices for one market There is no single oil price. There are benchmark prices tied to qualities of crude and delivery locations. Two dominate global discussion. Brent, from the North Sea, is a light sweet crude delivered by sea. It is the global maritime benchmark and directly or indirectly anchors a large share of internationally traded crude. WTI, West Texas Intermediate, is also light and sweet but delivered at Cushing, Oklahoma, an inland pipeline hub. It is the U.S. marker. This geography explains why the two prices differ. The Brent-WTI spread reflects transport costs, quality differences and regional imbalances. Since 2015, Brent has usually traded $2 to $8 above WTI. During the March 2026 Hormuz stress, the spread was around $4.70; by late June, after the risk premium faded, it had fallen below $3. A spread often says more than the outright price. Asia has another reference: Dubai/Oman, a heavier, sourer crude used for Gulf cargoes to the East. OPEC also publishes its own reference basket. Reading a quote means first asking which benchmark you are looking at. Paper price and physical barrels The price on screens is not the price of a barrel changing hands on a dock. It is formed first in futures markets: Brent on ICE in London, WTI on NYMEX/CME in New York. Paper volume vastly exceeds physical delivery, and that is where liquidity, hedging and speculative positioning concentrate. Physical cargoes then price off benchmarks through differentials: a specific crude trades at “Dated Brent minus $1.20” or at a premium. The differential captures quality, sulphur, density, location, shipping and refinery demand. This structure means price moves can reflect either physical scarcity or financial positioning. The positioning side is visible in the CFTC COT report, which breaks down futures positions by trader category. Contango and backwardation: reading the curve Each future maturity has its own price. Connecting them produces a futures curve. Its shape is often more informative than spot. In backwardation, spot is above future prices. The curve slopes downward. The market is paying for immediate barrels, which usually signals tightness. Backwardation discourages storage: holding oil to sell later earns less than selling now. It is typical of supply shocks or strong demand. In contango, spot is below future prices. The curve slopes upward. The market has immediate surplus and pays for storage. Deep contango can make it profitable to buy physical oil, store it and sell futures. The extreme case was April 20, 2020, when the front WTI contract settled around -$37 because holders could not take delivery at saturated Cushing storage. A negative oil price was not a market joke. It was a physical storage failure expressed through a futures contract. Supply: OPEC+, shale and spare capacity On the supply side, three forces matter. The first is OPEC+, the OPEC group plus non-OPEC producers led by Russia. The group manages a decisive share of global supply through quotas and voluntary cuts. In 2026 it began unwinding cuts, adding roughly 600,000 barrels per day between April and June and 188,000 more in July. The second is U.S. shale. Shale made the United States the world's largest producer and acts as a flexible supply response: wells start and decline quickly, so production reacts faster to price than conventional projects. The third is spare capacity, the volume that can be brought online within roughly thirty days. It is the global cushion. In mid-2026, OPEC+ spare capacity was around 5 million barrels per day, with roughly 3 million in Saudi Arabia alone. As the group reopens supply, that cushion shrinks. A market with little spare capacity is one crisis away from a price spike. Demand: where the barrel goes Global oil demand is around 103 to 104 million barrels per day. It is concentrated in transport, petrochemicals and industry, with growth shifting toward Asia. China, the world's largest crude importer, acts as a swing buyer: it can buy aggressively when oil is cheap and draw on stocks when oil is expensive. Real-time demand is difficult to measure. That is why the IEA, OPEC and the U.S. EIA publish monthly estimates that often disagree. The IEA tends to reflect consuming-country concerns; OPEC often takes a more optimistic view of demand; the EIA provides an independent U.S. statistical view. Reading oil means comparing all three rather than treating one report as truth. Inventories: cushion and signal Inventories absorb the gap between supply and demand. Commercial stocks are held by refiners and traders. Strategic petroleum reserves are held by governments. IEA members are required to hold the equivalent of 90 days of net imports. Strategic stocks are geopolitical. China's strategic reserve, estimated around 1.24 billion barrels in early 2026, may have become the world's largest. The U.S. Strategic Petroleum Reserve, around 409 million barrels in spring 2026 versus a peak of 727 million in 2009, has been heavily drawn down since 2022. Filling a reserve supports demand; drawing it down suppresses prices temporarily but reduces protection against the next shock. Weekly commercial inventories move prices. A surprise build is bearish; a surprise draw is bullish. Cushing matters especially for WTI: when Cushing fills, WTI weakens versus Brent. The data calendar Oil has a rich, mostly free data calendar. - API, Tuesday 16:30 ET. The American Petroleum Institute publishes a private estimate of U.S. inventories. Less authoritative, but watched as a preview. - EIA, Wednesday 10:30 ET. The Weekly Petroleum Status Report is the key short-term release: crude and product inventories, Cushing, refinery utilisation and U.S. production. - Baker Hughes, Friday 13:00 ET. The U.S. rig count is a leading indicator of shale activity. - Monthly reports. The IEA Oil Market Report, OPEC Monthly Oil Market Report and EIA Short-Term Energy Outlook each provide a different read on global supply-demand balance. Reading oil in practice Oil is not one number. The price level gives mood. Brent-WTI gives geography. The futures curve gives tightness or surplus. Spare capacity gives fragility. Inventories give near-term direction. Together, they say what the headline does not. The 2026 Hormuz shock shows the system. Brent moved toward $80, with stress scenarios above $100 if the Strait remained closed, while the curve shifted into backwardation. De-escalation and OPEC+ supply then pulled the barrel toward $71 by early July. The energy shock also fed directly into U.S. inflation, as explained in the CPI guide. Oil is one of macroeconomics' main transmission belts. --- Main sources: U.S. Energy Information Administration, Weekly Petroleum Status Report and Short-Term Energy Outlook; IEA, Oil Market Report · June 2026; RBN Energy on the Brent-WTI spread; Charles Schwab on WTI versus Brent; OilPrice on OPEC+ spare capacity; EIA on China strategic stocks and the U.S. Strategic Petroleum Reserve; CME Group on the API/EIA data calendar. ============================================================================ REFERENCE GUIDE: GENIUS Act: who enforces what in U.S. stablecoin regulation? URL: https://l0g.fr/en/guides/map-genius-act-stablecoin-regulators/ Canonical French source: https://l0g.fr/guides/qui-applique-le-genius-act/ Date: 2026-07-01 (reviewed 2026-07-01) ---------------------------------------------------------------------------- The GENIUS Act was signed on July 18, 2025. One year later, the question is no longer whether payment stablecoins will be regulated in the United States, but who regulates them and how. The law does not create a single authority. It divides supervision across several federal agencies and state regulators, depending on the issuer's legal form, and turns every permitted issuer into a financial institution subject to sanctions law. This guide is the operational companion to Stablecoins and the GENIUS Act. That guide explains how a stablecoin holds its peg and what a reserve must contain. This one maps the regulators, licensing routes and compliance obligations. One law, many regulators The natural assumption is that the GENIUS Act creates a single stablecoin regulator. It does not. Implementation is shared by the OCC, the Federal Reserve Board, the FDIC, the NCUA, the Treasury and state regulators. The key concept is the primary federal payment stablecoin regulator, assigned according to the issuer's legal structure. The OCC has the broadest perimeter. It supervises federally qualified non-bank issuers, national bank subsidiaries, uninsured national banks and trusts, federal branches of foreign banks, and registered foreign issuers. The other banking agencies generally supervise issuing subsidiaries of institutions they already regulate. The map of supervision | Issuer type | Main supervisor | |---|---| | Federally qualified non-bank payment stablecoin issuer | OCC | | National bank or uninsured national trust subsidiary | OCC | | State member bank subsidiary | Federal Reserve | | State non-member bank or savings association subsidiary | FDIC | | Credit union service organisation | NCUA | | State-qualified issuer at or below $10 billion | State regulator | | Foreign payment stablecoin issuer | OCC registration | On top of that prudential map, two Treasury arms apply across the whole sector: FinCEN for anti-money laundering and countering terrorist financing, and OFAC for sanctions. Three entry points to issue To issue a payment stablecoin legally in the United States, an entity must be a permitted payment stablecoin issuer. There are three main routes. The first is an insured depository institution subsidiary, supervised by the primary federal regulator of the parent institution. The second is a federally qualified non-bank issuer supervised by the OCC. The third is a state-qualified issuer supervised at state level, subject to federal constraints and size thresholds. Foreign issuers need registration to offer tokens into the U.S. market. The route matters. It decides who grants the licence, who examines the issuer, and which rulebook applies. Federal versus state supervision The state route is real but capped. A state-qualified issuer can remain under a state regime while its stablecoin issuance stays at or below $10 billion. Above that threshold, the issuer must move into federal supervision unless the OCC grants an exception. There is another condition: the state regime must be found “substantially similar” to the federal framework. A Stablecoin Certification Review Committee must determine unanimously that the state framework meets or exceeds federal standards. Without that certification, the state path becomes much narrower. This is designed to prevent state supervision from becoming a regulatory escape hatch. Deposit insurance does not pass through to the token holder A subtle but important FDIC point breaks a common intuition. Dollars deposited at a bank as stablecoin reserves are not automatically insured for the final token holder. Deposit insurance protects the issuer's bank deposit as a corporate account; it does not transform a stablecoin into an insured bank deposit for the end user. Holding a stablecoin backed by bank deposits is therefore not the same as holding an insured deposit. Stablecoin issuers become financial institutions This is the most consequential shift. The GENIUS Act requires permitted issuers to be treated as financial institutions under the Bank Secrecy Act. FinCEN and OFAC proposals translate that into two obligations. First, issuers need a bank-like AML/CFT programme. Second, they need an effective sanctions compliance programme. That is a major change for an instrument that often presented itself as technical infrastructure rather than regulated finance. Issuers must be able to freeze, seize or burn tokens under a lawful order. The law defines that order as a final action by a competent court or federal agency identifying the token or account with reasonable specificity. A stablecoin is therefore not simply code; it is a state-reachable payment instrument. That is exactly where this guide meets the OFAC and SDN List guide. Neither security nor commodity The law classifies a compliant payment stablecoin as neither a security nor a commodity. That moves authorisation away from the SEC and CFTC and into the banking-regulator architecture. The trade-off is clear. Issuers get legal clarity and access to a dedicated regime. In return, they cannot pay interest or yield to holders simply for holding the token, and they become subject to banking-style supervision, AML and sanctions. The implementation calendar Most rules must be issued within one year of enactment, meaning by July 18, 2026. The law becomes fully effective no later than January 18, 2027, or earlier, 120 days after final rules are published by the primary federal regulators. By mid-2026, the main agencies had published proposed rules and comment periods had largely closed. But timing remains uncertain. A later customer-identification proposal published on June 22, 2026 may not be finalised before 2027, and the statute does not contain an automatic fallback if an agency misses its rulemaking deadline. The calendar should therefore be read as a likely path, not an ironclad promise. How to read the architecture in practice Use five checks. First, identify the issuer's legal form. Second, map it to the right supervisor. Third, determine whether the relevant rule is proposed or final. Fourth, check when the effective date is triggered. Fifth, remember that AML and sanctions obligations apply to everyone, regardless of the prudential supervisor. Methodology This guide maps the regulatory architecture from the GENIUS Act and known federal implementation proposals as of July 1, 2026. It is not legal or investment advice. Proposed rules can change before finalisation, and implementation dates may shift. Thresholds and responsibilities are drawn from the statute and published Federal Register proposals. --- Main sources: GENIUS Act (S.1582, 119th Congress); Federal Register proposed rules from the OCC, FDIC, FinCEN and OFAC; U.S. Treasury releases; OCC issuer reporting bulletins; Chapman and Cutler GENIUS Act rulemaking tracker; Sullivan & Cromwell, Mayer Brown, Morgan Lewis and Troutman Pepper Locke analyses of implementation proposals. ============================================================================ REFERENCE GUIDE: How to Read TIC Data: who really finances U.S. debt URL: https://l0g.fr/en/guides/read-tic-data-us-debt/ Canonical French source: https://l0g.fr/guides/lire-les-donnees-tic/ Date: 2026-06-30 (reviewed 2026-06-30) ---------------------------------------------------------------------------- The question returns at every Treasury auction and every yield spike: if foreigners step back from U.S. debt, who buys, and at what yield? The reference source is the Treasury International Capital system. It tells how many Treasuries non-residents hold and where they are recorded. But it speaks in a language that must be learned: custody accounting, where Belgium can matter more than Saudi Arabia and a Cayman hedge fund can disappear into a domestic line. What TIC measures Treasury International Capital is the set of reports maintained by the U.S. Treasury, with the New York Fed as agent, tracking portfolio capital movements between U.S. residents and non-residents. Two pieces matter for U.S. debt: monthly net purchases and sales of long-term U.S. securities by foreigners, and holdings data, especially the Major Foreign Holders table for Treasuries. At the end of 2025 non-residents held roughly $9.27 trillion of Treasuries, with early-2026 totals above $9.4 trillion. That is huge, but it is around 30% of marketable debt, a share that has drifted down over time as domestic ownership has grown. Foreigners matter; they are not the whole marginal buyer. The top line is familiar: Japan around $1.19 trillion, the United Kingdom around $866 billion, China around $684 billion. But the rest of the ranking is a warning: Belgium, Luxembourg, the Cayman Islands, Ireland. These are custody and fund-domicile centers as much as final investors. Custody bias: the core problem The Treasury says it explicitly: the monthly table is collected on a custody basis and cannot assign ownership with perfect accuracy. A Treasury bought by an investor in one country but held through a custodian in another is attributed to the custodian's country. Belgium is large because Euroclear is there. Luxembourg and Ireland are large because investment funds are registered there. The UK is inflated by London's custody role. The Cayman Islands line captures hedge funds, including vehicles involved in the Treasury basis trade. That last point matters for stability. Fed research has estimated that Cayman hedge-fund Treasury positions were undercounted by about $1.4 trillion at the end of 2024, with some exposure classified as domestic. Read naïvely, TIC can turn a leveraged arbitrage into “American savings.” Holdings, flows and price effects Do not mix stock and flow. The holdings table gives a market-value stock at a date. Monthly flows show net purchases and sales. The two can diverge sharply because Treasury holdings are valued at market prices. If yields rise, bond prices fall and a country's reported holdings can decline even if it sold nothing. That is why headlines about a country “dumping Treasuries” often misread valuation effects. To judge appetite, compare the change in holdings with net monthly flows. In early 2026, monthly TIC inflows reached $184.5 billion in February and $150.7 billion in March, while October 2025 had shown a negative balance around -$37.3 billion. Flows, not just stocks, tell the demand story. China, the UK and the shift away from official buyers Over the long run the data tells two stories. The first is China's decline. China held more than $1.3 trillion of Treasuries in the mid-2010s and was below $700 billion at the end of 2025. Some Chinese holdings are booked elsewhere, but the direction is real. The second is the rise of custody centers and private buyers. The share of foreign holdings owned by official institutions, central banks and sovereign funds, has fallen from above 53% at the end of 2021 to roughly 42% at the end of 2025, even as total foreign holdings hit records. The marginal foreign holder is increasingly a private investor sensitive to yield, not a central bank mechanically accumulating reserves. That changes the nature of external financing. How to read TIC in practice The useful sequence is simple. Separate flows from holdings. Correct mentally for custody centers: Belgium, Luxembourg, Ireland, the Caymans and part of the UK are not straightforward national demand. Beware of valuation effects when yields move. Use the annual benchmark survey to refine true owner nationality, because monthly data is a fast but distorted map. TIC should be read with the domestic side of debt ownership: the Fed balance sheet, money-market funds, banks and now stablecoin issuers buying T-bills. TIC gives the foreign window; it is not the full financing map. The lesson is broader: transparency is not legibility. TIC gives precise numbers, but arranged according to the grammar of custody. Read well, it is one of the best windows on U.S. external financing. Read badly, it makes Belgium speak for Beijing and hides leverage inside a country label. --- Main sources: U.S. Treasury, Treasury International Capital system; Treasury Major Foreign Holders table; Treasury TIC monthly releases; Congressional Research Service, Foreign Holdings of Federal Debt; Wolf Street analysis of foreign Treasury holders and basis-trade undercounting. ============================================================================ REFERENCE GUIDE: Reading on-chain data: what the blockchain shows, and what it hides URL: https://l0g.fr/en/guides/read-on-chain-data/ Canonical French source: https://l0g.fr/guides/lire-la-donnee-on-chain/ Date: 2026-06-29 (reviewed 2026-06-29) ---------------------------------------------------------------------------- A listed company reveals positions through quarterly filings. A public blockchain records transactions continuously in a permanent ledger that anyone can inspect. It looks like the opposite of regulatory opacity: no delay, no filer, no gatekeeper. But opacity has not disappeared. It has changed shape. This guide explains how to read the ledger, which metrics carry signal, and why radical transparency is never immediate readability. The public permanent ledger A public blockchain is a replicated, timestamped database where validated transactions remain permanently recorded. Amounts, addresses, timestamps and balances are visible. Unlike a 13F filing or a Form 4, there is no quarterly delay. The reference source is the network itself; running a node gives access to the full history without an intermediary. Two accounting models matter. Bitcoin uses UTXOs, unspent transaction outputs: a balance is the sum of fragments received and not yet spent. Ethereum uses an account model, closer to a balance ledger and better suited to smart contracts. This difference shapes which metrics are native and which are approximate. The promise is powerful: an open, permissionless audit trail. The problem is interpretation. Complete data is not interpreted data. Reading activity: addresses, transactions, flows The most cited activity metric is active addresses: addresses that sent or received a transaction over a period. It approximates network use, but it is not a user count. One person can control hundreds of addresses, while an exchange can aggregate millions of users behind a cluster of addresses. Transaction count and volume have the same problem. A Bitcoin transaction can return change to the sender. Exchange internal transfers can inflate activity without representing economic usage. That is why data providers publish adjusted versions, removing transfers between addresses controlled by the same entity or very short-lived outputs. Without adjustment, you are measuring noise. Exchange flows are watched closely by institutional crypto desks. Coins moving onto exchanges may be available for sale; coins leaving exchanges for self-custody may signal longer-term holding. But even that signal has become more ambiguous. Bitcoin exchange reserves fell from roughly 3.2 million BTC in February 2020 to about 2.40 million BTC, roughly 12% of supply, by late April 2026. Read naively, that looks like accumulation. Read carefully, it is blurred by spot ETFs: since 2024, more than 1.45 million BTC have moved into institutional custody, and some providers classify those addresses alongside exchange infrastructure. A withdrawal is no longer automatically a retail cold-storage signal. Valuation: realised cap, MVRV, SOPR, NVT On-chain data makes possible a kind of valuation that does not exist in traditional markets: an estimate of the market's cost basis. Realised capitalisation values each coin at the price when it last moved on-chain, not at today's price. Dividing market capitalisation by realised capitalisation gives MVRV, a proxy for the market's average unrealised profit or loss. Introduced by Murad Mahmudov and David Puell in 2018, it is a regime indicator. Above roughly 3.5, Bitcoin has historically been in euphoric territory; below 1, the average holder is underwater, a zone often associated with capitulation and accumulation. SOPR, the spent output profit ratio, asks whether coins moved on a given day were spent in profit or loss. A value above 1 means coins moved at a profit; below 1, at a loss. NVT compares market capitalisation to on-chain transaction volume, roughly like a price-to-activity multiple. None of these ratios should be read alone or across assets without context. Bitcoin and Ethereum do not play the same role, and the same level can mean different things on different networks. DeFi: TVL and double counting In decentralised finance, the flagship metric is total value locked, or TVL: the dollar value of assets deposited in protocols. The open reference is DefiLlama, whose protocol adapters are public and auditable. By mid-June 2026, total DeFi TVL was around $71.8 billion across more than 450 chains, down about 37% year-to-date and almost 60% below the November 2021 peak of roughly $177 billion. Ethereum concentrated about 53% of TVL, showing reconcentration rather than unstoppable multichain diffusion. TVL has three traps. First, it rises when token prices rise, even without new capital inflow. Second, double counting is real: an asset can be deposited, tokenised and redeposited across protocols. Third, TVL is not the size of the crypto market. Stablecoin market capitalisation, around $314 billion in mid-2026, was more than four times DeFi TVL. What the chain hides This is the core issue. A blockchain shows all transactions, but it does not show identity. An address is not a person. Address clustering depends on probabilistic heuristics. Exchange clusters can contain tens of millions of addresses and grow constantly. Entity labels are not facts; they are provider inferences. Most trading also remains off-chain. A spot trade inside a centralised exchange updates an internal ledger, not the public blockchain. The chain sees deposits and withdrawals, not the internal order book. Reading only on-chain data is like watching a building's entrances and exits without seeing what happens inside. Labels are also revised. Glassnode warns that exchange-balance series can change as labels improve. Two providers can publish different exchange reserve numbers for the same day. Wrapped assets, bridges, layer-2s and lost coins add more distortion. Stablecoin supply issued by an issuer is not always active circulating supply. The data is exhaustive, but interpretation never is. Reading on-chain data in practice Use block explorers for raw truth. Use Glassnode, CryptoQuant, DefiLlama, Dune, Nansen or Arkham for interpreted views, knowing each provider has assumptions. For continuous monitoring, l0g aggregates signals in its dashboards. The method is simple: seek convergence, not one metric. Falling exchange reserves, low MVRV and SOPR below 1 together tell a stronger story than any single line. Always compare a metric to the same asset's own history. And cross on-chain data with macro liquidity, because crypto remains highly sensitive to global liquidity conditions. The core lesson The blockchain made public what traditional finance often hides: balances, transactions and flows. But it did not abolish opacity. It moved opacity into labelling heuristics, off-chain settlement and calculation conventions. Reading on-chain data means knowing exactly where the data stops and interpretation begins. --- Main sources: Glassnode research on exchange metrics and address clustering; Glassnode documentation for active addresses, MVRV, SOPR, NVT and realised price; Newhedge on the MVRV ratio and its origin; TRdesk on Bitcoin exchange reserves and ETF custody; DefiLlama methodology and TVL/stablecoin data; CoinLaw DeFi market statistics based on DefiLlama snapshots; Bitget Academy on exchange reserves and interpretation limits. ============================================================================ REFERENCE GUIDE: How to Read M2 Money Supply: liquidity signal, bad readings and risk-on narratives URL: https://l0g.fr/en/guides/read-m2-money-supply-risk-on/ Canonical French source: https://l0g.fr/guides/m2-masse-monetaire-risk-on/ Date: 2026-06-29 (reviewed 2026-06-29) ---------------------------------------------------------------------------- M2 money supply is one of the most quoted numbers in risk-on investing, and one of the most abused. It has been turned into a “global liquidity clock” that supposedly predicts Bitcoin with a neat lag. Reality is more careful. M2 carries a signal, but only when treated as a stock of money, not as a hose pointed at markets. The exact definition of M2 M2 is a monetary aggregate published monthly by the Federal Reserve in the H.6 release. It nests inside M1. M1 includes the most liquid money: currency in circulation, demand deposits and, since May 2020, other liquid deposits such as savings accounts and money market deposit accounts. M2 adds less-liquid components: small time deposits below $100,000 and retail money market fund shares. Three details matter. First, M2 counts money held by the U.S. nonbank public: federal government deposits and interbank deposits are excluded. Second, a May 2020 regulatory change moved savings deposits into M1, complicating composition comparisons across that date. Third, from the July 28, 2026 H.6 release, the Fed changes how IRA and Keogh balances are treated in the seasonally adjusted series. The crucial point: M2 is a stock, not a flow of money into equities or crypto. A rising M2 means liquid money balances are rising. It does not tell what holders will do with those balances. 2020-2023: explosion, then contraction The 2020-2023 episode is the cleanest case study. Fed asset purchases and fiscal transfers pushed M2 growth to nearly 27% year over year in early 2021, a record in the series. Monetarists saw an inflation warning when the central bank still played down the risk. Inflation followed. Then tightening and balance-sheet reduction pulled M2 into a rare contraction, down 4.6% year over year in April 2023, the deepest decline in the modern data. That episode supports both sides of the debate. M2 did carry a powerful signal when its move was extreme. But it was also an exceptional fiscal-monetary shock. In normal regimes, the signal is weaker. By mid-2026, M2 was above $22.8 trillion and growing roughly 5.6% year over year, close to its long-run rhythm. The link with risk-on The usual shortcut is that more money pushes risk assets higher. The mechanism is not absurd: abundant liquidity can seek return in equities, credit, commodities and scarce digital assets. But the transmission depends on velocity, the speed at which money turns over. The quantity theory identity links money times velocity to nominal income. If velocity falls, higher M2 may feed neither inflation nor markets. Velocity has trended down since the 1990s and remains unstable. That is why mechanical M2 readings fail. The composition also matters: rising retail money market balances can reflect risk-off behavior, as households move from low-yielding deposits into safer cash-like assets, not a rush into speculation. Bitcoin as liquidity barometer Crypto is where the M2 narrative became most popular. The serious version comes from Lyn Alden and Sam Callahan: from May 2013 to July 2024, Bitcoin moved in the same direction as global liquidity 83% of the time over twelve-month windows and 74% over six-month windows. That is a meaningful directional relationship. The weak version is the social-media chart claiming that global M2 leads Bitcoin by a fixed lag. The lag changes by author and by cycle: ten weeks, twelve weeks, fifteen weeks. When the “best” lag changes after the fact, it is overfitting, not law. Global M2 itself is a fragile construction: it aggregates many central-bank money supplies into dollars, so FX moves can inflate or depress the series without any real liquidity impulse. Recently the relationship broke. Over the twelve months into early 2026, global M2 rose about 12% while Bitcoin fell by a similar amount, according to CF Benchmarks. Liquidity dominated in 2020 and 2022; other factors dominated in 2025-2026: ETF flows, leverage, regulation and geopolitics. Common traps The first trap is confusing stock and flow. A high M2 level is not a reservoir waiting to flood markets. The second is confusing level and change: M2 and asset prices both trend upward over long periods, creating spurious correlations between non-stationary series. The useful information is in changes, not in two lines rising on the same chart. The third trap is ignoring the dollar when discussing global M2. A weaker dollar mechanically raises non-U.S. money aggregates when translated into dollars. The fourth is treating a historical lag as a future clock. How to use it without fooling yourself M2 remains useful if read as a conditional monetary climate gauge. Look at year-over-year growth rather than the raw level. Cross-check velocity and composition to see whether the move reflects risk appetite or a move into cash-like safety. Then compare it with other liquidity measures: reserves, the Treasury General Account, reverse repo, and money-market funding stress. M2 is one piece of the broader plumbing described in the guides on the Fed balance sheet and net liquidity. It is a good weather indicator, not a crystal ball. --- Main sources: Federal Reserve, H.6 Money Stock Measures definitions and current release; FRED M2SL, M2V and M2REAL series; Lyn Alden and Sam Callahan, Bitcoin: A Global Liquidity Barometer; CF Benchmarks, The M2-Bitcoin Relationship: What the Data Actually Shows; Kokabian working paper on global liquidity and Bitcoin. ============================================================================ REFERENCE GUIDE: MiCA, acronym by acronym: decoding Europe's crypto rulebook URL: https://l0g.fr/en/guides/decode-mica-crypto-regulation/ Canonical French source: https://l0g.fr/guides/mica-sigle-par-sigle/ Date: 2026-06-21 (reviewed 2026-06-21) ---------------------------------------------------------------------------- MiCA is the world's first full-scale crypto rulebook, and it is also an alphabet soup: ART, EMT, CASP, NCA, EBA, ESMA. This guide decodes the acronyms, explains who supervises what, and highlights the provision that is less about finance than geopolitics: caps on dollar stablecoins designed to protect the euro's monetary sovereignty. The key deadline is the end of the transition period on July 1, 2026. MiCA, the Markets in Crypto-Assets Regulation, entered into force in 2023 and has been fully applicable since December 30, 2024. It classifies crypto-assets, strictly regulates stablecoins, licenses service providers, and splits supervision between national authorities and European agencies. Because of its scale, it has become the benchmark against which other crypto regimes are compared. The three asset categories Everything begins with classification. MiCA divides crypto-assets into three groups. The first group is e-money tokens, or EMTs: tokens that reference one official currency. The second is asset-referenced tokens, or ARTs: tokens that reference a basket of currencies, commodities such as gold, other crypto-assets, or a combination of assets. The third is the residual category of “other crypto-assets,” including many utility tokens. One detail matters: the word “stablecoin” is not the legal category. In MiCA, stablecoins are mostly captured through EMTs and ARTs. Financial instruments already covered by MiFID are outside the regime, and non-fungible tokens are in principle excluded unless they are issued in a way that makes them functionally interchangeable. EMTs and ARTs: the two MiCA stablecoins The EMT/ART distinction is legal and practical. An EMT references a single currency and can only be issued by a credit institution or an e-money institution. Holders must be able to redeem it at par at any time. An ART references a basket or asset and can be issued by a wider range of authorised entities. Redemption depends on the market value of the reserve assets. Both regimes impose strict requirements: full reserves, segregation, custody with independent depositaries, own funds, an approved white paper, and no interest paid to holders. Algorithmic stablecoins are not explicitly banned by name, but the full-reserve logic makes a purely algorithmic payment stablecoin almost impossible to fit into MiCA. CASP: the licence to operate A company providing crypto-asset services in the Union needs a CASP licence: exchange, custody, trading, advice, transfers and related services. The authorisation is granted by a national competent authority and then passported across the Union. That passport is the core of the regime. MiCA replaces the previous mosaic of national crypto licences with a single European framework. A service provider with more than 15 million active users in the Union becomes a significant CASP and is subject to reinforced supervision. Who supervises what: NCA, EBA, ESMA Supervision is divided, which is a common source of confusion. The national competent authority, or NCA, authorises CASPs and ART issuers in its jurisdiction and grants the European passport. The European Banking Authority, or EBA, directly supervises issuers of significant stablecoins and sets prudential standards for reserves and capital. The European Securities and Markets Authority, or ESMA, covers market conduct, non-stable tokens, CASPs and market abuse. ESMA also maintains the central registers of white papers, authorised service providers and non-compliant entities. In practice, ESMA's register is the starting point for verification. The political lever: monetary sovereignty MiCA's most political provision is also its most under-discussed. To protect the euro's monetary sovereignty from dollar stablecoins, the regulation caps the use of EMTs and ARTs denominated in a non-European currency when they are used as a means of exchange. Above one million transactions or €200 million per day, the issuer must stop issuing the token concerned. Given that almost 99% of global stablecoin supply is dollar-denominated, the point is obvious: MiCA is also a monetary defence instrument. The thresholds for a “significant” token follow the same size logic: around ten million holders, €5 billion in outstanding value, or large daily volumes can bring the issuer into direct EBA supervision. The calendar and the 2026 cliff MiCA was phased in by risk. Rules on stablecoins, EMTs and ARTs have applied since June 30, 2024. The rest of the regime, including CASP licensing and the Transfer of Funds Regulation travel rule, has applied since December 30, 2024. A transition clause allows firms already operating under national regimes to continue temporarily, but that transition expires no later than July 1, 2026, with national variations before that date. After the cliff, serving European crypto clients without MiCA authorisation is no longer a grey zone but a breach of EU law. More than seventy CASPs were authorised by the end of 2025. Mid-2026 is therefore the moment that separates compliant actors from the rest. MiCA versus GENIUS: two philosophies Europe and the United States now have two very different stablecoin regimes. Both forbid issuers from paying interest to stablecoin holders. But the philosophy differs. The U.S. GENIUS Act builds a banking-style framework that supports the expansion of the digital dollar. MiCA inherits the logic of e-money, gives the EBA and ESMA major supervisory roles, and caps non-European stablecoins. One regime tries to export its currency; the other tries to protect itself from that currency. That asymmetry explains why Circle pursued MiCA authorisation while Tether did not, leading to restrictions on USDT in parts of Europe. Blind spots MiCA does not solve everything. Truly decentralised finance with no identifiable intermediary is mostly outside the framework for now, and the Commission still needs to design a dedicated regime. Calling a token an NFT does not automatically remove it from MiCA if it is issued in a large interchangeable series. Transition periods differ by member state, creating uneven protection and potential regulatory arbitrage. The overlap with payment services law also complicates EMT issuance. Classification is not credit analysis. MiCA tells you where an asset sits in the regulatory grid; it does not tell you whether the issuer is sound. How to read it in practice Start with the asset category: EMT, ART or other crypto-asset. Then check the ESMA register to verify whether the issuer or service provider is authorised and by which national authority. For a stablecoin, check whether EBA significance thresholds apply and whether non-euro caps matter. Finally, separate regulatory status from issuer risk: the quality of the reserve still needs its own analysis. The European passport improves access. It does not replace due diligence. Methodology This guide is based on primary sources: Regulation (EU) 2023/1114 on markets in crypto-assets, ESMA registers and implementation statements, EBA technical standards for ART and EMT issuers, and official application and transition calendars. Thresholds, dates and categories were checked against those sources. Practical MiCA analysis on l0g starts with the ESMA register, identifies the asset class, checks authorisation and competent authority, and separates regulatory status from issuer quality. --- Main sources: Regulation (EU) 2023/1114 on markets in crypto-assets (MiCA), Titles II to VII; ESMA MiCA pages and registers; EBA technical standards for significant ART and EMT issuers; official implementation calendar, including stablecoin rules from June 30, 2024, full regime from December 30, 2024, and Article 143 transition cliff no later than July 1, 2026. ============================================================================ REFERENCE GUIDE: OFAC and the SDN List: can sanctions target code? URL: https://l0g.fr/en/guides/read-ofac-sdn-list/ Canonical French source: https://l0g.fr/guides/ofac-sdn-list/ Date: 2026-06-21 (reviewed 2026-06-21) ---------------------------------------------------------------------------- A name on the SDN List, and dollar access shuts down. It is the most powerful financial weapon in the world, administered by a U.S. Treasury office that few outside compliance teams know well. Crypto has tested its limits: in 2025, U.S. courts forced OFAC to remove Tornado Cash from the list, holding that immutable code cannot be sanctioned as property. This guide explains how the list works, how it reaches crypto, and where it stops. OFAC, the Office of Foreign Assets Control, is the U.S. Treasury office that administers and enforces economic sanctions in support of foreign-policy and national-security objectives. Its flagship tool is the Specially Designated Nationals and Blocked Persons List, or SDN List. U.S. persons are generally forbidden to deal with listed persons and entities, and their property under U.S. jurisdiction is blocked. How the SDN List works The mechanism is deliberately harsh. When a person or entity is added to the list, two things happen. First, all property and interests in property under U.S. jurisdiction are blocked. Second, U.S. persons are prohibited from transacting with them. Liability is strict. A violation can matter even without intent or knowledge, which is why the global financial system tends to over-comply. The reach is far beyond U.S. borders. Because the dollar runs through global trade and international banks need access to the U.S. system, an SDN designation can effectively cut a target off from global finance. The legal basis is often the International Emergency Economic Powers Act of 1977, activated through executive orders targeting countries, groups or activities. The trap: the 50 percent rule A compliance check cannot stop at the exact name on the list. Under OFAC's 50 percent rule, an entity owned 50 percent or more, directly or indirectly, by one or more blocked persons is itself treated as blocked, even if its own name is not on the list. Ownership by several blocked persons is aggregated. The absence of a name from the SDN List is therefore not enough. You must trace beneficial ownership. This is one of the most common compliance failures. Beyond the SDN List The SDN List is not OFAC's only instrument. Sectoral sanctions restrict certain activities without fully blocking an entity. Other non-SDN lists impose narrower restrictions. Secondary sanctions can target non-U.S. actors that deal with certain sanctioned parties by threatening their own access to the U.S. system. The SDN List remains the sharpest tool: full blocking. The SDN List in the crypto era Since 2018, OFAC has also listed cryptocurrency wallet addresses. The logic is the same: transacting with a listed address is prohibited. North Korean laundering networks, including Lazarus-linked activity, have been a major driver of these designations. But crypto enforcement often moves through centralised chokepoints. A centralised stablecoin issuer can freeze tokens held at sanctioned addresses. Tether, for example, has frozen USDT linked to addresses designated by OFAC, including in cases connected to Russian exchange activity. This is the paradox of centralised digital money: it can be moved like crypto but frozen like regulated finance. Tornado Cash: can OFAC sanction code? Tornado Cash is the landmark case. In August 2022, OFAC listed Tornado Cash, an Ethereum mixing protocol accused of laundering billions of dollars, including funds linked to North Korea. For the first time, OFAC targeted not just persons or companies but a set of autonomous smart contracts. The challenge came quickly. In Van Loon v. Department of the Treasury, users argued that immutable smart contracts were not “property” under IEEPA because no one could own, control or modify them, not even the developers. On November 26, 2024, the Fifth Circuit agreed. OFAC had exceeded its authority. The result was confirmed in 2025. On March 21, 2025, OFAC removed Tornado Cash from the SDN List. In April 2025, a federal district court issued a permanent injunction preventing OFAC from reimposing sanctions on the immutable contracts. OFAC maintained sanctions on one developer under the North Korea programme, and criminal proceedings against founders continue separately. The lesson is precise: OFAC's authority over truly immutable decentralised code is legally bounded. Enforcement must focus on persons and entities, not code itself. What this changes for compliance risk Tornado Cash's delisting is not a green light. Mixers remain high-risk money-laundering infrastructure. Strict liability remains. The risk profile of a transaction does not disappear because one sanction designation is lifted. The boundary drawn by courts separates code from persons. A developer, operator or entity can still be targeted. Immutable software is harder to treat as sanctionable property. Courts also suggested that Congress could update IEEPA to address these technologies directly. Until then, sanctions reach into DeFi remains uncertain. How to verify, step by step Start with the official OFAC sanctions search, not a third-party aggregator. Do not stop at a name match: trace ownership for the 50 percent rule. For crypto, check listed addresses but separate sanctions status from broader laundering risk. Date the check, because listings and delistings change constantly. A sanctions check reduces legal risk. It does not replace real risk analysis. Methodology This guide is based on primary sources: OFAC resources on the SDN List and the 50 percent rule, the International Emergency Economic Powers Act, the Treasury's Tornado Cash delisting release of March 21, 2025, the Fifth Circuit decision in Van Loon v. Department of the Treasury, and the April 2025 district-court injunction. Practical sanctions analysis on l0g starts from the official OFAC list, applies the 50 percent rule and separates legal sanctions status from money-laundering risk. --- Main sources: U.S. Department of the Treasury, OFAC SDN List resources and 50 percent rule guidance; International Emergency Economic Powers Act; Treasury/OFAC Tornado Cash delisting release of March 21, 2025; U.S. Court of Appeals for the Fifth Circuit, Van Loon v. Department of the Treasury, 122 F.4th 549; U.S. District Court for the Western District of Texas, April 2025 injunction; CoinDesk and specialist legal coverage for timeline and stablecoin freezes. ============================================================================ REFERENCE GUIDE: Schedule 13D vs 13G: activism or passive ownership URL: https://l0g.fr/en/guides/read-schedule-13d-13g-sec/ Canonical French source: https://l0g.fr/guides/13d-vs-13g-sec/ Date: 2026-06-21 (reviewed 2026-06-21) ---------------------------------------------------------------------------- The same crossing of the same threshold, 5% of a company's equity, can trigger two filings with very different meanings. Schedule 13D is the long form for an investor who wants to influence the company. Schedule 13G is the short form for a holder who says it is passive. The value is in that difference of intent, and in one specific item of the 13D where an activist explains what it plans to do. Since 2024 these filings arrive faster, which has made them more useful again. What 13D and 13G are Schedules 13D and 13G are disclosures required by the SEC when a person or group crosses 5% of a class of equity securities of a U.S.-listed company. A 13D is for holders with an intent to influence or control. A 13G is the lighter form for holders that qualify as passive, exempt or qualified institutional investors. The same economic stake can therefore mean two very different things. A 5.2% stake filed on 13G says, in effect, “we own this, but we are not trying to change the company.” A 5.2% stake filed on 13D says, “we may act.” That distinction is the whole point. Where the forms come from The two schedules come from Sections 13(d) and 13(g) of the Securities Exchange Act of 1934, as amended by the Williams Act of 1968. The logic is control transparency: once a holder crosses 5%, the market and other shareholders should know who is there, how much they own, and what they intend to do with that ownership. This is why these filings matter in takeover fights, activist campaigns and slow accumulations. They are not just position reports; they are early warnings about possible governance pressure. The dividing line: intent The dividing line is not the size of the stake, but the project behind it. A Schedule 13D is the form used by an investor that may influence management, seek board seats, push a sale, press for capital returns or otherwise affect control. A Schedule 13G is reserved for holders that meet one of the passive or institutional categories and do not seek control. The important signal is the switch. A passive holder that develops an intent to influence must move from 13G to 13D within five business days. That migration is often more informative than the initial stake itself: it says that a formerly quiet shareholder has decided to act. Deadlines changed in 2024 A lot of old pages are now stale. The SEC accelerated beneficial-ownership reporting. Since February 5, 2024, an initial 13D must be filed within five business days after crossing 5%, down from ten calendar days. Material 13D amendments are due within two business days, replacing the older, vaguer “promptly” standard. Since September 30, 2024, 13G deadlines have also tightened: five business days for passive investors, forty-five days after quarter-end for qualified institutional investors, and quarterly rather than annual amendments when material changes occur. Since December 18, 2024, these reports use structured data. The filing cut-off has also been extended to 10 p.m. Eastern time. Freshness changed the analytical value of the documents. Item 4: the activist signal In a 13D, one section carries most of the signal: Item 4, Purpose of Transaction. This is where the investor describes its plans: board representation, strategic alternatives, asset sales, capital allocation, buybacks, governance demands, or even control intentions. A 13D with a vague Item 4 is not the same as one that states the investor has contacted management, requested board seats or prepared proposals. Reading a 13D means reading Item 4 first. The group problem The subtle trap is the idea of a “group.” When several investors act together to acquire, hold or vote securities, their stakes are aggregated for the 5% threshold. The SEC has clarified that a group can arise when a future filer intentionally communicates its plan to file a 13D in order to induce another investor to buy, and that other investor buys. This matters because modern activism is often not a single investor moving alone. Informal “wolf packs” can form around a campaign. Reading one 13D can therefore understate the real coordinated pressure around a stock. How to read the filings on EDGAR Everything is public on EDGAR. Search either the company or the reporting person, then filter for SC 13D, SC 13G and amended versions marked /A. The reading order is simple. First identify the filing type: a 13D is more aggressive than a 13G. Then read Item 4 of the 13D to understand intent. Watch for conversions from 13G to 13D. Finally, reconstruct possible groups by looking at filings on the same issuer over a short window. Limits These disclosures have blind spots. Positions below 5% stay invisible. Synthetic exposure through derivatives has long complicated the regime and is still not perfectly captured. A holder can remain passive in form while approaching a level that makes it influential in practice. The shortened deadlines still leave an accumulation window. Group doctrine remains fact-intensive. A filing is not an investment thesis; it is the start of the investigation. Confluence with 13F and Form 4 The value rises when 13D/13G data is crossed with other signals. Form 13F shows the long equity book of institutional managers, quarterly and with delay. Form 4 shows insider transactions within two business days. A 13D adds the missing layer: intent. An activist 13D, insiders buying, and respected managers appearing in 13F filings create a much stronger convergence than any one signal alone. This is the logic behind 13FLOW . Methodology This guide is based on primary sources: Sections 13(d) and 13(g) of the Securities Exchange Act, SEC final rules modernizing beneficial-ownership reporting, and the instructions to Schedules 13D and 13G. Dates, thresholds and compliance deadlines have been checked against SEC materials. A practical l0g reading starts from the raw EDGAR filing, distinguishes 13D from 13G, reads Item 4, watches for switches and reconstructs possible groups. --- Main sources: SEC, Sections 13(d) and 13(g) of the Securities Exchange Act of 1934 and Schedules 13D and 13G; SEC, SEC Adopts Amendments to Rules Governing Beneficial Ownership Reporting (October 10, 2023, Release 2023-219) and related final rules; SEC compliance dates for 13D, 13G and structured data. ============================================================================ REFERENCE GUIDE: Stablecoins and the GENIUS Act: reading the promise of the digital dollar URL: https://l0g.fr/en/guides/read-stablecoins-genius-act/ Canonical French source: https://l0g.fr/guides/stablecoins-genius-act/ Date: 2026-06-21 (reviewed 2026-06-21) ---------------------------------------------------------------------------- A stablecoin promises one dollar, always, on demand. That promise rests on nothing except the reserve behind it. Most of the time, the reserve is visible through a monthly report produced by the issuer. The GENIUS Act, signed into law in July 2025, creates the U.S. federal framework for payment stablecoins. But an attestation is not an audit, a snapshot is not continuous proof, and this market of roughly $320 billion has quietly become a major buyer of U.S. Treasury bills. This guide explains how to read the promise. At its simplest, a payment stablecoin is a digital token backed one-for-one by a currency, in practice the dollar in nearly 99% of cases according to European Central Bank data. The issuer takes in a dollar, issues a token, and holds the dollar or a cash-like asset in reserve. The token's stability depends on whether that reserve exists, is liquid, and can be redeemed at par. How a stablecoin holds its peg The peg is enforced by redemption arbitrage. If a token trades at $0.99 but can be redeemed for $1, an arbitrageur can buy it and redeem it, pushing the price back toward par. The mechanism works only if redemption is credible. That is why reserve-backed stablecoins such as Tether's USDT or Circle's USDC differ radically from purely algorithmic models. Reserve-backed tokens depend on collateral. Algorithmic stablecoins tried to hold the peg through incentives and auxiliary tokens rather than equivalent reserves. TerraUSD's collapse in May 2022 showed what happens when that loop breaks. The GENIUS Act draws the lesson: payment stablecoins require full reserves, effectively closing the door to a purely algorithmic payment-stablecoin model. What the GENIUS Act actually requires Signed on July 18, 2025, after Senate passage on June 17 and House passage on July 17, the Guiding and Establishing National Innovation for U.S. Stablecoins Act is the first U.S. federal payment stablecoin framework. Its first move is legal classification: a compliant payment stablecoin is neither a security nor a commodity. Authorisation therefore moves away from the SEC/CFTC lane and into a banking-supervision architecture. The core requirements are straightforward: - only permitted issuers may issue payment stablecoins; - issuers fall into three broad classes: insured depository institution subsidiaries, federally supervised non-bank issuers under the OCC, and state-qualified issuers; - state-qualified issuers are capped at $10 billion in outstanding stablecoins before a federal regime becomes mandatory; - reserves must cover at least 100% of tokens outstanding; - eligible reserve assets are limited to high-quality liquid assets: dollars, bank deposits, Treasury bills with maturities of 93 days or less, repos backed by those bills, and government money market funds; - rehypothecation of reserves is banned; - reserve composition must be disclosed monthly; - management must certify the accounts; - reserves must be examined by a registered accounting firm; - issuers above $50 billion in outstanding value must produce annual audited financial statements; - issuers may not pay interest to holders. Two protections complete the structure. In an issuer bankruptcy, reserves are separated for the benefit of holders, with a super-priority claim if there is a shortfall. And issuers must be able to freeze, seize or burn tokens under a lawful order. Full effectiveness comes no later than January 18, 2027, or earlier if regulators finalise rules before then. Many implementing rules are due around July 18, 2026. An attestation is not an audit This is the point most commentary skips. A monthly reserve report is usually an attestation, not a full audit. An attestation checks a specific fact at a specific date. It can confirm that declared reserves existed at that snapshot. It does not necessarily express an opinion on internal controls, continuous solvency, counterparty quality, or what happens between two reporting dates. Tether illustrates the gap. After a New York Attorney General investigation found misleading claims about backing, Tether was required to publish periodic reserve reports. It has since published attestations, long before producing a full independent audit. Circle has used transparency and regulated status as a commercial advantage, with USDC reserve reports and a more explicitly regulated posture. The GENIUS Act raises the floor with monthly attestations and annual audits for the largest issuers. But a monthly snapshot is still a snapshot. How to read a reserve A reserve report must be read line by line. Not all “dollar” reserves are equal. Short Treasury bills are liquid and close to risk-free. Bank deposits carry counterparty risk. Repo exposure depends on collateral, tenor and counterparty. Money market funds add another layer of structure. Maturity matters because a reserve can be solvent on paper and still be hard to liquidate instantly. The Silicon Valley Bank episode made the point. In March 2023, Circle held roughly $3.3 billion of USDC reserves at SVB. When the bank failed, USDC briefly traded around $0.87 before the peg recovered after deposits were protected. A fully reserved stablecoin can still depeg if part of the reserve is stuck at a failed counterparty. The mechanics of depeg risk A depeg happens when the market doubts redemption at par. There are three main triggers. The first is insufficient or illiquid reserves. The second is counterparty failure. The third is an algorithmic spiral, where the mechanism meant to support the token collapses as confidence disappears. In every case, the immediate mechanism is a run: too many holders try to exit at once and the reserve cannot meet the demand cleanly. The analogy with banking is obvious, except that stablecoin holders do not automatically receive deposit insurance or central-bank lender-of-last-resort protection. The macro angle: stablecoins as Treasury buyers Stablecoins are no longer just a crypto-market instrument. To back their tokens, issuers hold large amounts of short-term Treasury bills. Collectively, they held around $155 billion of Treasury bills by late 2025; Tether alone claimed more than $100 billion. This makes stablecoin issuers an increasingly important marginal buyer of U.S. short-term debt. A 2026 BIS working paper showed that stablecoin flows now affect safe-asset prices, while U.S. officials have explicitly described stablecoin demand as a new force in the Treasury bill market. The same logic also creates a risk. If stablecoins become a buyer of Treasury bills when they grow, a large redemption wave can turn them into a forced seller of Treasury bills at the worst possible moment. The stability of payment tokens and the liquidity of the Treasury bill market are now connected. GENIUS versus MiCA The European Union's MiCA framework also imposes reserve and redemption requirements for e-money tokens and asset-referenced tokens. Both regimes ban interest paid directly to stablecoin holders. But the models differ. GENIUS builds a three-class banking-style regime designed around the dollar. MiCA caps the use of major non-euro stablecoins and requires European authorisation. The United States is building an export rail for the digital dollar; Europe is trying to limit its monetary spillover. How to read the risk in practice Start with the monthly reserve report: composition, Treasury bill share, deposit share, repo share, maturity and counterparties. Then check whether the report is an attestation or an audit, and which firm performed it. Verify the issuer's regulatory status under the GENIUS Act and MiCA. Check market history: capitalisation, peg deviations, redemption depth and chain concentration. Finally, be suspicious of yield promises: the GENIUS Act bans issuer interest to holders, so any yield offered by an intermediary means the risk has moved, not disappeared. Methodology This guide is based on primary sources: the GENIUS Act, Congressional Research Service analysis, Federal Reserve Bank of Richmond explanations, BIS work on stablecoins and safe-asset prices, European Central Bank data on dollarisation of stablecoin supply, public reserve reports from Tether and Circle, and MiCA for the European comparison. Market capitalisation, reserve and share figures are dated to early 2026 and should be treated as moving data. --- Main sources: GENIUS Act (S.1582, Public Law 119-27, signed July 18, 2025); Congressional Research Service analysis; Federal Reserve Bank of Richmond, Stablecoins and the GENIUS Act: An Overview; BIS Working Paper No. 1270, Stablecoins and safe asset prices; ECB material on the dollar share of stablecoin supply; public reserve reports from Tether and Circle; Regulation (EU) 2023/1114 (MiCA). ============================================================================ REFERENCE GUIDE: How to Read Private Credit: the risk you cannot see URL: https://l0g.fr/en/guides/read-private-credit-risk/ Canonical French source: https://l0g.fr/guides/analyser-credit-prive/ Date: 2026-06-21 (reviewed 2026-06-21) ---------------------------------------------------------------------------- Private credit has become one of the largest blind spots in the financial system: roughly $1.3 trillion in the United States, more globally, lent outside public markets, without daily prices and with limited mandatory transparency. This guide does not predict a crash. It teaches how to read what the usual market data cannot show. What private credit is Private credit is lending made directly by nonbank asset managers to companies, usually without a public quote or active secondary market. The manager raises capital from investors, lends to companies, often middle-market borrowers, and holds the risk outside bank balance sheets. The structure explains both the appeal and the danger: an illiquid, less regulated asset class that exchanges transparency and liquidity for yield. Why it exploded The boom began after 2008. Bank regulation and capital requirements pushed banks away from some corporate lending. Nonbank financial institutions filled the gap. U.S. private credit grew from roughly $500 billion five years ago to nearly $1.3 trillion, comparable in scale to parts of the bank loan and corporate bond markets. Globally, the market exceeds $2 trillion in 2026, with Moody's expecting a path toward $4 trillion by 2030. The vehicle mix matters. Direct lending dominates. Traditional investors, pensions and insurers, are now joined by retail channels: U.S. retirement savings and Europe's ELTIF 2.0 regime are opening the asset class to individuals. That democratization is the new fragility. A first methodological warning: there is no harmonized definition of private credit. The Financial Stability Board highlighted this in May 2026. The first blind spot is the headline number itself. Data that no one really sees The defining feature is not the yield; it is the absence of a market price. A syndicated loan can trade and reprice daily. A private loan is valued by the lender using a fair-value model, often quarterly. The manager is grading its own homework. The Fed and the FSB both stress the same issue: limited transparency and poorly mapped interconnections make systemic risk hard to assess. In early 2026 large pension funds bought loans from stressed private-credit vehicles at discounts. If they were willing to buy only below marks, they were implicitly saying the marked values were too high. The four red flags First: PIK interest, payment-in-kind. The borrower does not pay cash interest; it capitalizes the interest into more debt. For the fund, income exists on paper, not in the bank account. Rising PIK means borrowers cannot fully service debt in cash. Second: model valuation. Without a market, the lender marks the asset. Marks are smoothed and delayed, especially when no forced sale occurs. Third: NAV loans. A fund borrows against the net asset value of its portfolio, sometimes to fund distributions or meet redemptions. This adds leverage at the fund level and can become circular: borrowing against asset values to satisfy investors who doubt those values. Fourth: liquidity mismatch. Semi-liquid vehicles promise periodic redemption windows while owning illiquid loans. When exits rise, managers gate redemptions. The problem is not illiquidity itself; it is liquidity promised and then withdrawn. Vehicles: from readable to dangerous The wrapper matters. A listed BDC trades on an exchange and files with the SEC: it is the most transparent form. A non-traded BDC, whose assets have grown above $200 billion, offers periodic valuations and capped redemptions. Interval and evergreen funds operate similarly. A classic closed-end fund locks capital for years and does not pretend otherwise. The trap is the semi-liquid wrapper that sells illiquid credit as if it were nearly liquid. The 2025-2026 stress sequence The recent test was revealing. In September 2025, the failures of First Brands and Tricolor shook credit markets. Jamie Dimon's “cockroach” warning followed on October 15. Those cases involved fraud and were not pure senior private-credit defaults, but they exposed the system's sensitivity to hidden leverage and opaque collateral. Liquidity was the clearer test. In Q1 2026, redemption requests hit semi-liquid vehicles. Cliffwater reportedly faced requests around 14% of its flagship fund and capped redemptions at 7%. Morgan Stanley, Blue Owl and Blackstone vehicles also had to restrict or support exits. For the first time, the difference between perceived and actual liquidity became visible. The counterpoint matters. Losses remained contained, and Cliffwater's private debt index returned 9.33% in 2025 with realized losses near 0.70%. The episode looked less like a 2008-style collapse than an idiosyncratic fraud shock plus liquidity mismatch in retail-facing vehicles. How to read the risk The good news: BDCs, listed and non-listed, file with the SEC on EDGAR. Read them for the four red flags. Measure PIK as a share of interest income. Watch interest coverage, non-accrual loans, NAV leverage and valuation methodology. Compare distributions with cash flows and redemption queues. Look for sector concentration, especially software, tech and AI-linked borrowers. Private credit should also be read against public credit spreads. Public credit reprices daily; private credit reprices slowly. When public spreads move sharply and private marks do not, the gap may be delayed recognition, not an opportunity. The systemic question Is private credit the next crisis? The honest answer is debated. The cautious camp points to opacity, NAV leverage, growing bank-fund interconnections and retail distribution. The reassuring camp notes long capital lock-ups, moderate fund-level leverage and generally stable borrower fundamentals. The l0g reading is more about slow repricing than sudden apocalypse: a gradual recognition of losses, plus liquidity stress where semi-liquid vehicles meet impatient investors. Regulators agree the topic matters; the ECB and Bank of England have both built exploratory private-market stress scenarios. Methodology This guide uses primary institutional sources: the FSB's May 2026 report on private-credit vulnerabilities, Fed and IMF financial-stability materials, Fed research on bank lending to private credit, and SEC filings by BDCs. Practical l0g analysis begins with EDGAR filings, PIK share, interest coverage, non-accruals, fund-level leverage and valuation notes. --- Main sources: Financial Stability Board, Report on Vulnerabilities in Private Credit (May 6, 2026); IMF Global Financial Stability Report chapters on private credit; Federal Reserve Financial Stability Report and FEDS Note Bank Lending to Private Credit; Moody's Private Credit Outlook 2026; BIS work on AI-related corporate financing; Cliffwater Direct Lending Index; Financial Times, Bloomberg and Reuters coverage of 2025-2026 private-credit stress. ============================================================================ REFERENCE GUIDE: How to Read SEC Form 4: insider transactions without the noise URL: https://l0g.fr/en/guides/how-to-read-sec-form-4/ Canonical French source: https://l0g.fr/guides/analyser-form-4-sec/ Date: 2026-06-21 (reviewed 2026-06-21) ---------------------------------------------------------------------------- Form 4 is the freshest insider data in the U.S. market: two business days, versus forty-five days for a 13F. But freshness is not signal. Most Form 4 lines are accounting noise: stock grants, option exercises, tax withholding, sales scheduled months earlier. The real signal is rarer and more expensive to fake: an open-market purchase. What Form 4 is Form 4 is a filing that officers, directors and holders of more than 10% of a class of registered equity securities must file with the SEC within two business days after a change in their holdings. It is governed by Section 16 of the Securities Exchange Act. The premise is simple: those closest to a company must show the market what they do with their own securities. The current two-day deadline was introduced after Sarbanes-Oxley in 2002; before that, insiders could often wait until the tenth day of the following month. That change turned Form 4 into a near-real-time signal. Who files, and when Three groups are covered: executive officers, directors and beneficial owners of more than 10% of a registered equity class. They live inside a three-form system. Form 3 is the initial ownership report. Form 4 reports changes within two business days. Form 5 is an annual clean-up for exempt or missed transactions. A 2026 change widened the perimeter: on February 27, 2026, the SEC adopted a rule extending Section 16 to officers and directors of foreign private issuers, effective around March 18, 2026. For those foreign issuers, however, 10% shareholders remain outside the new regime. The filer universe therefore expanded in 2026. What is inside a Form 4 The filing has two tables. Table I covers non-derivative securities, mainly common stock. Table II covers derivatives: options, warrants, convertibles. Each line includes the transaction date, a one-letter transaction code, the amount, an acquisition or disposition marker, the price, and the number of securities owned after the transaction. A final column tells whether ownership is direct or indirect, for example through a trust or family vehicle. The “owned after transaction” number is often more useful than the transaction itself. A director selling 1,000 shares while retaining 200,000 is not saying the same thing as a CFO selling nearly everything. Transaction codes: filter before interpreting This is where most readings fail. Every line has a code, and not all codes are equal. Two codes show a voluntary market decision. P means purchase, open-market or private. S means sale. These are the only codes where the insider is deploying or recovering money at a market price. Most other codes are compensation or mechanics. A is a stock award. M is an option exercise. F is a surrender of shares to pay an exercise price or tax withholding; it looks like a sale but is often just tax plumbing. G is a gift, C is a conversion. Treating grants or tax withholding as buying or selling is the classic error. The first filter is therefore P and S. Why insider buying weighs more than selling Buying and selling are asymmetric. An insider buys for one plausible reason: they think the shares are attractive. They sell for many reasons: diversification, taxes, estate planning, option expiry, a house. Decades of finance research find insider purchases more predictive than insider sales. The law reinforces the signal. Section 16(b) requires insiders to disgorge short-swing profits from matched purchases and sales within six months. Section 16(c) prohibits insiders from short-selling their own company's stock. An open-market purchase is therefore costly: the insider cannot quickly flip it for a profit without giving up the gain, and cannot hedge it with a short. That makes it a more credible signal. The 10b5-1 checkbox Sales have a decisive filter: Rule 10b5-1. It lets insiders adopt prearranged trading plans when they do not possess material non-public information. A sale executed months later under such a plan may say almost nothing about the insider's current view. Since April 1, 2023, Forms 4 and 5 include a mandatory checkbox indicating whether a transaction was made under a 10b5-1 plan, along with the plan adoption date. The SEC also added cooling-off periods, restrictions on overlapping plans and good-faith certifications. For the reader, this is a major improvement: a sale checked as 10b5-1 and adopted well before the event can often be treated as programmed noise. A discretionary sale deserves more attention. Routine versus opportunistic The strongest filter is behavioral. Cohen, Malloy and Pomorski, in a 2012 Journal of Finance study, separate “routine” insiders who trade at the same time every year from “opportunistic” insiders whose trades break their own pattern. The result is sharp: after removing routine trades, opportunistic trades carry most of the predictive power. Routine sales, especially from executives who always sell in March, teach little. An unusual purchase by a CEO or CFO who rarely buys is the signal to isolate. How to read it on EDGAR Everything is public and free. Search the company on EDGAR, then filter for filing type 4. Open both the readable filing and the XML source. The workflow is simple. Identify the code: isolate P and S, set aside A, M and F. For sales, check the 10b5-1 box and the plan date. Read the filer's role: CEO and CFO purchases usually matter more than non-executive director purchases. Scale the transaction relative to the holder's remaining position. Finally, look for clusters: several insiders buying the same company within a short window is often the strongest pattern. Limits Form 4 is partial. One company, one insider, one transaction is statistically fragile. Sales are noisy by nature. Gifts and indirect ownership blur who economically holds what. A 10% shareholder may not know the business like an operating executive. Late filings still happen. Derivative tables require technical reading. One insider purchase is never a full thesis. Confluence with 13F A fast, narrow signal becomes stronger when crossed with a slow, broad one. Form 13F shows where institutional managers held long positions last quarter. Form 4 shows whether those inside the company are committing personal capital now. A position reinforced by respected managers and opportunistic insider purchases weighs more than either signal alone. This is the logic behind 13FLOW . Methodology This guide is based on primary sources: Section 16 of the Securities Exchange Act, SEC forms and instructions, SEC rules on 10b5-1 amendments and the 2026 extension of Section 16 to foreign private issuer officers and directors, plus peer-reviewed academic research on insider trading signals. A practical l0g reading starts from the raw EDGAR filing, isolates P and S, checks 10b5-1, weights the filer's role and refuses to turn one purchase into a complete investment case. --- Main sources: SEC, Section 16 of the Securities Exchange Act of 1934, Forms 3, 4 and 5 and instructions; Investor.gov, Insider Transactions and Forms 3, 4, and 5; SEC rules and compliance guidance on Rule 10b5-1 amendments; SEC 2026 final rule extending Section 16 to officers and directors of foreign private issuers; Cohen, Malloy and Pomorski, Decoding Inside Information, Journal of Finance, 2012. ============================================================================ ANALYSIS: KOSPI: anatomy of a liquidation concentrated in two stocks URL: https://l0g.fr/en/analysis/kospi-concentrated-liquidation-samsung-sk-hynix/ Canonical French source: https://l0g.fr/posts/kospi-liquidation-concentree-samsung-sk-hynix/ Date: 2026-08-10 (reviewed 2026-08-10) Topics: KOSPI, markets, leverage, ETFs, semiconductors, South Korea ---------------------------------------------------------------------------- On 28 July 2026, Samsung Electronics fell 13.4%, SK Hynix 14.7% and the KOSPI 10.8%. Concerns about AI infrastructure financing and Chinese competition supplied the initial shock. The structure of the Korean market did the rest: the two chipmakers jointly represented 52% of KOSPI market capitalisation on 15 July, up from 34% at the end of 2025. A fall in two stocks therefore became a fall in the index, then a mechanical constraint on leveraged products and credit-financed accounts. (Reuters, 28 July; Financial Services Commission, 16 July) The word “liquidation” still covers three different events. A 2x ETF has to rebalance its exposure every day. A broker can sell a client’s shares when collateral becomes insufficient. It can also unwind a purchase when settlement is not paid on time. All three channels can sell together, but their statistics measure neither the same object nor the same period. Adding them would produce a dramatic and false number. A 52% index built on two stocks Concentration is more than a weighting detail. It is the condition that turns local leverage into an index problem. According to the Korean regulator, the combined share of Samsung Electronics and SK Hynix in KOSPI market capitalisation rose from 34% at the end of 2025 to 41% at end-April 2026, 49% on 26 May and 52% on 15 July. In less than seven months, their weight gained 18 percentage points. (FSC, 16 July 2026) This dominance immediately transmits a memory-chip shock to every indexed portfolio. It also reduces the explanatory value of the headline level: “the KOSPI fell” can describe a broad crisis across Korean companies or, as here, an extraordinarily concentrated correction. A synthetic KOSPI excluding Samsung and SK Hynix is therefore the first useful counterfactual. Leverage enters an already concentrated market The regulatory calendar matters almost as much as the chart. On 21 April, the Korean government approved single-stock ETFs with maximum exposure of 200%. Eligibility notably required a stock representing at least 10% of market capitalisation and 5% of trading volume, plus an investment-grade credit rating. The regulator framed the reform as a way to close the gap with overseas products and reduce capital outflows to Hong Kong. (FSC, 21 April) On 27 May, eight asset managers launched 16 ETFs: fourteen long 2x products and two inverse 2x products, split evenly between Samsung Electronics and SK Hynix. Two leveraged ETNs were added on the same day. Prior training and a KRW10 million minimum deposit applied, but the entire range targeted two stocks that already accounted for 49% of the index one day earlier. (FSC, 26 May) For an individual, a 2x ETF can look safer than a margin account: the loss is capped at invested capital and the fund manages the derivatives. For the market, it nevertheless creates daily demand or supply that follows the underlying move. The wrapper pools leverage for investors, then concentrates rebalancing with the manager. The daily mechanics of 2x A 2x ETF targets twice the daily change in a stock, not twice its performance over several weeks. After a gain, it must increase exposure to start the next day at 200%. After a loss, it must reduce it. The rule makes it buy into gains and sell into losses. Daily compounding also creates decay when the market alternates sharply between gains and losses, even if the stock ends close to where it started. (FSC risk guide, 26 May) The Korea Capital Market Institute estimated the size of this mechanism after launch. On 19 June, SK Hynix 2x ETFs had roughly KRW8.7 trillion in assets on the previous day and the share gained 2.9%. The theoretical rebalance represented about KRW260 billion of cash-equity purchases and KRW270 billion in futures. Over its study period, estimated cash rebalancing averaged 1.6% of Samsung’s daily trading value and 2.1% of SK Hynix’s. These are portfolio estimates, not transaction-level measurements. Two accounting cautions matter. ETF assets include inventory held by liquidity providers and therefore do not map entirely to final demand. Futures do not erase impact either: arbitrageurs can transmit an imbalance between derivatives and cash shares. Flow size matters, but timing and market depth matter just as much. Three liquidation ledgers The first ledger belongs to the leveraged fund. Selling after a fall is contractual rebalancing. It can be forced by the replication rule without any investor default and without a net outflow from the fund. The second is margin credit, or 신용거래융자. The broker lends the cash used to buy securities and takes them as collateral. If the maintenance ratio falls below the required threshold and the client fails to add enough collateral, Korea Financial Investment Association rules allow the broker to sell. The national balance reached a record KRW38.63 trillion on 24 June, then fell to KRW34.37 trillion on 15 July, according to KOFIA data cited by Reuters. The KRW4.26 trillion drop measures total deleveraging, not forced sales alone. It also includes voluntary repayments and changes in balances. (Reuters, 20 July) The third is an unpaid settlement receivable, or 위탁매매 미수금. The broker briefly advances the funds for a purchase that the client must settle. If payment fails, it sells. From 1 to 10 July, these forced counter-trades tied to unpaid receivables amounted to KRW425.8 billion, including KRW142.2 billion on 9 July. This KOFIA statistic, reported by SBS, covers neither all margin credit nor ETF rebalancing. (SBS, 14 July) This breakdown changes the interpretation. The fall in the credit balance does show a leverage purge, but not its execution price. The unpaid-receivables statistic measures actual triggered sales, but for a narrow subset. ETF rebalancing can occur without investor redemptions. Liquidation is real, but no single public counter provides a consolidated total. The July purge By 15 July, the market capitalisation of the 16 leveraged products had risen from KRW4.4 trillion to KRW11.9 trillion since launch. Their daily trading value had increased from KRW10.4 trillion to KRW13.0 trillion. From 26 May to 10 July, the regulator calculated annualised volatility of 96% for Samsung and 113% for SK Hynix. (FSC, 16 July) These figures make amplification plausible, but they do not prove the ETFs created the reversal. The KCMI study also observes a sharp rise in volatility at Micron, in the SOX index and across other global semiconductor shares around the same period. Expectations for AI spending, data-centre financing and memory competition therefore remain independent triggers. Our analysis of the global chip relapse documents this move outside Seoul. The defensible causal statement is narrower: concentration transmitted almost the entire shock in two shares to the index; 2x ETFs added procyclical demand or supply; credit made some holders sensitive to the path of prices and collateral. Leverage can explain why the journey became so violent without explaining by itself why the market changed destination. The regulator applies the brake On 16 July, less than two months after launch, the FSC suspended new listings and advertising for single-stock products. It decided to raise the minimum deposit from KRW10 million to KRW30 million in cash, require an extra hour of training and increase the minimum trading unit from 1 to 20 shares. On 24 July, it brought forward the stronger cash requirement to 31 July. (FSC, 16 July; FSC, 24 July) On 28 July, the regulator asked asset managers to spread rebalancing through the session instead of concentrating it at the close. It acknowledged the trade-off: a more dispersed execution may reduce instantaneous impact but increase tracking error if the stock moves before the close. The FSC also floated an individual limit, such as 20% of an investor’s financial assets, without adopting it at that stage. (FSC, 28 July) This sequence forms a rare regulatory experiment. The April opening sought to repatriate demand that had moved to Hong Kong. The July brake admits that domesticating the product does not remove its microstructure risk. It only makes that risk more visible and more directly regulable. A closing-auction hypothesis to test A preprint published on 4 August pushes the analysis further. It argues that predictable end-of-day rebalancing lets other market participants buy or sell ahead of the ETFs, then transfer the position to them at a less favourable price. The loop would no longer be merely procyclical: it would be partly anticipated by the market. This thesis is not established fact. The paper labels itself preliminary and incomplete, has not been peer reviewed and discloses AI assistance. Its wealth-transfer estimates depend on a counterfactual model. It nevertheless provides a falsifiable hypothesis consistent with the regulator’s concern over order concentration at the close. Four tests would help settle it: 1. compare Samsung and SK Hynix returns in the final minutes with the rest of the session, before and after 27 May; 2. match predicted rebalancing flows against actual closing-auction and futures volume; 3. track a KOSPI excluding the two shares to separate broad Korean stress from concentration; 4. use Micron, Kioxia and the SOX as controls for the global chip cycle. If the closing effect disappears, if the KOSPI without Samsung and Hynix falls just as much and if foreign peers follow the same path, the leverage explanation weakens markedly. If a gap appears precisely after launch, grows with product assets and recedes as execution is dispersed intraday, it becomes stronger. Blind spot and verdict Public data do not yet allow transaction-level consolidation of the three layers. The FSC publishes product market capitalisation and volumes. KOFIA publishes credit and counter-trades on receivables. KCMI reconstructs rebalancing from portfolios. None of these datasets links each order to the fund, credit account or arbitrageur that submitted it. Estimates of total liquidation must therefore remain attributed and accompanied by their method. The robust diagnosis fits in one sentence: a global chip shock hit two shares that had become half the KOSPI, then distinct leverage layers could amplify its path. Concentration made the market fragile. 2x ETFs made part of demand mechanical. Credit turned volatility into a collateral constraint. Liquidation is not a rival explanation to the fundamental shock; it is the mechanism through which that shock could become a cascade. For related plumbing, read our work on the Treasury basis trade and forced unwinds, the carry-trade guide and the definition of a margin call. For the sector trigger, see the stack of semiconductor constraints. Sources - Financial Services Commission, single-stock ETF authorisation and 200% cap, 21 April 2026 - Financial Services Commission, launch of 16 ETFs and two ETNs, 26 May 2026 - Financial Services Commission, concentration, product data and emergency measures, 16 July 2026 - Financial Services Commission, accelerated cash-only deposit, 24 July 2026 - Financial Services Commission, intraday rebalancing and possible individual limit, 28 July 2026 - Korea Capital Market Institute, single-stock ETFs and retail investor flows, June 2026 - Korea Financial Investment Association, rules on margin credit and collateral - Reuters, KOSPI concentration and margin credit, 19 July 2026 - Reuters, retail margin-credit risks, 20 July 2026 - Reuters, mechanics of Korean leveraged ETFs, 29 July 2026 - SBS, KOFIA data on forced counter-trades tied to unpaid receivables, 14 July 2026 - Yinhong Zhao, “Preying on Leveraged ETFs”, preprint dated 4 August 2026 Limitations Market moves quoted here are dated and sourced; they are not real-time prices. KCMI rebalancing flows are estimates based on portfolio structure. The decline in margin-credit balances does not measure forced sales alone. Unpaid settlement receivables cover a narrower scope than margin credit. The 4 August preprint is neither final nor peer reviewed. Finally, no public source reviewed for this article yet provides a consolidated register linking every July sale to its leverage layer. ============================================================================ ANALYSIS: Epstein and Europe's banks: six relationships that require separate proof URL: https://l0g.fr/en/analysis/epstein-european-banks-six-relationships-not-to-confuse/ Canonical French source: https://l0g.fr/posts/epstein-banques-europeennes-six-relations-a-ne-pas-confondre/ Date: 2026-08-09 (reviewed 2026-08-09) Topics: Jeffrey Epstein, Ghislaine Maxwell, European banks, Deutsche Bank, Edmond de Rothschild, HSBC, BNP Paribas, UBS, Barclays, compliance, investigation ---------------------------------------------------------------------------- Six European banks appear in the public records reviewed for this investigation. But they do not appear in the same capacity or with the same standard of proof. Deutsche Bank provided banking services directly to Jeffrey Epstein. HSBC and BNP Paribas held accounts in his name under different circumstances. UBS and Barclays are documented mainly through their relationships with Ghislaine Maxwell. Edmond de Rothschild Holding appears as the principal behind an engagement awarded to Southern Trust, a company chaired by Epstein. Reading these records together can show how risk travels. Merging them into a single accusation would distort the sources. NYDFS, Deutsche Bank order of 6 July 2020 · Bloomberg, 21 November 2025 · Le Monde, 21 February 2026 · Reuters, 8 February 2026 · Upper Tribunal, 26 June 2025 · agreement dated 5 October 2015, EFTA00587465 The distinction prevents two errors. The first would be to treat every appearance in a file as proof of an account. The second would be to turn an account or transfer into automatic proof of an offence. A banking record first establishes a relationship, date or transaction. Separate evidence is needed to characterise its purpose, economic beneficiary or possible criminal nature. This investigation complements our series on Epstein's money. The mechanisms already established at Deutsche Bank are detailed in a separate investigation. The Staley case and Maxwell's Barclays account are covered in our dedicated analysis. The purpose here is comparative: which relationship does each source establish, and where does the evidence stop? Edmond de Rothschild: a documented paid engagement An agreement dated 5 October 2015 states that Southern Trust Company had been collaborating with Ariane de Rothschild, on behalf of Edmond de Rothschild Holding, on outstanding matters between the holding company and the United States. It provides for the Rothschild Group to pay $25 million to Southern Trust no later than three days after Edmond de Rothschild Holding made a payment to the United States. The document identifies Epstein as Southern Trust's president and says the work was to continue as agreed from time to time between Epstein and Ariane de Rothschild. EFTA00587465, pages 1 and 2, indexed copy from FBI VOL00009 A transaction table published among the US records then lists two payments to Southern Trust: $10 million on 17 December 2015, with Edmond de Rothschild (Suisse) SA as the originator, then $14,999,980 on 21 December, with Benjamin Edmond de Rothschild as the originator. The total is $24,999,980. The two entries do not have the same payer, and the record does not explain why the combined amount is twenty dollars below the contractual fee. Transaction table, first page of EFTA00027019 PDF On 18 December 2015, the US Department of Justice announced that Edmond de Rothschild (Suisse) SA and Edmond de Rothschild (Lugano) SA would pay $45.245 million under the Swiss Bank Program. The release concerns undeclared US accounts and potential tax offences. It does not concern Epstein's crimes. The proximity of the dates to the agreement and payments is documented; on its own, it does not establish corruption, intervention with the department or an unlawful quid pro quo. DOJ, 18 December 2015 · EFTA00587465 The rigorous formulation is therefore narrow: the records establish an engagement of Southern Trust connected to the holding company's US matters and payments almost equal to the stipulated fee. By themselves, they do not establish that Epstein held an account at Edmond de Rothschild, had power over the DOJ resolution or caused the penalty to change. HSBC: a French closure, three Swiss accounts without a timeline A letter dated 21 December 2007, obtained by Bloomberg, told Epstein that HSBC did not intend to maintain the relationship and was closing his Paris account at the bank's request. The letter gives no reason. Bloomberg reports, based on two people familiar with the matter, that compliance staff had flagged suspicious transactions. That explanation is attributed to the anonymous sources; it does not appear in the letter itself. The closure became effective on 21 January 2008, according to an email from Epstein's lawyer also published by Bloomberg. Bloomberg, 21 November 2025 That closure does not end the HSBC record. A suspicious activity report filed by JPMorgan after Epstein's death and later unsealed lists him with three accounts at HSBC's Swiss private bank. Reuters notes that the record provides neither amounts nor details of the relationship. It therefore does not show when the accounts were opened or closed, or whether they remained active after the Paris closure. A JPMorgan report establishes information sent to the US Treasury, not a judicial finding about HSBC. Reuters, 4 November 2025 The material gap is chronological: a French entity closed a relationship before Epstein's 2008 conviction, while Swiss accounts appear in a later record that gives no dates. Nothing in the public sources reviewed establishes whether the French and Swiss teams shared the same file at the time, or whether they did not. BNP Paribas: a Fortis retail account inherited in an acquisition The BNP Paribas case begins at another institution. Le Monde reports that Fortis France opened a retail current account in Epstein's name in February 2008. BNP Paribas subsequently inherited it when it acquired Fortis, along with thousands of other accounts. Records reviewed by the newspaper also show a savings account and bank card. BNP Paribas told Le Monde that this was not a private banking relationship and that the bank closed the account on its own initiative, for compliance reasons, in mid-2018. Le Monde, 21 February 2026 The opening preceded Epstein's Florida conviction in June 2008 by several months. The Fortis acquisition therefore creates a problem different from knowingly accepting a new client who had already been convicted: it raises the question of reviews conducted when a portfolio is integrated and when public risk changes. The published records do not establish that the BNP Paribas account funded Epstein's crimes. Le Monde says that use remains unknown. Le Monde, 21 February 2026 · DOJ notification of the 30 June 2008 plea and sentence, EFTA00013888 UBS: the client rejected, the linked person retained The documents analysed by Reuters describe two separate decisions. UBS provided Epstein with a credit card in 2014 after his JPMorgan relationship ended. According to an email from his accountant, the bank closed it in September that year because of “reputational risk”. At the same time, UBS opened personal and business accounts for Ghislaine Maxwell in 2014 and managed as much as $19 million for her in later years. The account holder, products and duration were therefore different. Reuters, 8 February 2026 Reuters saw records showing that UBS conducted due diligence before Maxwell's accounts moved from JPMorgan, but it could not establish the details. The news agency said there was no evidence of wrongdoing by UBS or its advisers. It nevertheless documented a question of scope: JPMorgan had classified Maxwell as a high risk client in 2011 because of her links to Epstein, while UBS rejected Epstein for reputational risk but maintained its relationship with Maxwell. Reuters, 8 February 2026 On 22 July 2019, sixteen days after Epstein's arrest, UBS moved $130,000 from Maxwell's savings account to her current account to help pay an American Express bill, according to the documents. On 16 August, the bank received a grand jury subpoena concerning Maxwell and provided the FBI with wire transfer information. Reuters could not determine whether or when UBS closed her accounts. Executing a transaction after an arrest or complying with a subpoena does not itself prove a breach; the dates make a complete timeline of monitoring and exit decisions necessary. Reuters, 8 February 2026 Deutsche Bank and Barclays: two records already established Deutsche Bank is the most clearly regulated case in this comparison. The New York financial regulator's order describes a direct relationship with Epstein, more than forty accounts connected to him and his entities, known risks at entry and repeated monitoring failures. The relationship, its exceptions and its exit are detailed in our Deutsche Bank investigation and the second instalment of our banking series. NYDFS, order of 6 July 2020, pages 6 to 24 Barclays must remain on two planes. In 2025, the UK Upper Tribunal found that Jes Staley had knowingly approved two misleading statements about his relationship with Epstein. Separately, documents analysed by Reuters indicate that Maxwell held $2.4 million at Barclays at the end of 2018 and that more than $600,000 moved from that account to UBS in the three weeks after Epstein's arrest. Neither record establishes a Barclays account in Epstein's name. The full chain is set out in our Barclays investigation. Upper Tribunal, paragraphs 496 to 503 · Reuters, 27 March 2026 Public questions for the holders of the records The sources support precise questions without presuming their answers. The institutions and people named below may hold some or all of the information. Banking secrecy, an investigation, data protection or other obligations may limit publication. A lack of publication would confirm no hypothesis. 1. Who approved the Southern Trust engagement, and on the basis of which checks? Edmond de Rothschild Holding, its 2015 governance bodies and Ariane de Rothschild could publish a redacted chronology of the decision, requested work, checks performed and deliverables received. The agreement establishes the fee and a general purpose, not the concrete substance of the service. EFTA00587465 2. Why were the payments split between Edmond de Rothschild (Suisse) SA and Benjamin Edmond de Rothschild? Records held by the bank, holding company and Benjamin de Rothschild's estate could document the accounting and contractual basis for that division. The table establishes two originators; it does not give the reason. EFTA00027019, first page 3. When were the three HSBC Switzerland accounts opened and closed? HSBC Private Bank (Suisse) and group archives could provide a redacted timeline. HSBC France could say whether the 2007 Paris closure decision and due diligence file were shared with the Swiss entity. Reuters documents the accounts, not their dates. Reuters, 4 November 2025 · Bloomberg, 21 November 2025 4. When did BNP Paribas reassess the account inherited from Fortis? BNP Paribas and the Fortis integration archives could specify the reviews conducted after the acquisition, after the June 2008 conviction and before the 2018 closure, together with the exact blocking date. Those answers could be given without revealing the existence or absence of a confidential report. Le Monde, 21 February 2026 5. Which perimeter did UBS use to assess Maxwell? UBS and its compliance archives could explain which Epstein links, entities and beneficial owners were examined in 2014 and reassessed after July 2019. Reuters saw evidence that checks existed but could not learn their details. Reuters, 8 February 2026 6. On which dates were transfer capabilities and Maxwell's UBS accounts restricted or closed? UBS could publish a redacted timeline of decisions taken before and after the 16 August 2019 subpoena. The FBI and federal prosecutors could index any releasable letters and records that establish the rest of that chronology. Reuters, 8 February 2026 7. Have European regulators reconstructed the cross-border flow of information? The ACPR, FINMA, FCA and other competent authorities could publish, within legal limits, the dates and scope of their reviews and the expectations applicable to information sharing between entities in the same group. The question concerns supervisory method, not disclosure of an individual report. 8. Can the US Department of Justice connect every documentary claim to a stable record? A public index linking agreements, transaction tables, emails and statements to their full identifiers would allow further checking of press investigations without republishing personal information about victims or third parties. Method: identify each relationship and its evidence These six records do not describe a single European “Epstein bank”. They show an architecture that is harder to control: a direct client can leave one institution and reappear elsewhere; an account can be inherited in an acquisition; a linked person can retain a separate banking relationship; a company controlled by the client can become a service provider to a financial group; and a senior executive can become the principal source for a review of his own conduct. The strongest conclusion is therefore not a ranking of banks. It is a requirement of method: identify the exact account holder, legal entity, product, period, payment purpose and source that establishes each point. Whenever one is missing, the question should remain open and clearly name who may be able to answer it. Lire la version française. Sources - New York State Department of Financial Services, Deutsche Bank order, 6 July 2020 - Upper Tribunal, James Edward Staley v. Financial Conduct Authority, 26 June 2025 - US Department of Justice, Swiss Bank Program resolution with Edmond de Rothschild, 18 December 2015 - EFTA00587465, agreement dated 5 October 2015 between Southern Trust and Edmond de Rothschild - EFTA00027019, transaction table showing payments to Southern Trust - Bloomberg, closure of the HSBC Paris account, 21 November 2025 - Reuters, HSBC Switzerland accounts listed in a JPMorgan report, 4 November 2025 - Le Monde, BNP Paribas account inherited from Fortis, 21 February 2026 - Reuters, Maxwell's UBS accounts, 8 February 2026 - Reuters, Maxwell's Barclays account and transfers to UBS, 27 March 2026 Limitations - Some facts concerning HSBC, BNP Paribas, UBS and Barclays rely on journalistic analysis of documents that have not all been published in a stable, indexed form. Those facts are attributed to the news organisations that established them. - EFTA00587465 was reviewed through an indexed copy with automated text extraction. EFTA00027019 is a public copy from a US case file. Their presence in the files does not by itself provide a legal characterisation of the service or payments. - Edmond de Rothschild's tax resolution with the DOJ is separate from Epstein's crimes. The proximity of dates and the agreement's wording justify a question about the engagement; they do not prove corruption, influence or a reduced penalty. - An account, card, transfer or relationship with a linked person does not itself prove an offence by the bank. This article does not characterise any funds as criminal without specific evidence. - Maxwell's accounts are legally distinct from Epstein's accounts. Their existence does not permit the account holder's name to be changed or the funds to be automatically attributed to Epstein. - The detailed reasons for the HSBC Paris closure rely on two anonymous sources cited by Bloomberg. The closure letter itself gives no reason. - Other institutions appear in political lists, emails or press investigations. They are not added to this comparison because the public evidence reviewed is insufficient to establish a banking or contractual relationship of a comparable nature. - The source record is current to 9 August 2026. The absence of a public document does not prove the absence of an internal review, confidential report or investigation. ============================================================================ ANALYSIS: Barclays, Staley and Maxwell: two records, one boundary URL: https://l0g.fr/en/analysis/barclays-staley-maxwell-two-records-one-boundary/ Canonical French source: https://l0g.fr/posts/barclays-staley-maxwell-deux-dossiers-une-frontiere/ Date: 2026-08-09 (reviewed 2026-08-09) Topics: Barclays, Jes Staley, Ghislaine Maxwell, Jeffrey Epstein, FCA, governance, banks, investigation ---------------------------------------------------------------------------- Barclays now appears in the public record on two separate planes. The first concerns Jes Staley and the bank's governance: in 2025, the UK Upper Tribunal found that its former chief executive had knowingly approved two inaccurate statements that risked misleading the Financial Conduct Authority. The second concerns Ghislaine Maxwell: US Department of Justice documents analysed by Reuters indicate that she held $2.4 million at Barclays at the end of 2018 and that more than $600,000 was later transferred from that account to UBS in the three weeks after Jeffrey Epstein's arrest. These two findings do not prove that Epstein himself held an account at Barclays. Upper Tribunal, decision released 26 June 2025 · Reuters, 27 March 2026 That boundary is essential. The Staley ruling concerns the accuracy of a response sent to a regulator and the integrity of a senior manager. The Reuters investigation concerns an account held by another person, Ghislaine Maxwell, who was convicted in the United States and is legally distinct from Epstein. Placing these records on the same timeline can illuminate Barclays' controls. Merging them into a single accusation would produce a conclusion that the sources do not support. Upper Tribunal, paragraphs 496 to 503 · Ghislaine Maxwell criminal judgment, SDNY, 29 June 2022 The comparison with other banks is covered in our investigation into the signals financial institutions could see. Deutsche Bank's commercial exception mechanism is examined in a separate analysis. The question here is narrower: what did Barclays verify, whom did the bank rely on, and what does the Maxwell account now reveal? First plane: an assurance built around Staley On 15 August 2019, the FCA asked Barclays for written assurance: had the board sufficiently informed itself, and was it satisfied with any association between Jes Staley or Barclays and Jeffrey Epstein? The question was not limited to possible knowledge of the crimes. The Tribunal found that it also concerned the nature of the association and the checks conducted by the bank. Upper Tribunal, paragraphs 318 to 371 Several Barclays officials prepared the response using conversations with Staley and documents assembled during the summer of 2019. The Tribunal found that, by 8 October, the bank did not have a complete and accurate picture of the relationship. It did not know, among other things, its depth, the frequency of contact after Epstein's conviction, or that the contact continued until late October 2015. This finding does not mean that the board knew the letter was false. It means that the information it received from Staley was incomplete and differed from the judges' later assessment. Upper Tribunal, paragraphs 316 and 317 On 1 October, a first draft stated that Barclays' financial crime team had reviewed the bank's records and found no client relationship with Epstein or his "known affiliates". The reference to affiliates disappeared from the final text. The letter that was sent said only that the review had found no client relationship with Epstein. The public record establishes this edit, but not its reason or author. Upper Tribunal, paragraphs 405 to 433 That negative finding must be read literally. It was the stated result of a review of Barclays' records. It supports neither a claim that the method was exhaustive nor the invention of an account that was not found. The public ruling does not list the names, entities, spelling variants, counterparties, periods or databases that were checked. Upper Tribunal, letter text and paragraphs 405 to 433 Circular validation in the emails On 6 October, Barclays' general counsel Bob Hoyt sent the nearly final draft to Staley and asked whether its language was "fair and accurate". The two spoke for about five minutes the next day. The Tribunal accepted Hoyt's account: he told Staley that only Staley knew the facts and had to decide whether the wording was appropriate. At 12:37 pm, Hoyt wrote to Higgins that Staley had reviewed the letter and was comfortable with its language. The only change then attributed to Staley concerned the tense of a sentence unrelated to the two disputed statements. Higgins sent the text on 8 October. Upper Tribunal, paragraphs 422 to 433 The control weakness is specific: Barclays officials believed the relationship was not close because Staley had described it that way, then asked Staley to certify that a summary derived from his own account was accurate. Hoyt said that he did not know about the emails that emerged later and relied on Staley's judgment to characterise the relationship. Higgins testified that, with the information later known, Barclays would probably have responded differently, might not have sent a letter, or might have opened a discussion with the FCA. Upper Tribunal, paragraphs 475 and 476 In November 2019, JPMorgan told the FCA that it had identified documents in the course of US investigations. The FCA then compelled the bank to produce the exchanges. The record includes more than 1,200 emails between 2008 and 2012. Those emails concern the JPMorgan period, not banking services supplied by Barclays. They illuminate what Staley knew about his own relationship and what he had not described to Barclays. Upper Tribunal, paragraphs 7, 105 and 434 On 26 June 2025, the Tribunal found that Staley knew the two statements were inaccurate, understood the risk of misleading the FCA and acted without integrity in approving them. The FCA's final notice dated 23 July 2025 set the fine at £1,107,306.92 and barred him from senior management or significant-influence functions in UK regulated financial activities. These are personal sanctions against Staley, not a fine imposed on Barclays. Upper Tribunal, paragraphs 496 to 503 and 559 to 561 The decision answers two questions: the closeness of the relationship and the date of the last contact. It does not find that Staley participated in Epstein's crimes, nor does it find that the Barclays board knew the letter was false. When Staley stepped down on 1 November 2021, Barclays stressed that the regulators' preliminary findings did not say he had seen or known about Epstein's crimes. That limit does not alter the later conclusion about the letter. Barclays, 1 November 2021 statement filed with the SEC · Upper Tribunal, paragraphs 473 to 503 Second plane: a Barclays account in Maxwell's name The second plane rests on a Reuters investigation published on 27 March 2026. The reporters examined files made public by the US Department of Justice, banking exchanges and a 2020 wealth report prepared for the federal court in Manhattan by accountants hired by Maxwell's lawyers. According to that analysis, Barclays was, from 2017, the only institution outside the United States where Maxwell held funds. She held $2.4 million there at the end of 2018. Reuters, 27 March 2026 Reuters also reported that, in the three weeks after Epstein's arrest on 6 July 2019, UBS received more than $600,000 from Maxwell's Barclays account. The exchanges reviewed indicate that she was gathering money to pay a credit-card bill. This wording documents the immediate banking origin and destination of the transfers. It establishes neither the ultimate economic origin of each dollar nor whether the funds were lawful or unlawful. Reuters, 27 March 2026 The account cannot be described as Epstein's account. The reported holder was Maxwell. Proximity between two people, even where one was convicted in a related criminal case, does not erase their separate legal identities or the evidentiary requirements attached to a bank account. No public source examined for this article establishes that Epstein was a Barclays customer or that an account was opened in his name. Reuters, 27 March 2026 · Upper Tribunal, text of the 8 October 2019 letter That caution does not cancel the compliance questions. Maxwell had been publicly associated with Epstein well before 2017. The material now published therefore permits questions about Barclays' onboarding, risk classification, monitoring and response after Epstein's arrest. In its current public form, it does not provide the account's exact opening or closing date, the relevant Barclays entity, the contents of the customer due-diligence file, any alerts or the decisions made about the transfers to UBS. Reuters, 27 March 2026 The records review described in the 2019 letter makes the issue more concrete. The first draft mentioned Epstein and his known affiliates, while the final version named only Epstein. The public record does not say whether Maxwell was searched by name, whether she fell within the intended category, or whether her account was connected to the review concerning the chief executive. This absence of information does not prove that she was omitted. It simply prevents a claim that she was included. Upper Tribunal, paragraphs 405 to 433 · Reuters, 27 March 2026 Investor litigation narrows without a liability ruling In Merritt v. Barclays, investors brought a putative class action concerning New York-listed ADRs and London-listed ordinary shares bought between 22 July 2019 and 12 October 2023. On 10 July 2025, the federal court in California allowed several US-law claims to move beyond the pleading stage, while finding the claim under section 90A of UK law insufficiently pleaded. At that stage, the judge was required to treat properly pleaded allegations as true without finding that they had been proved. C.D. California, amended order dated 10 July 2025 On 28 July 2026, the same judge dismissed the revised UK claim without leave to amend. The order required the plaintiffs to file a third amended complaint omitting that claim alone and expressly said that the federal Exchange Act claims continued to be litigated. The dismissal of the UK claim is therefore neither a finding of liability against Barclays nor the documented end of the entire case. C.D. California, docket 160 order dated 28 July 2026 Political pressure in 2026 targets the board's method On 22 July 2026, Senator Elizabeth Warren and Representatives Ro Khanna and Raja Krishnamoorthi sent eight groups of questions to Barclays chair Nigel Higgins. They asked, among other things, which checks preceded Staley's recruitment, who drafted and approved the 2019 letter, why the full board was not consulted, what disciplinary measures were considered, and whether US banking regulators had contacted Barclays after the new DOJ releases. The letter requested a response by 5 August. These congressional questions are not judicial findings. Letter dated 22 July 2026 · Congressional press release, 23 July 2026 On 23 July, Barclays told Reuters that it had investigated on the basis of the information then available and noted that the UK regulator had found that Staley misled the bank. Barclays added that new information had since emerged concerning activities before he joined the bank. This position is consistent with the Tribunal's finding that Barclays lacked a complete picture in 2019. In the sources identified by 9 August, it does not publicly answer the detailed questions about recruitment, the scope of the records review or the Maxwell account. Reuters, 23 July 2026 · Upper Tribunal, paragraphs 316 and 317 Eight public questions that remain open The documents support precise questions without presuming their answers. The people and institutions named below may hold some or all of the information. Banking secrecy, employment law, investigative confidentiality or other duties may also prevent them from publishing it. Their silence would confirm no hypothesis. 1. Which checks preceded Staley's recruitment in 2015? Barclays and the members of its nomination committee at the time could specify the questions asked about his relationship with Epstein, the independent references sought and the written declarations obtained. The 22 July congressional letter already asks Barclays to release this material. Congressional letter, questions 1 and 2 2. Why did the 2019 assurance rely mainly on Staley to characterise his own relationship? Nigel Higgins, Bob Hoyt, Crawford Gillies and the board records could document which independent checks were considered, performed or rejected. The FCA and PRA could clarify the expectations communicated to the bank, subject to their confidentiality rules. Upper Tribunal, paragraphs 318 to 433 3. Who proposed and approved the removal of "known affiliates" between the 1 October draft and the final letter? Versions and emails held by Barclays, together with FCA records, could provide a chronology. The change is established; its reason is not. Upper Tribunal, paragraphs 405 to 433 4. What was the exact scope of the records review? Barclays could identify, in redacted form, the names, entities, variants, counterparties, periods and systems searched. The FCA could say whether it tested or audited that method. The Tribunal published the stated result, not the complete protocol. Upper Tribunal, letter text and paragraphs 405 to 433 5. Did the 2019 review include Ghislaine Maxwell and her account? Barclays is best placed to answer. The FCA could clarify whether it asked for a search of people closely associated with Epstein. An answer should distinguish the existence of an account, its risk classification and any alerts without exposing unnecessary personal data. Reuters, 27 March 2026 · Upper Tribunal, paragraphs 405 to 433 6. When was the Maxwell account opened, reassessed and closed? Barclays could publish a redacted chronology naming the relevant entity, review dates and decisions taken after Epstein's arrest. The DOJ documents and wealth report cited by Reuters could supplement that chronology if they are published in indexed and redacted form. Reuters, 27 March 2026 7. How were the transfers of more than $600,000 to UBS reviewed? Barclays and UBS could describe the controls applied without revealing any SAR, whose existence or absence must not be inferred. The relevant authorities could say whether the movements were included in a regulatory review. Reuters documents the transfers, not Barclays' compliance reasoning. Reuters, 27 March 2026 8. Did Barclays answer the US lawmakers by 5 August, and will that response be published? Nigel Higgins, Elizabeth Warren, Ro Khanna and Raja Krishnamoorthi could release a redacted response or state its status. As of 9 August 2026, no detailed public response was identified in the sources consulted for this article. Congressional letter, page 5 One institution, two standards of proof The Staley record shows a bank that questioned its chief executive, reviewed its records and answered the regulator, but did not have a complete account of the relationship it was assessing. The Tribunal placed regulatory responsibility on Staley: he knew the facts, approved the two statements and knew they could mislead the FCA. The governance question remains because Barclays' method allowed the subject of the review to become the principal source validating his own account. Upper Tribunal, paragraphs 316, 317 and 473 to 503 The Maxwell account opens a different field. It does not make Barclays "Epstein's bank". It documents a banking relationship with a person closely associated with Epstein, $2.4 million present at the end of 2018 and more than $600,000 transferred to UBS after the July 2019 arrest. The rigorous question is not whether the account holder's name can be changed. It is how the account was accepted, monitored, connected to known risk and handled when the context changed. Reuters, 27 March 2026 Lire la version française. Sources - [Upper Tribunal, James Edward Staley v. Financial Conduct Authority, [2025] UKUT 00203, 26 June 2025](https://www.gov.uk/tax-and-chancery-tribunal-decisions/james-edward-staley-v-the-financial-conduct-authority-2025-ukut-00203-tcc) - Financial Conduct Authority, final notice concerning James Staley, 23 July 2025 - Financial Conduct Authority, statement on the Tribunal decision, 25 July 2025 - Barclays, leadership change, SEC filing dated 1 November 2021 - Reuters, Maxwell's Barclays account and transfers to UBS, 27 March 2026 - US District Court, SDNY, Ghislaine Maxwell criminal judgment, 29 June 2022 - US District Court, C.D. California, Merritt v. Barclays, amended order dated 10 July 2025 - US District Court, C.D. California, Merritt v. Barclays, docket 160 order dated 28 July 2026 - Elizabeth Warren, Ro Khanna and Raja Krishnamoorthi, letter to Nigel Higgins, 22 July 2026 - Reuters, Jes Staley hearing and Barclays' public position, 23 July 2026 Limitations - The Upper Tribunal decision concerns Staley's conduct and his letter to the FCA. It is neither a criminal conviction of Staley nor a liability ruling against Barclays for Epstein's crimes. - The Maxwell account and transfer amounts rest on Reuters' analysis of DOJ documents, emails and a wealth report. Indexed public records do not allow an independent reconstruction of every movement or the account's full operation. - Holding an account or making a transfer does not by itself prove an offence. This article does not characterise the funds as criminal and infers no SAR from their existence. - The account attributed to Maxwell does not prove an account in Epstein's name, Epstein's control of that account or services supplied directly to Epstein by Barclays. - The US lawmakers' letter asks questions and repeats some media allegations. It is not a court decision and does not prove its own hypotheses. - The Merritt decisions concern the sufficiency of civil claims. They do not decide liability on the merits. The 28 July 2026 order dismisses the UK claim without leave to amend and leaves the US federal claims pending as of that date. - The source review and the absence of a detailed public response from the lawmakers or Barclays are current to 9 August 2026. Non-publication does not prove the absence of a confidential response, investigation or regulatory work. ============================================================================ ANALYSIS: Deutsche Bank and Epstein: the compliance exception URL: https://l0g.fr/en/analysis/deutsche-bank-epstein-compliance-exception/ Canonical French source: https://l0g.fr/posts/deutsche-bank-epstein-exception-conformite/ Date: 2026-08-09 (reviewed 2026-08-09) Topics: Deutsche Bank, Jeffrey Epstein, compliance, money laundering, NYDFS, banks, investigation ---------------------------------------------------------------------------- In April 2013, a memorandum prepared for Deutsche Bank’s wealth-management leadership placed two facts on the same page. One was Jeffrey Epstein’s criminal history, prison sentence and the public allegations surrounding his relationships with young women. The other was a commercial forecast: $100 million to $300 million of flows and $2 million to $4 million of annual revenue over time. The risk was not hidden. It was already part of the client-acquisition file. That juxtaposition, documented in the New York financial regulator’s consent order, does not prove that expected revenue bought a compliance decision. It establishes something narrower: the bank knew the criminal record and reputational risk before opening the first accounts while also quantifying the relationship’s economic appeal. Deutsche Bank has since acknowledged that onboarding Epstein in 2013 was an error, regretted its historical association with him and pointed to its investment in stronger controls. The NYDFS, for its part, called the handling of the accounts a major compliance failure while crediting the bank’s cooperation and remediation. Both elements belong in the record. Deutsche Bank statement, updated in 2026 · NYDFS, paragraphs 55 to 59 and 109 to 117 The issue is therefore not to invent a hidden motive for individual bankers. It is to understand an institutional mechanism: how a client correctly labelled high-risk could become, transaction after transaction, the internal benchmark for his own normality. The comparison across several institutions appears in our investigation into what the banks saw. The analysis here focuses on Deutsche Bank: its decision chain, the gap between an announced exit and its implementation, and the exact limits of the public penalties and civil litigation. The risk was in the file The relationship originated with a relationship manager who moved from a competitor to Deutsche Bank in November 2012. According to the NYDFS, he soon presented Epstein to senior management as a potential source of millions of dollars in revenue and introductions to other wealthy clients. A coordinator prepared the April 2013 memorandum summarising the criminal record and civil settlements already in the public domain. NYDFS, paragraphs 16 to 21 The 5 May approval did not result from a formal meeting of the regional reputational-risk committee. An executive wrote that he had consulted the legal and anti-money-laundering heads orally. The regulator said the bank could find no other record of that conversation and that the committee did not meet for the onboarding. The emailed approval then became the basis for opening more accounts. NYDFS, paragraphs 22 to 25 The first accounts opened on 19 August 2013 for Southern Trust Company and Southern Financial LLC. Epstein, related entities and associates would eventually open and fund more than 40 accounts. The bank rated the relationship high-risk and subjected it to enhanced due diligence. The NYDFS nevertheless found that the monitoring was not tailored to the specific risks already identified. NYDFS, paragraphs 24 to 26 That distinction matters. A high KYC rating is not a safeguard by itself. It works only through the control rules it triggers, the people who receive them and the decisions made when real activity does not fit the expected profile. See KYC, AML and SAR in the glossary. In 2015, control tightened on paper New reporting and legal developments prompted an internal escalation at the end of 2014 and the start of 2015. Two executives met Epstein at his home on 22 January 2015. The bank told the NYDFS that it did not possess a contemporaneous record of the discussion and was not aware of other steps taken at the time beyond asking the client himself about the allegations. NYDFS, paragraphs 33 to 37 On 30 January, the reputational-risk committee allowed the relationship to continue. No detailed minutes were retained, contrary to the internal policy described by the regulator. Three conditions were nevertheless imposed: identify unusual or suspicious activity, monitor new reputational developments and review transactions unusual in size or structure. NYDFS, paragraphs 37 to 40 The system then broke in two places. The conditions were not communicated to every member of the relationship team. Nor did they properly reach the transaction-monitoring team. The NYDFS finally describes an inverted test: instead of comparing a transaction with Epstein’s specific risk profile, the team compared it with this client’s own historical activity. In March 2017, an alert involving payments to a Russian model and publicity agent was closed because that type of activity was deemed “normal for this client”. NYDFS, paragraphs 38 to 42 That sentence does not establish an illicit purpose for the payment. It establishes a weak benchmark. If atypical behaviour repeats for long enough, the client’s past becomes the standard of normality. Monitoring turns circular: what looks like the same client’s previous behaviour is treated as ordinary. The limited reach of account signals The Butterfly Trust illustrates the problem. Deutsche Bank opened checking and money-market accounts for it in January 2014. The NYDFS later counted more than 120 wires totalling $2.65 million to trust beneficiaries, with stated purposes including hotels, tuition and rent. The regulator also cited more than $7 million in apparent settlement payments through law firms and more than $6 million in apparent legal expenses. NYDFS, paragraphs 28 to 32 Those categories are not convictions. Rent, tuition, a civil settlement or legal fees can be legitimate. The order says something else: given the known profile, the payments required questions better tailored to the risk, stronger documentation and escalation where appropriate. The same safeguard applies to cash. An Epstein agent made 97 withdrawals of $7,500 at the Park Avenue branch between 2013 and 2017. The NYDFS also records $100,000 withdrawn at another nearby branch. In total, over roughly four years, the agent withdrew more than $800,000. The regulator says the bank properly filed mandatory currency transaction reports when the relevant thresholds were reached. It also found that the bank obtained no explanation beyond travel, tips and expenses, even after questions about reporting thresholds and an episode that appeared to involve structuring. NYDFS, paragraphs 48 to 52 A SAR is a signal to authorities, not proof that funds are criminal. Conversely, filing a currency transaction report correctly does not answer every question about customer knowledge or economic purpose. The obligations are complementary. An exit decided in December, completed after the arrest In November 2018, the Miami Herald investigation into the 2008 plea deal prompted another review. On 21 December, Deutsche Bank informed Epstein that it would no longer service his accounts. The NYDFS order treats that date as the decision to terminate the relationship. It also says that, despite the decision, a relationship manager prepared reference letters for two other financial institutions. NYDFS, paragraphs 53 and 54 DOJ records reviewed by Reuters in February 2026 provide a longer operational chronology. The exit letter gave Epstein until 28 February 2019 to transfer assets. Services continued beyond that deadline. Reuters identified at least nine accounts with combined balances of $1,776,680 on 3 May. After Epstein’s 6 July arrest, an urgent list still called for 28 accounts to be closed. Deutsche Bank told Reuters that asset transfers continued over the following months and reiterated its regret. Reuters, 11 February 2026 The most sensitive item is the 18 March reference letter. Reuters reported that it told the receiving institution that Deutsche Bank knew of no problem with the operation or use of the accounts. The NYDFS order confirms the existence of reference letters and reproduces the same language without dating each one. The letter does not prove an intention to conceal. Public documents do not identify who approved it or what information accompanied the asset transfer. NYDFS, paragraph 54 · Reuters, 11 February 2026 Three amounts that measure different things The public record places three eye-catching figures next to each other. Adding them or presenting them as three measures of the same harm would be wrong. | Amount | Exact category | What it does not measure | |---|---|---| | $150m | NYDFS penalty covering failures linked to Epstein, FBME Bank and Danske Bank Estonia | The order does not allocate an Epstein-only share | | $75m | Civil class settlement approved in 2023 | Neither a merits judgment nor an admission of liability | | Over $250m | Wires the Wyden report says were reported retrospectively as suspicious | Neither proven criminal funds, bank revenue nor victim losses | The $150 million penalty appears in the NYDFS order. It resolves failures across three high-risk relationships and includes oversight by an existing independent monitor. The document does not divide the amount among Epstein, FBME and Danske Estonia. NYDFS, paragraphs 7, 114 and 117 The $75 million figure is the civil class settlement with the plaintiffs. In May 2023, a federal judge had allowed some claims to survive a motion to dismiss. At that procedural stage, the court had to assume the well-pleaded factual allegations were true to test plausibility, without deciding the bank’s liability. In October, the court approved the $75 million settlement, which expressly excluded any admission of fault or liability. SDNY, 1 May 2023 decision · SDNY, 20 October 2023 final judgment Finally, the Senate Finance Committee staff report published on 4 August 2026 says Deutsche Bank failed to report in real time more than $250 million of suspicious wires and later filed retrospective reports. The report relies in part on confidential SARs reviewed at the Treasury. The public does not have the appendices needed to reconstruct and deduplicate every transaction. Reuters said it could not verify the report’s details. Deutsche Bank’s position cited by Reuters is that it cooperated with authorities, addressed deficiencies transparently and strengthened its controls. Reuters, 4 August 2026 The documents support only this conclusion: more than $250 million in wires were retrospectively reported as suspicious, according to the Wyden report. They do not support writing that the entire amount was criminal, financed particular offences or represented Deutsche Bank revenue. New York imposed a penalty, Germany has not publicly closed the subject The most detailed public action came from New York, which supervised the local branch and US trust company. The consent order binds the NYDFS and Deutsche Bank as to the settlement it describes. It does not bind US federal agencies, other states or criminal authorities. NYDFS, paragraphs 120 to 122 In Germany, a federal-government response published in July 2026 said that the government then had no information establishing criminally relevant links to Germany in the matters reviewed. It also said BaFin had made two information or document requests following concrete triggers and that no further supervisory action was considered necessary at that stage. The Bundestag summary does not identify the institutions concerned or the content of the responses, while the authorities said they continued to examine possible German links. That document does not support saying that BaFin cleared Deutsche Bank. It describes the public state of German work at a particular date. No new public penalty is not proof that confidential checks did not occur, just as an information request is not proof of an offence. Six questions remain open The public record stops short in several places. Those gaps prove neither additional wrongdoing nor concealment. They do, however, support six specific questions. 1. On what date was each account closed, and which services remained available until then? Deutsche Bank might be able to publish a redacted account-by-account chronology. The NYDFS and the independent monitor named in its order might be able to clarify whether this closure period was reviewed, subject to banking secrecy and their confidentiality duties. NYDFS, paragraphs 53, 54 and 117 · Reuters, 11 February 2026 2. Who authorised the services maintained after 28 February 2019, and under what rules? Deutsche Bank might be able to clarify the rules applied during that period. Further DOJ records or court documents could also provide evidence if they become public. Reuters, 11 February 2026 3. Who approved the reference letters, and what information was sent to the receiving banks? Deutsche Bank and the institutions that received the letters might be able to describe what was communicated without disclosing personal or protected information. The public NYDFS order does not name the receiving institutions. NYDFS, paragraph 54 4. How was the figure above $250 million calculated and deduplicated? Senate Finance Committee staff might be able to publish a redacted methodology. FinCEN, the US Treasury and Deutsche Bank might be able to clarify the number and timing of the reports, within the legal confidentiality limits governing SARs. Senate Finance Committee staff report, 4 August 2026 · Reuters, 4 August 2026 5. How much revenue did the relationship actually generate each year compared with the 2013 forecasts? Deutsche Bank is best placed to publish redacted, aggregate figures. The public documents reviewed quantify the initial commercial target, not the revenue ultimately realised. NYDFS, paragraphs 16 to 21 6. Did the two information requests cited by the German government concern Deutsche Bank? BaFin and the federal government might be able to clarify their scope. The response published in July 2026 does not name the institutions and therefore cannot settle the question. German federal government, response 21/7048 · Bundestag summary Naming these actors does not mean that all of them hold the answer, can lawfully publish it or are required to do so. Any silence would not confirm a hypothesis. The exception changes the definition of abnormal The Deutsche Bank record is not a story of compliance being entirely absent. It is a story of compliance appearing at almost every stage but being weakened by governance. The risk is written down before onboarding. The client is rated high-risk. The relationship is reviewed. Conditions are imposed. Alerts are generated. Currency reports are filed. An exit is decided. Every one of those actions exists in the record. The problem lies between them: approval without a formal meeting, a visit without a written record, a committee without compliant minutes, conditions not transmitted, monitoring that compares the client with himself, and a closure spread over months. The lesson is less comfortable than a simple blind spot. Controls can know the risk, label it and still learn to live with it. When a commercial exception becomes durable, it does more than evade the rule. It moves the internal boundary between ordinary and abnormal. Read the French version. Sources - New York State Department of Financial Services, Consent Order, 7 July 2020 - New York State Department of Financial Services, press release, 7 July 2020 - Deutsche Bank, public position on the 2013 onboarding, updated in 2026 - SDNY, Doe 1 v. Deutsche Bank, motion-to-dismiss decision, 1 May 2023 - SDNY, final approval of the class settlement, 20 October 2023 - US Senate Finance Committee, staff report, 4 August 2026 - US Department of Justice, Epstein disclosures portal - Reuters, investigation into Deutsche Bank’s prolonged exit, 11 February 2026 - Reuters, Wyden report and banks’ responses, 4 August 2026 - German federal government, response 21/7048, July 2026 Limitations - The NYDFS findings come from a consent order resolving an administrative action. They are not a criminal judgment and do not bind other authorities. - The DOJ records reviewed by Reuters may not be comprehensive. This article republishes no personal data, account number or detail capable of identifying a victim. - SARs are confidential. The figure above $250 million remains a conclusion attributed to the Wyden report, not a total that can be verified transaction by transaction from the public record. - The May 2023 court decision concerns the sufficiency of civil claims. The later settlement ended the litigation without a merits judgment or admission of liability. - The German response is current only to the public documents reviewed through 9 August 2026. No publication cannot establish that no confidential investigation or supervisory work exists. ============================================================================ ANALYSIS: Epstein’s money, 4/4: the final ledger URL: https://l0g.fr/en/analysis/epstein-money-final-ledger/ Canonical French source: https://l0g.fr/posts/argent-epstein-dernier-grand-livre/ Date: 2026-08-08 (reviewed 2026-08-09) Topics: Jeffrey Epstein, estate, 1953 Trust, survivors, U.S. Virgin Islands, IRS, Valar, investigation ---------------------------------------------------------------------------- At the end of June 2024, Jeffrey Epstein’s estate held just $2.19 million in cash. It had sold almost every property, paid more than $121 million through the compensation program, settled other survivor claims and completed most of a nine-figure settlement with the U.S. Virgin Islands. Published reporting then identifies an IRS refund of roughly $111.6 million. Three months later, total estate assets stood at $145.08 million. That refund rewrote the ending. An estate expected to run low on money was back above nine figures. The latest quarterly accounting shows $107,644,728.02 in gross assets at June 30, 2026. It also reports $22.5 million still payable under a survivor class settlement and $4.5 million in incurred but unpaid professional fees. Other claims, estate and inheritance taxes, and possible gift-tax liabilities remain unknown or to be determined. The story cannot be reduced to subtracting $107.6 million in 2026 from $577.7 million in 2019. The first number was a provisional inventory. The second combines cash, receivables and private companies, some still carried at August 2019 values. In between, the same dollars were sometimes sold, taxed, borrowed, repaid and counted again in a different form. This final installment follows the cash without mistaking it for the fortune. It comes after the documented origins and gaps in Epstein’s wealth, what his banks could see and the Highbridge, Valar, Apollo and ESW portfolio. The $577.7 million starting point The probate petition filed on August 15, 2019 lists $577,672,654 in personal and estate property. It includes $56.55 million in cash, $14.30 million in fixed income, $112.68 million in equities, $18.55 million in aircraft, vehicles and boats, and $194.99 million in hedge funds and private equity. Fine art remains marked “TBD,” with no amount in the total. The six properties add to $180,603,063 in this primary document: New York $55.93 million, Zorro Ranch $17.25 million, Palm Beach $12.38 million, Paris $8.67 million, Great St. James $22.50 million and Little St. James $63.87 million. The $117 million property figure repeated in some coverage roughly matches the real estate subtotal without Little St. James. It is not the complete petition total. The petition’s footnote matters. Values remained subject to appraisal and date-of-death updates. This was a photograph taken before final tax work, physical inventory and the liquidation of private assets. It established a starting point, not an exit price. The houses became cash, then the cash moved on The property liquidation invites a tidy story: six assets, six sales, one total. The public record is messier. Palm Beach sold for $18.5 million in 2021. The East 71st Street townhouse brought $51 million. The Paris residence sold on June 29, 2022 for roughly $10.4 million at the exchange rate used in contemporaneous coverage. Little and Great St. James sold together for about $60 million in May 2023. Zorro Ranch found a buyer in August 2023, but New Mexico’s disclosure rules left the price private. The subtotal of known public prices is therefore $139.9 million, plus an unknown amount for Zorro. Forbes reconstructed the sales and prices. The eleventh quarterly accounting confirms the Paris sale and says the net proceeds were still held by the French notary when that report was filed. Public sale prices are gross. Commissions, closing costs, taxes, creditors and survivor-related transfers sit between the price and the estate’s retained cash. The 2022 settlement also sent half of Little St. James’s net proceeds to the Virgin Islands. Complete closing statements are not public for all six assets, so a single figure for what the estate “kept” would mix published amounts with assumptions. The first line of claimants: survivors The Epstein Victims’ Compensation Program operated in 2020 and 2021 under independent administrator Jordana Feldman. The estate’s seventeenth accounting reports $121,127,339.05 as “claims and other amounts paid pursuant to the EVCP.” That wording matters. The filing does not label every dollar as net compensation received by a claimant; it combines claims and other program amounts. Settlements outside the EVCP form a second line. The same seventeenth accounting gives a cumulative $34.23 million at December 31, 2023. Later filings add $75,000 in Q2 2024, $8,344,610.90 in Q3, $1.275 million in Q4, $1.675 million in Q2 2025 and $3 million in Q4. The reconstructed total at year-end 2025 is $48,599,610.90. Reuters’ reported $49 million rounded figure is consistent with that ledger. A third line opened in 2026 in Ward v. Indyke. The February 19 settlement agreement provides $35 million if at least forty people are found eligible, and $25 million if there are fewer than forty. The estate must place $12.5 million in a qualified settlement account after preliminary approval, then pay the balance after final approval and the eligibility count. Indyke and Kahn deny the allegations, and the agreement contains no admission of fault or liability. The estate accounts fit the $35 million branch exactly: $10 million transferred in Q1, $2.5 million in Q2 and $22.5 million still payable at June 30. The first $12.5 million is a transfer into the Qualified Settlement Account, not proof that it has already been distributed to class members. The money remains under court jurisdiction. The settlement hearing is scheduled for September 16, 2026. The public notice adds a possible deduction. Class counsel will ask for up to 30% of the global amount, plus expenses of up to $1 million, subject to the judge’s decision. The amount available for allocations will be lower than the global settlement if the court awards some or all of those requests. The Virgin Islands published settlement total: $117.28 million The settlement announced by the U.S. Virgin Islands Department of Justice on November 30, 2022 required $105 million in cash, $450,000 for environmental remediation around Great St. James and half the proceeds from Little St. James. The release also says the deal returns more than $80 million in economic-development tax benefits to the territory. Those tax benefits are included in the $105 million cash payment. Adding them would create a false $185 million settlement before the island proceeds. The seventeenth accounting closes the sequence at $117,282,494.01 paid by December 31, 2023. Subtracting the $105 million, the $450,000 remediation amount and $456,245.01 of final interest leaves $11,376,249 attributable to the net Little St. James share. This figure is derived by subtraction; it is not separately labeled that way on the cover sheet. The estate borrowed $30 million at 9% to finance part of the settlement. The twentieth accounting records its August 12, 2024 repayment: $31,768,611.14, principal and accrued interest. That is financing debt, not another survivor or government settlement. The check that brought the estate back The same twentieth accounting captures the summer 2024 reversal. Total assets rose from $84.44 million on June 30 to $145.08 million on September 30. “Income collected and other increases” reached $112,015,145.16. The public form does not break that line into a precise tax refund. Reporting by The New York Times, repeated with the figures by the New York Post and Forbes, identifies $111.6 million as an IRS refund. The reports trace it to roughly $190 million in estate taxes paid in July 2020, before several assets, particularly real estate, sold below expected values. The court filing reports $112,015,145.16 in quarterly increases; published reporting identifies $111.6 million of that increase as the federal refund. The underlying Form 706, IRS notice and detailed refund computation are not among the public records reviewed for this story. Economically, the refund looks less like new income than the late return of an excessive advance. The estate had paid tax on provisional valuations before knowing the assets’ realized prices. Cash reappeared in 2024, but the transaction began in 2020. The last balance sheet still contains the past The twenty-seventh accounting, dated July 30 and docketed July 31, 2026, is the latest snapshot, at June 30: - cash: $25,704,907.99; - jewelry and watches: $4,055; - loans receivable: $3,432,264; - entities owned by Epstein: $78,503,501.04; - reported total: $107,644,728.02. The four components add to $107,644,728.03. The filed total is one cent lower. The discrepancy has no economic significance, but it is methodologically useful: the primary filing is reproduced with its anomaly instead of silently repaired. The material valuation issue lies elsewhere. The $78.50 million entity line still relies on date-of-death appraisals, according to the filing’s notes. In March 2026, co-executor Darren Indyke told the House Oversight Committee that the two Valar interests might be worth roughly $172 million, while stressing that the estimate was unrealized and the fund duration uncertain. That estimate does not sit on top of the balance sheet. Valar is already inside the estate-owned entity line. The relevant question is how much an old carrying value should be adjusted, not whether to calculate $107.6 million plus $172 million. A current audited LP statement or realized distribution would be needed to answer it. The 1953 Trust waiting room Epstein signed his will and amended the 1953 Trust on August 8, 2019, two days before his death. The will filed in the Virgin Islands pours the residue into the trust. It also grants Darren Indyke and Richard Kahn $250,000 each for serving as executors, payable when probate is complete, with no other executor compensation beyond reasonable reimbursed expenses. The trust copy published by the U.S. Department of Justice, EFTA01266204, operates at a different scale. Debts, administration expenses and taxes come first. Then three priority bequests follow: 1. $50 million in cash to Karyna Shuliak, plus the purchase of a $50 million annuity; 2. $50 million to Darren Indyke; 3. $25 million to Richard Kahn. These three ranks total $175 million in nominal bequests, before items 4 through 41. The trust directs the remaining bequests to be paid in order and to lapse where assets are insufficient. It also establishes nominal reserves, including $50 million for claims against the assets and $10 million for challenges to the will or trust. Shuliak was also to receive the six properties, household items and diamonds. All six properties were sold before final trust funding. The public instrument does not settle what substitute property, if any, would be due after those sales. That question belongs to probate administration and governing law, not a mechanical conversion of sale proceeds into a bequest. The $250,000 executor fees and the $50 million and $25 million trust bequests to Indyke and Kahn are distinct. The former compensates a role after probate is complete. The latter depends on the trust’s priority rules and whatever survives the claims process. No accounting reviewed for this article documents a distribution to trust beneficiaries. So where did the money go? Three major estate-funded outflows are documented: $121.127 million under the EVCP label, $48.600 million in settlements outside the EVCP through year-end 2025, and $117.282 million paid to the Virgin Islands. The estate then transferred $12.5 million into the 2026 class settlement account, with $22.5 million still payable. Even these categories are not perfectly homogeneous: the EVCP line includes “claims and other amounts,” and the class deposit still awaits distribution. The JPMorgan and Deutsche Bank settlements belong to a different ledger. Those banks funded their own agreements. Their money did not leave the estate. Combining them with estate outflows would measure the wider settlement landscape around Epstein, not the path of his $577.7 million inventory. The houses generated cash, but complete net proceeds remain unavailable. Reporting says roughly $190 million went to federal estate tax in 2020 and $111.6 million came back in 2024. A $30 million bridge loan funded part of the territory settlement and later cost $31.769 million to repay. Valar remains locked in private funds whose eventual realization may differ sharply from its old carrying value. The final ledger therefore ends with a precise, provisional answer. At June 30, 2026, the estate reported $107.645 million in gross assets. At least $27 million of liabilities were quantified. Other claims and taxes remained open. Most of the carrying value sat in private entities. The beneficiaries of the 1953 Trust were still behind that line. Epstein’s fortune did not fall through one hidden trapdoor. It fragmented among survivors, the territory, tax authorities, creditors, professionals and illiquid holdings. Any future inheritance is the residual variable in that equation, not its starting point. Conclusions supported by the accounts The amounts cited here are distinguished by their basis: primary document, reported figure, derived calculation or unknown. Quarterly accountings 17 through 27 are read alongside the probate inventory, the USVI settlement, the 2026 class agreement, the will and the 1953 Trust. Reported property prices never become net proceeds without closing records. Valar estimates are never added to a balance that already includes them. The documents trace a coherent path: the 2019 inventory, reported sales, the Virgin Islands settlement, settlements outside the EVCP, the class settlement and the first three priority bequests fit together. They do not, however, supply a ready-made closing balance. The June 30, 2026 balance sheet even retains a one-cent difference between its components and its total. That cent does not change the estate’s economics; it is a reminder to read the accounts as published rather than silently repair them. The Form 706 and IRS notice, property closing statements, the Zorro Ranch price, current Valar statements, cumulative professional fees, final taxes, other claims and any trust distributions are not public in the records reviewed. Read the French version. Main sources - Probate petition, August 15, 2019, Superior Court of the Virgin Islands, ST-2019-PB-00080. - Twenty-seventh quarterly accounting, through June 30, 2026, dated July 30 and docketed July 31, 2026. - Twentieth quarterly accounting, through September 30, 2024, loan repayment and $112.015 million increase line. - Full seventeenth quarterly accounting, through December 31, 2023, EVCP and USVI figures. - USVI DOJ, settlement with the estate and co-defendants, November 30, 2022. - Ward v. Indyke settlement documents and official schedule, accessed August 7, 2026. - Jeffrey Epstein’s will and 1953 Trust, EFTA01266204. - U.S. House Oversight Committee, Richard Kahn and Darren Indyke depositions, March 2026. - Reuters, class settlement of up to $35 million, February 20, 2026. - Forbes, reconstruction of the property sales, July 22, 2025. - New York Post, reported amount and origin of the tax refund, January 15, 2025. Limitations Quarterly cover sheets publish asset stocks and selected categories of movement; they do not by themselves provide a consolidated statement of every cash flow since 2019. The $111.6 million IRS refund and roughly $190 million paid in 2020 are reported figures; the primary Q3 2024 increase line is broader at $112,015,145.16. Property prices are gross and Zorro Ranch remains undisclosed. The $172 million Valar figure is an oral estimate, not an audited statement or realized distribution. Tax liabilities and some claims remain to be determined. Information reviewed through August 7, 2026. ============================================================================ ANALYSIS: Epstein’s money, 3/4: client, broker or investor? URL: https://l0g.fr/en/analysis/epstein-money-client-broker-investor/ Canonical French source: https://l0g.fr/posts/argent-epstein-client-courtier-investisseur/ Date: 2026-08-08 (reviewed 2026-08-09) Topics: Jeffrey Epstein, Highbridge, Valar Ventures, Apollo, Leon Black, ESW, private equity, investigation ---------------------------------------------------------------------------- On December 29, 2004, Financial Trust Company received $15 million for “merger and acquisition advice” connected to the sale of Highbridge to JPMorgan. Ten years later, two Jeffrey Epstein-related entities held $58.4 million at Highbridge. Within months, they withdrew $59.45 million. Three figures, three transactions, and one trap: adding them as though they measured the same thing. The first number is income. The second is a balance at a point in time. The third is gross redemption proceeds. Some of the capital had been contributed before the 2004 fee; intervening withdrawals and transfers remain incomplete. The $15 million, $58.4 million and $59.45 million therefore do not produce a $132.8 million fortune or a calculable return. The same confusion runs through Epstein’s wealth. The documents variously cast him as paid adviser, possible finder, fund client, public shareholder and member of private companies. Around Apollo, he owned shares in the publicly traded manager, but no identified interest in an Apollo-managed fund. He invested in two LLCs formed by Apollo executives, but legally separate from Apollo’s funds. At Valar, $40 million first describes commitments; the same interests were estimated at $90.44 million at the end of 2018 and roughly $172 million in 2026 testimony. Those are not three layers of wealth. They are successive states of the same investments. After examining the documentary origins of Epstein’s wealth and what the banks could see, this third installment follows the capital itself. It relies on a deduplicated ledger: every line carries a date, legal entity, economic category, scope and source. Unsigned drafts are excluded from realized income. Whole-fund financials never replace an individual limited partner’s account. Unrealized estimates remain estimates. Four separate ledgers A financial number only makes sense with its accounting unit. We separated four ledgers: - income: fees and commissions received; - capital: subscriptions, capital calls and purchases of securities; - ownership: share count, percentage and legal vehicle; - value or liquidity: capital account, market price, distribution or redemption. The same asset can move from one ledger to another over time. A $10 million contribution may become an interest valued at $30 million and later generate a $35 million redemption. Adding 10 + 30 + 35 would manufacture $75 million from one investment. Epstein’s files invite exactly that mistake: bank statements, internal reports, tax forms, emails and estimates are separated by years yet appear side by side in the public corpus. Highbridge: the invoice that exists The strongest starting point is a bank record. In Government of the U.S. Virgin Islands v. JPMorgan Chase, case 1:22-cv-10904-JSR, Exhibit 246 attached to Document 285-88 contains a wire screen marked “processed.” Date: December 29, 2004. Amount: $15,000,000. Destination: Bear Stearns, for further credit to Financial Trust Company. The next page is a December 28 invoice to Highbridge’s Ron Resnick for merger and acquisition advice. Its DOJ corpus identifier is EFTA02816421. Court filings identify the payer as Dubin & Swieca Holdings and connect the fee to JPMorgan’s acquisition of Highbridge. Another filing states that Epstein received no fee from JPMorgan itself. The precise formulation is therefore: Financial Trust received $15 million in connection with the Highbridge transaction, not “JPMorgan paid Epstein $15 million.” The payment does not establish what Financial Trust actually delivered. The invoice contains no hours, team or work product. It proves the amount and stated purpose, not the value of the service. A $2.25 million agreement that remains a draft After the sale, Highbridge considered a longer arrangement. A February 15, 2005 internal email, EFTA02811823, sketches a five-year deal: $100,000 a year in direct fees plus employee pricing for certain investments. At $7 billion in Highbridge assets, the author estimated the present value of that discount at about $8 million. The same email says Epstein then had about $35 million directly with Highbridge and $200 million to $300 million mostly with D.B. Zwirn, which the author believed was largely client money. That sentence does not identify the beneficial owners. It does show that Highbridge viewed the potential relationship as combining Epstein’s capital, third-party capital and commercial access. The revised agreement raises the direct fees to $2.25 million, in five $450,000 installments from June 1, 2005 through December 31, 2009. “Qualified Funds” could invest up to 0.85% of Highbridge Capital Corp’s assets at employee pricing. But the released signature page is blank. We found no wire corresponding to the five installments. The $2.25 million therefore belongs in the proposed-contract ledger, not realized income. Including it would turn an offer into cash. The Highbridge capital came before the fee A June 9, 2014 email from Richard Kahn to Epstein provides the best bridge between old contributions and later balances. It says: - Haze Trust invested $10,041,666 on April 20, 1999 and had withdrawn nothing by June 2014; - Financial Trust invested $25,044,521 on January 11, 2001 and withdrew $25 million on February 28, 2006; - at May 31, 2014, Southern Financial held $20,472,425 and Haze Trust $37,903,950, totaling $58,376,375. Both documented contributions predate the 2004 invoice. That forecloses a simple account in which the $15 million fee created the entire Highbridge portfolio. Some fee proceeds may have been reinvested, but the available documents do not isolate them. A separate September 30, 2006 statement, EFTA01592294, also identifies a “Highbridge · Financial Trust Co.” managed account with a $12.21 million functional portfolio value. It cannot automatically be combined with the Haze and Southern fund interests. The manager’s name is the same; the product and scope may not be. The redemption notices document the exit. At October 31, 2014, Haze received $38,485,360 and Southern Financial $20,785,189. Final payments in February 2015 added $113,297 and $62,943. Total: $59,446,789 in gross proceeds. That amount is not profit. The 2006 withdrawal, possible additional subscriptions, entity transfers, earlier distributions and fees make a full return calculation impossible. Valar: $40 million is not a valuation At Valar, private-fund terminology changes what the numbers mean. Southern Trust Company subscribed to two commitments: $15 million in Valar Global Fund II, L.P. and $25 million in Valar Global Fund III, L.P. The $40 million is capital Southern promised to supply if called. It says neither how much had been funded nor what the interests were worth. The June 30, 2018 limited-partner statements separate those measures. For Fund II, $14.55 million was paid in, $450,000 remained, and ending partner capital was $21,000,471. For Fund III, $21 million was paid in, $4 million remained, and ending partner capital was $33,345,270. At that date: - commitments: $40 million; - capital actually paid in: $35.55 million; - combined LP capital-account value: $54,345,741. Those figures describe the same commitments at different stages. They do not add to $129.9 million. The Valar wire missing from the chart The Fund III reconstruction illustrates the difference between a lawyers’ list and a ledger. EFTA00080250 is an August 13, 2019 legal email with charts of selected transactions. Its authors refer to transactions they “highlighted”; the chart is not described as exhaustive. Its Fund III wires total $22.5 million, $2.5 million short of the commitment. Bank wire report EFTA01299550 supplies the missing item: on April 4, 2017, Southern Trust wired $2.5 million to Valar Global Fund III, L.P. Restoring that wire makes the sequence match the statements exactly: $21 million funded by June 30, 2018; another $2.5 million in January 2019; and $1.5 million in April. Total: $25 million. This is not a new cash flow layered onto the corpus. It is an omitted cash flow that closes the gap. Fund II remains less granular. The records track $6.3 million in 2015, followed by a sixth call of $2.25 million, bringing the balance to $8.55 million in April 2016. The individual amounts for calls seven and eight are not available in the OCR reviewed; their implied aggregate is $3.6 million. After $1.5 million in February 2017, $600,000 in October and $300,000 in the first half of 2018, the LP statement reaches $14.55 million. The ledger preserves the aggregate rather than inventing two precise amounts. From $54 million to $90 million to $172 million Three months after the June statements, the communicated values were nearly unchanged: $21,166,482 for Fund II and $33,165,591 for Fund III, or $54,332,073 at September 30, 2018. Then the step-up became dramatic. February 2019 emails, including EFTA01029015, conveyed unaudited December 31, 2018 estimates: $40.75 million for Fund II and $49.69 million for Fund III. Total: $90.44 million. The manager had already communicated high gross performance figures: 3.7x and a 44.6% IRR for Fund II; 2.6x and a 63.6% IRR for Fund III. Those indicators are not cash distributions. They depend on underlying company valuations and on the use of a gross metric. One discrepancy remains open. The Fund II LP statement records $14.55 million paid in, while later internal reports use a $14.25 million cost basis. The $300,000 gap may reflect an accounting definition or an update lag; no identified document resolves it. Both bases must remain visible. The final number comes from testimony, not a statement. In a March 19, 2026 deposition released by the House Oversight Committee, co-executor Darren Indyke gave a rough value of about $170 million for the two Valar interests, then specified $172 million. He also cautioned that those values were not meaningful until realization and that fund terms could be extended. Indyke separately put the estate’s current accounting at just over $100 million and said it already included a value for Valar. The $172 million therefore cannot be added to the $100 million. The precise overlap is not public. Apollo: documented ownership at company level “Apollo” covers at least four distinct scopes. The first is Apollo Global Management, LLC, the company listed in 2011. The second consists of private funds managed by Apollo. The third includes LLCs formed by individual executives for opportunities the funds declined. The fourth is made up of outside companies in which those executives, Epstein or Leon Black may have invested. The Dechert report Apollo filed with the SEC in January 2021 says it found no evidence that Apollo retained Epstein or that Epstein invested in an Apollo-managed fund. That conclusion came from an investigation commissioned by Apollo’s conflicts committee, not from a regulator. It is nonetheless consistent with the financial records reviewed here. Financial Trust did purchase publicly traded Apollo shares in the IPO directed share program. The prospectus set the price at $19. Internal report EFTA00811897 shows 263,157 shares at a cost of $4,999,983, exactly 263,157 × $19. At September 30, 2018, the position was worth $9,092,074, for an unrealized gain of $4,092,091. Dechert, however, writes 263,257 shares. Bank statement EFTA01510763 provides a second control: on November 30, 2012, Financial Trust received $105,262.80 at $0.40 a share. The quotient is 263,157. Dechert’s figure is likely a 100-share typo, submitted to Apollo for confirmation. According to Dechert, the shares were transferred to Southern Financial in 2013 and remained held through at least September 2019. They were an interest in the listed manager. They conveyed no direct economic rights in Apollo’s fund portfolios. AP SHL and AP Technology: vehicles that blur the map The records also identify AP SHL Investors LLC and AP Technology Partners LLC. Draft agreements collected in EFTA00586106 assign Financial Trust a 40% interest in AP SHL and 5.834262% in AP Technology. John Hannan is named as AP SHL’s managing member; Andrew Africk, Hannan, Mark Rowan and Michael Weiner as AP Technology managers. The Dechert report says Apollo executives formed the vehicles for opportunities declined by Apollo funds. The “AP” prefix, executive involvement and proximity to Apollo deal flow create ambiguity. Legally and financially, that does not turn the LLCs into Apollo funds. An AP SHL 2012 K-1, EFTA00593329, shows ending capital of $40,864 after a $13,867 current loss. A tax capital figure is not necessarily fair market value. Again, the category controls the meaning. ESW: the payment is certain, the final contract is missing The ESW case, often called ESWW in the records, shows what an EDGAR reconstruction can establish and what it leaves open. Financial Trust’s Schedule 13G filed July 20, 2011 reports 13,350,205 shares, or 6.1% of the company. Financial Trust had sole voting and dispositive power. Inventory EFTA00299927 lists a $1 million 9% convertible funded in March 2010 and several certificates composing the block. In November 2012, drafts set out a package sale of ESW stock and the AP SHL and AP Technology interests. The November 2 draft still prices the package at $18 million, with an indicative allocation among the three assets. November 26 drafts name Black Family Partners, L.P. as buyer and cut the aggregate price to $5.5 million. But the released signature pages are blank; one page even bears the impossible date November 31, 2012. Two draft transfer instructions cover 13,198,711 and 151,494 ESW shares. Together they equal the precise 13,350,205 shares in the 13G. Yet transfer and signature fields remain incomplete. The bank statement then settles one central point. On November 30, 2012, Financial Trust received a $5,500,030 CHIPS credit from Black Family Partners LP. That payment is consistent with the latest draft’s $5.5 million price. The extra $30 is unexplained; the record does not support calling it a fee. We can therefore report that Black Family Partners paid $5.50003 million to Financial Trust and that consistent drafts describe an ESW/AP SHL/AP Technology package. We cannot treat the $18 million draft allocation as final or claim to have located the executed agreement. EDGAR leaves two gaps open ESW’s 2012 10-K, filed in March 2013, still identifies Financial Trust as holding 13,350,205 shares, or 5.85%. That post-payment disclosure may rely on an earlier record date, an unupdated register or an incomplete transfer. Without the shareholder register or signed agreement, selecting one explanation would be speculation. On May 24, 2013, ESW executed a one-for-2,000 reverse split. Financial Trust’s block mechanically yields 6,675 whole shares plus a fractional remainder. A February 2014 S-1/A attributes 6,671 direct shares to Black Family Partners, plus 4,706 shares received as interest payments. The proximity of 6,671 to the theoretical 6,675 corroborates the economic transfer; the four-share gap remains unexplained. In 2015, the 10-K assigns 14,389 shares, or 10.63%, to Black Family Partners. ESW’s April 1, 2015 Form 15 then terminated public reporting. The documentary trail ends. The picture after the third ledger Highbridge is the only set examined here in which a large advisory payment is established by an exact bank transaction: $15 million. But the Highbridge investments began before that payment. The story is not a single commission turned into a portfolio. It is a relationship in which fees, Epstein’s own capital, money attributed to clients and contemplated preferential terms could coexist. Valar marks a different phase. Southern Trust no longer appears to provide a service; it behaves as a private-fund investor, subject to capital calls, long holding periods and illiquid valuations. The apparent rise from $35.55 million funded to roughly $172 million estimated would be substantial if realized. The public record does not show that realization. Apollo and ESW finally demonstrate why names make poor accounting boundaries. “Apollo” can mean a listed share, a fund, an LLC formed by executives or simply a network of people. “Black” can mean Leon Black as a personal client, Black Family Partners as an asset buyer or family members involved with ESW. A financial investigation must follow the legal entity before following the reputation of the name. Nothing in these transactions, standing alone, proves criminal origin of the invested funds. Nor can an appraisal be converted into available liquidity. What the records establish is narrower: Epstein had durable access to sophisticated managers, invested through multiple entities, obtained or contemplated special terms, and left behind private assets whose value remains partly locked inside fund structures. The fourth and final installment will follow those assets after his death: the estate inventory, victim compensation, taxes, legal fees, asset sales and residual value. The same rule will apply: a Valar estimate already carried in the estate’s accounts cannot be added a second time. Method and unresolved discrepancies This investigation cross-checks DOJ releases, exhibits in the U.S. Virgin Islands litigation against JPMorgan, SEC EDGAR filings and testimony released by the House. EFTA identifiers are preserved even when a direct DOJ PDF URL is intermittent. Lawyers’ transaction charts are treated as selected lists unless shown to be exhaustive; unsigned drafts do not establish execution. The control ledger retains four unresolved discrepancies: 263,157 Apollo shares in the statements versus 263,257 in Dechert; $14.55 million Fund II paid-in capital versus a $14.25 million internal cost basis; 6,675 theoretical ESW shares after the reverse split versus 6,671 disclosed; and ESW still listing Financial Trust in March 2013 despite the November 2012 payment. Read the French original. Sources - USVI v. JPMorgan: public docket, No. 1:22-cv-10904-JSR, including Document 285-88, Exhibit 246, EFTA02816421. - DOJ Dataset 9: EFTA00589969, draft Highbridge consulting agreement. - DOJ Dataset 9: EFTA00640876, 2014 Highbridge balances and contribution history. - DOJ Dataset 9: EFTA01118143, principal Highbridge redemptions. - DOJ Dataset 9: EFTA00590651, final Highbridge redemption payments. - DOJ Dataset 9: EFTA00811791, Valar Fund II LP statement. - DOJ Dataset 9: EFTA00811797, Valar Fund III LP statement. - DOJ Dataset 10: EFTA01299550, April 4, 2017 Fund III wire. - DOJ Dataset 9: EFTA01029015, December 2018 Valar estimates. - House Oversight Committee: Richard Kahn and Darren Indyke deposition videos, March 2026. - SEC: Apollo prospectus, 2011 and Dechert report, 2021. - DOJ Dataset 10: EFTA01510763, Apollo distribution and Black Family Partners credit. - SEC: Financial Trust Schedule 13G for ESW, ESW 2012 Form 10-K, 2014 Form S-1/A and 2015 Form 15. Limitations The Highbridge $15 million payment is established by a bank record and invoice filed as a court exhibit; the public copy does not supply the underlying work product. Dechert’s report is a company-commissioned investigation, not a regulatory finding. EFTA emails and internal reports are contemporaneous records but may use accounting definitions that are not stated. The 2018 Valar year-end figures are expressly unaudited, and the 2026 figure is oral testimony about an unrealized value. ESW’s released purchase agreements and transfer instructions are drafts; the bank statement establishes payment but not the final price allocation or every transfer mechanic. No figure in this article is presented as criminal proceeds. Values at different dates and fund-level figures are not added. Corpus current to August 7, 2026. ============================================================================ ANALYSIS: Epstein’s money, 2/4: the banks’ record URL: https://l0g.fr/en/analysis/epstein-money-what-the-banks-saw/ Canonical French source: https://l0g.fr/posts/argent-epstein-ce-que-les-banques-voyaient/ Date: 2026-08-07 (reviewed 2026-08-09) Topics: Jeffrey Epstein, banks, JPMorgan, Deutsche Bank, BNY Mellon, Bank of America, money laundering, investigation ---------------------------------------------------------------------------- In March 2017, Deutsche Bank’s monitoring system generated an alert on a payment from Jeffrey Epstein to a Russian model and publicity agent. The alert was closed. Not because the bank had established the economic purpose of the payment, but because the activity was deemed “normal for this client”. Four words capture much of the banking record: repetition had normalised the risk. Epstein’s banks were not operating in an information vacuum. JPMorgan filed its first suspicious activity report in 2002. Deutsche Bank classified him as high-risk when it onboarded him in 2013. Both institutions examined cash withdrawals, unusual counterparties and his reputation. Both also made explicit decisions to retain, restrict or terminate the relationship. What the public record does not support matters just as much. An alert is not proof of an offence. A Suspicious Activity Report, or SAR, is neither an indictment nor a judgment. And a bank’s knowledge of a client’s conviction does not prove that each of its executives knew about each crime committed by that client. The record establishes something else: signals existed, but their consequences remained limited for years; the largest reports were not filed until 2019, after Epstein’s new arrest. As the first instalment showed, Epstein’s fortune cannot be reduced to a single number. Neither can his banking infrastructure. The JPMorgan, Deutsche Bank, BNY Mellon and Bank of America figures describe overlapping scopes. The same wire can be seen by the payer’s bank, the beneficiary’s bank and a correspondent bank. Adding them together produces a scandalous-looking number, not an accounting. The exact scope of a SAR The Bank Secrecy Act requires US banks to report certain transactions when they know, suspect or have reason to suspect that the activity may involve illegal funds, seek to evade reporting requirements or have no apparent business or lawful purpose. According to FinCEN, the clock starts when the institution initially detects facts that may form the basis for a filing: thirty days in principle, up to sixty when no suspect has yet been identified. That detail rules out an easy shortcut. Comparing a transaction date with a SAR filing date does not, by itself, establish a late filing. The relevant date is when the bank concluded that the facts reached the reporting threshold. That chronology of detection, escalation and decision-making is almost never fully public. SARs are also confidential. The available numbers come mainly from bank documents produced in litigation, regulatory orders and the investigation led by Senator Ron Wyden’s staff, who reviewed part of the Treasury file without being able to publish the reports in full. These are institutional and sometimes highly granular sources. They do not replace the missing transaction appendices. JPMorgan: a first alert in 2002 The Senate Finance Committee staff report published on 4 August 2026 dates the main Epstein · JPMorgan relationship to 1998. An earlier memorandum focused on the bank breaks down nine SARs that have become partly public. The first three were filed before the Palm Beach investigation: - 18 April 2002: $194,300 in reported activity; - 16 December 2002: $1,925,000; - 15 April 2003: $166,600. Four more reports were filed between 2008 and 2016. Together, the seven SARs preceding the two giant 2019 filings covered $4,316,424. JPMorgan can therefore say, accurately, that it began reporting transactions in 2002. But that early start raises the central question: how did a relationship already generating alerts survive until 2013, five years after Epstein’s Florida conviction? The documents supply part of the answer. A 2003 internal review described the Epstein accounts as one of the private bank’s largest annual revenue flows. By September 2009, balances linked to the relationship stood at no less than $142 million. In 2012, he ranked among the private bank’s top twenty clients by revenue, eighth according to a record cited by investigators. Senate staff estimate that JPMorgan collected more than $8.1 million in fees from 2009 to 2014. Revenue does not prove that a compliance decision was bought. It does establish that the risk was being weighed inside a commercially important relationship. Seven million dollars in cash, year after year An expert report submitted by the government of the US Virgin Islands identified 134 JPMorgan accounts connected to Epstein, his entities and his associates. That number comes from a party-appointed expert, not a judgment, but its annual cash table can be reconciled with internal records cited by Senate investigators. From 2002 to 2013, the listed withdrawals total exactly $7,159,475. Cash was withdrawn in every year. The peak came in 2002 at $2,119,300, followed by $840,000 in 2004, $904,335 in 2005 and $938,264 in 2006, the year Epstein was arrested in Florida. The series does not reveal who ultimately received each banknote. It documents a durable pattern that the bank observed in fragments. In March 2012, banker John Duffy exchanged messages with risk personnel about $160,000 in withdrawals. He said that he had asked Epstein to use an aviation account rather than personal accounts, Epstein’s explanation being that the cash paid for fuel. Compliance would later write that paying cash for aircraft fuel abroad was not normal. It would go too far to conclude from this that Duffy taught Epstein how to avoid a reporting requirement. The record establishes that withdrawals were shifted from one category of accounts to another, and that investigators identified no contemporaneous SAR. It does not document the legal intent required to establish evasion. In July 2013, compliance discovered roughly $800,000 in prior withdrawals that had not been escalated to it. An employee asked why the business had not reported them. The question exposes an organisational divide: the bank held the data, but the units seeing it did not necessarily assign it the same meaning. In 2011, a decision to retain Epstein The internal timeline becomes clearer from 2011. A due-diligence review recorded that Jes Staley had consulted Stephen Cutler, then the bank’s general counsel, and that a decision had been made to retain Epstein. In 2013, another review said that Mary Erdoes and John Duffy were aware of the relationship, which was classified as sensitive and subject to annual review. JPMorgan has since isolated Staley’s conduct. In a two-page response to Senator Wyden, the bank said that its other executives had acted with integrity and that no material produced in discovery established that Jamie Dimon knew of the relationship before 2019. That response belongs in the account. Emails bearing references such as “pending Dimon review” underpin investigators’ questions, but do not by themselves prove that Dimon reviewed or approved any particular decision. The direct relationship was closed in 2013 for risks connected, according to the Senate report, to money laundering and human trafficking. Yet on 14 August 2013, after the exit decision, Duffy asked Erdoes whether the bank could continue working with Epstein through third-party client accounts, including Leon Black’s. She approved. That point connects the first two instalments of this investigation. The bank no longer retained Epstein as a direct client, but still authorised activity in which he acted as another client’s adviser. Closing a legal relationship did not necessarily remove every economic transaction that depended on it. The billion that must not become $1.28 billion After the July 2019 arrest, JPMorgan filed two reports on an entirely different scale. On 13 August, an initial SAR covered 469 wires representing $200,979,535, over a period running from 1 October 2003 to 29 May 2019. On 26 September, a much broader filing covered 4,725 wires representing $1,081,819,653, from 1 October 2003 to 22 July 2019. Their arithmetic sum is exactly $1,282,799,188. The Senate report’s table presents that addition. But the two periods begin on the same day and overlap almost entirely. Reuters described the September filing as an amended report that added 44 subjects and expanded the initial review. Without the transaction lists, two boundaries are possible: - lower bound: $1,081,819,653, if the first set is wholly included in the second; - gross upper bound: $1,282,799,188, if the two sets are entirely separate. The second assumption is difficult to reconcile with the nested dates and the description of an amended filing, but it cannot formally be eliminated without the appendices. The strongest wording is therefore: the expanded September 2019 SAR covered 4,725 wires worth $1.0818 billion; an earlier $200.98 million filing may overlap with it. Even then, the billion is a cumulative face-value volume. It is not Epstein’s net worth, his income or proven criminal proceeds. A $10 million round trip counts as $20 million of transaction volume. Deutsche Bank: onboarding a known risk The handover came with almost no gap. A relationship manager who had moved from JPMorgan introduced Epstein to Deutsche Bank. The commercial memorandum projected $100 million to $300 million of flows and $2 million to $4 million in annual revenue. The onboarding file disclosed his conviction, registered-sex-offender status and seventeen civil settlements. On 19 August 2013, Deutsche opened the first Southern Trust Company and Southern Financial accounts. More than forty accounts would ultimately be linked to Epstein, his entities and trusts. The client was classified as high-risk and treated as an “honorary PEP”, requiring enhanced scrutiny. These facts come from the New York Department of Financial Services consent order, the strongest regulatory source in the banking record. It is a set of findings accepted by Deutsche Bank, not a civil litigant’s pleading. The problem was not the absence of procedure. It was execution. In January 2015, the reputational-risk committee agreed to continue the relationship after meeting Epstein. The order says that no minutes were kept, contrary to policy. Conditions were imposed, including enhanced monitoring of the accounts and transactions. Yet those conditions were communicated neither to the relationship manager nor to the transaction-monitoring team. The committee could be, in the document’s words, “comfortable with things continuing”. The computer system did not know what that continuation was supposed to prohibit. When abnormal activity becomes the client profile The NYDFS order documents more than 120 wires, totalling $2.65 million, to Butterfly Trust beneficiaries. Their apparent purposes included hotels, rent and tuition. At least eighteen payments of $10,000 or more went to people described in the record as alleged co-conspirators. The bank also processed more than $7 million in apparent settlements to law firms and more than $6 million in other apparent legal expenses. Taken individually, none of those payments proves a crime. Their significance lies in how the alerts were closed. In March 2017, the payment to the Russian model or publicity agent was treated as normal for Epstein. In May 2018, a tuition wire to a person with an Eastern European surname and an account at a Russian bank received a two-part explanation: she was a friend and the money paid for her studies. According to the regulator, the bank asked no further questions. The cash pattern is more concrete. From 2013 to 2017, one of Epstein’s lawyers made 97 withdrawals of $7,500 at the Park Avenue branch, totalling $727,500. That amount was the branch limit for third-party withdrawals. In 2014, the lawyer asked how frequently he could withdraw that sum without triggering an alert. In 2017, after an employee explained the reporting rule for amounts above $10,000, a larger withdrawal was split across two days. The bank examined possible structuring, accepted the lawyer’s denial and allowed the withdrawals to continue. Including a $100,000 cash transaction in 2018, withdrawals exceeded $800,000 over the Deutsche relationship. A termination letter was sent on 21 December 2018, after the renewed Miami Herald investigation. Despite that decision, the relationship manager drafted reference letters for two other financial institutions stating that he knew of no problems. Documents released since then indicate, according to Reuters, that some accounts and services remained active until the July 2019 arrest. The decision to exit, the notice and the operational closure of each account must therefore be treated as separate dates. The 2026 Senate report attributes more than $250 million in wires from 2013 to 2019 to Deutsche Bank’s retrospective SARs. It cites one filing covering 1,140 wires and $147 million. That example sits inside the amount above $250 million; it is not an additional sum. BNY Mellon: the same money at two banks The January 2026 letter to BNY Mellon says that a 2019 filing identified 270 incoming and outgoing wires worth $378 million. According to notes taken by investigators during their Treasury review, the bank had found no clearly verifiable business purpose for the transactions. The exact period and the list of all 270 wires are not public. The letter nevertheless supplies a decisive sample: eighteen round-number transfers of $1 million sent in 2007 from BNY accounts connected to Epstein or Financial Trust to JPMorgan accounts. The published table adds to $18 million; the letter refers to at least $20 million, indicating that the list is not exhaustive. Another chain is still more revealing. On 15 June 2007, an Epstein account at BNY sent $7.4 million to Ghislaine Maxwell’s account at JPMorgan. The same day, $7.4 million was transferred to Air Ghislaine. Three days later, that company paid $7.3 million to Sikorsky as a helicopter deposit. That is not three separate lots of $7.4 million in new money. It is a payment chain whose links can appear in more than one monitoring system. Bank of America and the payer-side record Bank of America occupied a different position. The records do not present it as Epstein’s main bank, but as the bank for Leon Black-controlled accounts that sent money to Financial Trust at JPMorgan and later Southern Trust at Deutsche Bank. The Senate report reconstructs eighteen wires worth $169.8 million from 2012 to 2017. The first instalment explained why that figure does not equal the $158 million of fees retained by the Dechert review: it includes, among other amounts, $5.5 million in 2012 and $6.3 million in 2016, outside Dechert’s fee schedule. For this article, the decisive information is the reporting date. According to Senate staff, Bank of America did not file the cited SARs until 2020, years after the wires and after Epstein’s death. The report says the bank could not identify a verifiable business purpose. Bank of America says that it did not facilitate Epstein’s crimes and that it takes its compliance obligations seriously. The proposed $72.5 million class settlement with survivors has received only preliminary approval. As of 7 August 2026, the final hearing remains scheduled for 27 August. The bank denies wrongdoing. A civil settlement is not a SAR, a regulatory fine or an admission. Penalties do not measure transaction flows Three categories are routinely mixed together: reported transactions, regulatory penalties and civil settlements. The NYDFS imposed a $150 million penalty on Deutsche Bank in 2020. That amount covered the Epstein matter alongside failures involving FBME Bank and Danske Bank Estonia. It cannot be assigned entirely to Epstein. In 2023, JPMorgan agreed to a $290 million settlement in the survivors’ class action and a separate $75 million agreement with the US Virgin Islands. Deutsche Bank settled the class action against it for $75 million. Those agreements ended litigation without a trial on the merits and without a general admission of liability. The case against BNY Mellon was dismissed with prejudice in February 2026 and an appeal was filed on 20 March. The dismissal means that the claims did not clear the legal threshold applied by the court. It does not certify the economic purpose of each of the 270 wires discussed in the Senate letter. Finally, the FCA permanently barred Jes Staley from senior management roles and fined him roughly £1.1 million for approving a misleading letter about his closeness to and last contact with Epstein while he ran Barclays. That decision confirms the nature of their personal relationship. It is not a ruling on JPMorgan transaction flows. Established alerts and remaining unknowns The public record supports five propositions. First, JPMorgan detected and reported transactions as early as 2002. The bank was therefore not devoid of signals before the 2008 conviction. Second, identifiable decision-makers chose to retain Epstein in 2011, and in 2013 authorised some activity through third-party clients after closing his direct relationship. Third, Deutsche Bank accepted a client whose history it knew, while anticipating large flows and revenues. It designed monitoring conditions that were not transmitted to the teams expected to apply them. Fourth, both institutions processed years of cash patterns, trust transfers and foreign counterparties. The records do not prove that every operation was unlawful. They show that the activity generated enough questions to trigger reviews and alerts. Fifth, the mass filings were retrospective. JPMorgan’s largest SAR came six years after the direct client exit. The giant reports attributed to Deutsche Bank, BNY Mellon and Bank of America followed the 2019 arrest. The decisive material required to move from a history of controls to a complete accounting remains unavailable: each SAR’s transaction list, the included account identifiers, the exact dates of internal detection and the decisions connecting each alert to a filing or a decision not to file. Reuters, which collected the banks’ responses to the 4 August 2026 report, said it could not independently verify all of its details. JPMorgan says that it reported transactions during and after the relationship and acted appropriately. Deutsche Bank says that it regrets the historical relationship, cooperated with authorities and addressed the deficiencies. Bank of America denies facilitating any crime. Those responses do not dissolve the timeline. They establish its adversarial boundary. The banking system did not simply “miss” Epstein. It observed him in fragments: a cash withdrawal here, a counterparty there, an annual review, a risk committee, a closed alert. Each fragment could receive an isolated explanation. Their organisational sum was reconstructed only after the fact. The demonstrable scandal is therefore not a round number in the billions. It is the length of time during which known signals remained administratively compatible with continuing the business. Read the French original. Sources - US Senate Finance Committee: Looking the Other Way: Wall Street’s Role in Financing Jeffrey Epstein, 4 August 2026 - US Senate Finance Democratic staff: memorandum on JPMorgan and Epstein, 20 November 2025 - JPMorgan Chase: response to Senator Wyden, October 2025 - New York Department of Financial Services: Deutsche Bank consent order, 6 July 2020 - US Senate Finance Committee: letter to BNY Mellon, 15 January 2026 - Jorge Amador / Axia Advisors: expert report, Government of the USVI v. JPMorgan, exhibit 238-31 - FinCEN: official answers on SAR obligations and filing deadlines - FCA: confirmation of the Jes Staley ban and penalty, 25 July 2025 - JPMorgan Chase: US Virgin Islands settlement, 26 September 2023 - SDNY: docket and appeal in Doe v. BNY Mellon, current to 23 March 2026 - Reuters: banks’ responses to the Senate report, 4 August 2026 Limitations The Senate reports are institutional staff work produced under the Finance Committee’s ranking Democratic member. They draw partly on SARs reviewed in camera, but remain political reports rather than judgments. The NYDFS order is a set of regulatory findings accepted by Deutsche Bank. The Amador report is expert evidence submitted by a party to litigation. Civil allegations, settlements without admissions and procedural rulings are identified as such. Bank aggregates are never added across institutions. The amounts measure reported transaction volumes, not Epstein’s income, net worth or established criminal proceeds. Corpus current to 7 August 2026. ============================================================================ ANALYSIS: Epstein’s money, 1/4: the fortune without a ledger URL: https://l0g.fr/en/analysis/epstein-money-fortune-without-ledger/ Canonical French source: https://l0g.fr/posts/argent-epstein-fortune-sans-grand-livre/ Date: 2026-08-07 (reviewed 2026-08-09) Topics: Jeffrey Epstein, finance, wealth management, tax, U.S. Virgin Islands, investigation ---------------------------------------------------------------------------- Jeffrey Epstein’s fortune is routinely described as a mystery. That is not quite right. Part of his income, assets and transfers can now be documented. The real problem is that the public records never form a continuous general ledger. They layer salaries, fees, loans, repayments and asset valuations whose scopes do not match. When Epstein died in August 2019, he reported a net worth of more than $577 million. Forty years earlier, his Bear Stearns personnel file showed an annual salary of $42,000. In between, there was no initial public offering, no documented company sale and no fund whose performance can be audited in public. There are, however, firm anchors. The first is Epstein’s relationship with Leslie Wexner: an exceptionally broad power of attorney, real-estate transactions and, eventually, a private $100 million repayment. Next came Epstein’s advisory companies in the U.S. Virgin Islands. For the older Financial Trust Company, financial statements provide large aggregates but almost no client detail. Southern Trust Company is the reverse: five reporting years, three identifiable sources of payment and $183,999,980 in fees that can be reconciled to the dollar. Finally, the Leon Black records explain both a large share of Epstein’s late-stage income and the continuing confusion between $158 million, $164.3 million, approximately $166 million and $169.8 million. This first instalment therefore does not claim to solve a biographical riddle. It builds an accounting of evidence: what is established, what is alleged, and what no public record yet connects. The salary that explains nothing The Bear Stearns file matters precisely because its numbers are modest. A copy of Epstein’s personnel records, released in the U.S. Department of Justice’s Epstein archive, traces a quick rise: $225 a week when he joined as a trainee in March 1976, $300 in August, $24,000 a year in 1977, $32,000 and then $36,800 in 1978, and $42,000 in April 1979. An October 1979 record also refers to a $20,000 loan, repayable from Epstein’s May 1980 bonus. That proves a loan, not a $20,000 bonus; still less annual compensation of $177,000 or $200,000, figures sometimes repeated without a payslip or tax form to support them. Epstein resigned as a limited partner in March 1981. The same personnel file records a $2,500 fine and suspension related to credit and new-issue allocation rules. These figures do not rule out bonuses or personal investment income. They simply set a limit on what the public record supports: the available employment documents do not explain the accumulation of a fortune worth hundreds of millions. A $9.2 million settlement Epstein received from Bear Stearns in 2011 does not alter that conclusion. It arose from an investment dispute three decades later; it was not 1970s employment income. The following decade is the first major blind spot. Epstein formed J. Epstein & Company and spent time around Towers Financial, but the public corpus supplies no W-2, commission statement or reconciled client list for 1981 · 1991. Assigning no amount to that period does not mean that he earned nothing. It means that no figure currently clears the evidentiary threshold used here. Wexner: documented authority, alleged misappropriation The decisive shift began in the early 1990s. Leslie Wexner, founder of The Limited, gave Epstein extraordinary management authority. In a public letter issued in 2019, Wexner wrote that he had granted Epstein a very broad power of attorney, allowing him to act on Wexner’s behalf in financial matters. Wexner said that in 2007 he discovered that “vast sums” had been misappropriated, severed the relationship and recovered part of the money. A federal prosecutors’ memorandum dated July 2019 supplies a number and, just as importantly, identifies the type of source. The memorandum summarizes a proffer made by Wexner’s lawyers to federal prosecutors in Manhattan. According to counsel, Epstein had misappropriated “several hundred million dollars” and paid himself fees without Wexner’s knowledge. Prosecutors wrote that the alleged conduct appeared to account for virtually all of Epstein’s wealth. They also recorded claims that Epstein had acquired Wexner properties before selling them to himself for a fraction of their value, and had obtained an aircraft at a steep discount. This is an official document, but its contents are neither an adversarial audit nor a judicial finding. It records the account given by Wexner’s counsel. The distinction is fundamental: “several hundred million dollars” is the primary amount of the allegation, not damages established by a court. The same memorandum says that a private settlement led Epstein to return $100 million in January 2008. A report commissioned by the Wexner Foundation describes part of the mechanism. On January 1, 2008, an entity called the COUQ Foundation transferred assets recorded at $12,377,844 and 201,939 Apple shares valued at $34,280,154 to the YLK Charitable Fund, a combined $46,657,998. The report characterizes the transaction as a partial recovery. It then refers to roughly $35 million remaining in YLK and transferred to the Wexner Family Charitable Fund in 2010 and 2011. Adding $100 million, $46.7 million and $35 million would be tempting, and wrong. Without the private settlement or complete entity ledgers, there is no proof that the charitable transfers sit outside the $100 million rather than forming components or later stages of the same recovery. The power of attorney presents the same documentary problem. Public sources establish that it existed by 1991 and that it was broad. We have not, however, located a complete signed copy in a freely accessible official filing that would allow every power and limitation to be inventoried. The existence and breadth of the instrument are documented; its exact text remains missing. The best-known property transfer was not free The townhouse at 9 East 71st Street in Manhattan illustrates the damage caused by misreading registries. The 1998 Wexner · Epstein transaction is often described as a mansion gifted for zero dollars. An archive index released as EFTA00300480 instead lists a file titled Leslie H. Wexner Sale of Nine East 71st Street Corporation to NES, LLC, with a sale, promissory note, personal guarantee and general ledger. Documents from that file described by Vanity Fair put the price at about $20 million: approximately $10 million in cash and $10 million financed by the note. Without a complete public facsimile of every instrument, the price should remain approximate. The zero-dollar ACRIS filing from 2011 records a later transfer between Epstein-related entities. It does not prove that Wexner donated the house to Epstein. The official ACRIS portal records real-property instruments; EFTA00300480 documents the existence of the transaction file, not the full execution of the transaction by itself. The difference between a deed and a sale of shares in a property-holding company is not cosmetic: collapsing the two turns a financed sale into a gift. Two Ohio transactions add context without independently proving improper enrichment. A New Albany property acquired around 1992 for $3.5 million was sold in 1998 for $8 million. Another house, at 7558 King George Drive, was transferred without consideration in 2007 to a trust for Abigail Wexner and later sold for $365,000. These chains must be checked instrument by instrument in the Franklin County Recorder’s official database. They show assets moving between the Wexner and Epstein spheres; they do not reveal the balance of their accounts or the net amount of any misappropriation. Financial Trust: $300 million without a public client ledger In 1996, Epstein obtained a U.S. Virgin Islands charter for Financial Trust Company, a business presented as serving clients worth at least $1 billion. Financial statements cited in the court record attribute roughly $300 million in fees to Financial Trust from 2000 through 2006, including $66 million in 2006 alone. The curve then collapsed: less than $4 million in 2007, roughly $100,000 in 2008 and about the same in each of the next three years, followed by no revenue in 2012. From 2008 through 2012, the company accumulated approximately $166 million in net losses. The figures matter, but they do not answer the central question: who paid? The problem is not a total absence of tax material. Financial statements, returns and U.S. Virgin Islands Economic Development Commission files exist and were produced in litigation. The problem is their public granularity. Income statements show annual revenue; they do not publish the customer subledger, contracts, invoices and wire records needed to allocate roughly $300 million dollar by dollar. The 2019 proffer and the scarcity of other demonstrable clients strongly suggest a concentration around Wexner. But that inference cannot honestly be converted into an exact client breakdown without Financial Trust’s books. The public record supports the conclusion that Epstein made enormous amounts from wealth management and advice before Leon Black. It does not support the sentence: “Here is every client and exactly what each one paid.” Southern Trust: an exact three-payer reconciliation Southern Trust Company, organized in 2012, is far less opaque for 2013 through 2017. Accounting expert Bruce G. Amador’s report in the U.S. Virgin Islands lawsuit against JPMorgan cites the company’s annual financial statements and identifies Southern Trust as Epstein’s only entity generating significant revenue during that period. Fee income was $51 million in 2013, $70 million in 2014, $54,999,980 in 2015, zero in 2016 and $8 million in 2017. Total: $183,999,980. Three sets of contracts, invoices and wires reconcile that amount exactly: - Leon Black: $158 million, comprising $50 million in 2013, $70 million in 2014, $30 million in 2015 and $8 million in 2017; - Edmond de Rothschild bank: $24,999,980 in 2015, paid through two wires of $10 million and $14,999,980 after a $25 million advisory agreement; - Steven Sinofsky: $1 million in 2013, confirmed by a wire instruction and completion notice to Southern Trust. The arithmetic is exact: $158,000,000 + $24,999,980 + $1,000,000 = $183,999,980. No unattributed fee income remains in the available statements. That reconciliation does not come from tax returns alone. It emerges from matching the accounts to the Rothschild agreement, its invoice, bank records and the Sinofsky wire. That completeness must be stated narrowly. It applies to Southern Trust’s fee income line over the five available reporting years. It does not mean that these three people or institutions encompass all of Epstein’s financial relationships, or that every payment had the same economic purpose. Nor does it independently establish the quality or legitimacy of the invoiced services. Leon Black: why four totals circulate The independent review commissioned by Apollo and conducted by Dechert produced the best-known figure: $158 million in fees paid by Leon Black to Epstein from 2013 through 2017. The report filed with the SEC breaks the payments down as $50 million in 2013, $70 million in 2014, $30 million in 2015, no fee in 2016 and $8 million in 2017. Dechert describes an agreement signed on February 13, 2013 for $23.5 million, paid in two instalments of $15 million and $8.5 million. A second draft for approximately $56.5 million appears in the spring 2013 records, but the copy located was unsigned. After 2013, according to the report, the relationship became largely ad hoc and services were not covered by written agreements. A $20 million invoice in 2014 concerned work on a tax-basis step-up. In 2015, a $35 million invoice resulted in a $30 million payment; in 2017, a draft $11 million invoice resulted in an $8 million wire. An invoice establishes a demand. The wire establishes the payment. Dechert reported finding no evidence that Black paid Epstein for an illegitimate purpose. It nonetheless found that the fees were far above those paid to Black’s other advisers and described documentation that was sometimes thin. Both findings must remain intact: no evidence of an improper purpose, but extraordinary compensation and weak contracting. The Senate Finance Committee report, based in part on banking information and suspicious-activity reports reviewed in camera, uses a different scope. Its annual Bank of America wire table totals $169.8 million from 2012 through 2017: $5.5 million in 2012, $50 million in 2013, $70 million in 2014, $30 million in 2015, $6.3 million in 2016 and $8 million in 2017. The reconciliation is straightforward. $169.8 million minus $158 million equals $11.8 million, exactly the $5.5 million in 2012 and $6.3 million in 2016 that Dechert did not count as 2013 · 2017 fees. For 2013 through 2017, the banking table totals $164.3 million. Elsewhere, the Senate report’s narrative rounds or describes the same period as “approximately $166 million,” creating an internal $1.7 million gap with its own table. Without the complete underlying wires, that $1.7 million should not be invented or assigned. The loans form yet another circuit. In 2017, Epstein advanced $22.5 million and then $8 million to Black-related entities under “Plan D,” for a total of $30.5 million. A promissory note initially drafted for $28.5 million was revised to $30.5 million. Ten million dollars was repaid on October 2, 2018; $20.5 million remained outstanding when Epstein died. The advances should neither be treated as Epstein income nor added to the $158 million in fees. They were assets, receivables whose repayment changed cash, not revenue. Highbridge: $15 million proved, $2.25 million only proposed Before Leon Black, Highbridge Capital Management provides one of the rare large payments that can be tied to a named client. In a statement of facts filed by the U.S. Virgin Islands, paragraph 379 states that in December 2004 Highbridge paid Financial Trust $15 million for merger-and-acquisition advice. Epstein later said $20 million in a deposition. The contemporaneous $15 million payment is the better measure. Another draft Highbridge agreement, nominally effective from June 2005 through December 2009, proposed $2.25 million in five $450,000 instalments, no minimum time commitment, and investment capacity on terms comparable to employees. But the available signature page is blank. Without an executed copy or payment record, the $2.25 million remains a contractual proposal, not earned income. The distinction captures the method required by the Epstein archive: a draft is not a contract, an invoice is not a payment, a wire is not necessarily income, and an asset value is not cash. A visible fortune before Black, but still not auditable Internal records show that Epstein was already extremely wealthy before the Black fees. A statement dated July 31, 2012 valued his assets at $289,022,838. Another reached $340,906,529 in January 2014. A 2009 · 2010 estimate approached $319.5 million while excluding residences, an aircraft and some personal property. These are internal management documents, not audited accounts; their exclusions and valuation methods differ. The real-estate portfolio confirms pre-Black accumulation: the Manhattan residence acquired in 1998 through a stock sale for about $20 million, Little Saint James bought that year for roughly $7.95 million, and the New Albany property acquired for $3.5 million and sold for $8 million. Other assets, including Palm Beach, Zorro Ranch, Paris, aircraft, securities and private holdings, appear in balance sheets or registries, but the public chain does not always disclose the original price, source of funds, attached debt and final sale proceeds. Financial Trust fees, the $15 million Highbridge payment, investment income and Wexner-related transfers therefore provide partial and sometimes enormous explanations. They do not form a year-by-year bridge from a $42,000 salary in 1979 to $289 million in assets in 2012. The missing link is not one secret. It is the set of general ledgers, bank statements, complete tax returns, executed agreements and recovery settlements required to eliminate double counting. Findings supported by the records Four conclusions survive a hostile audit. First, known Bear Stearns income cannot by itself explain Epstein’s later wealth. Higher claims about his compensation lack the necessary public tax or payroll records. Second, Wexner is the pivot of the first major phase of documentable wealth. The broad power of attorney, the $100 million settlement and the charitable transfers are supported by primary or near-primary records. But “several hundred million dollars” remains an allegation by Wexner’s lawyers recorded by prosecutors, and the exact scope of the recoveries remains hidden in a private agreement. Third, Southern Trust’s revenue is unusually legible: $183,999,980 in fees, comprising $158 million paid by Leon Black, $24,999,980 linked to Edmond de Rothschild and $1 million from Steven Sinofsky. That precision cannot be projected backwards onto Financial Trust, whose earlier roughly $300 million in fees still lacks a complete public client allocation. Finally, $158 million, $164.3 million, approximately $166 million and $169.8 million are not interchangeable estimates of Black’s payments. The first is Dechert’s fee total; the second is the Senate table’s wires for 2013 · 2017; the third is the report’s approximate and unreconciled narrative; the fourth adds 2012. The $30.5 million in loans is a separate circuit. Epstein’s fortune is therefore neither wholly mysterious nor fully explained. It is documented in fragments, with a revealing asymmetry: the closer the record gets to Southern Trust and Leon Black, the more traceable the payments become; the further back it moves toward Financial Trust and Wexner, the more the numbers become aggregated, private or alleged. That zone, rather than Epstein’s Bear Stearns salary, is where the central accounting gap remains. Lire la version française. Sources 1. U.S. Department of Justice, memorandum regarding Leslie Wexner counsel’s proffer, EFTA02731082, July 2019. 2. Leslie Wexner, Letter from Les, Wexner Foundation, 2019. 3. Kegler Brown Hill + Ritter, Report following independent review for the Wexner Foundation, 2020. 4. Bruce G. Amador, Expert report, Government of the United States Virgin Islands v. JPMorgan Chase Bank, especially pp. 72 · 74. 5. Dechert LLP, Report of the Conflicts Committee of Apollo Global Management, exhibit filed with the SEC, January 2021. 6. U.S. Senate Committee on Finance, The Wall Street Connections to Jeffrey Epstein, Ron Wyden investigative staff report, August 4, 2026. 7. Government of the U.S. Virgin Islands, Plaintiff’s statement of material facts, docket item 285-2, especially ¶ 379 on Highbridge. 8. U.S. Department of Justice, Epstein Library, official collection of EFTA documents and court records. 9. Accessible copy of EFTA00187050, Bear Stearns personnel file, OCR mirror of a DOJ-released record. 10. Accessible copies of EFTA00584904 and EFTA00586695, Edmond de Rothschild agreement and invoice, OCR mirrors of DOJ-released records. 11. Accessible copy of EFTA00675845, Sinofsky wire instruction and confirmation, OCR mirror of a DOJ-released record. 12. Accessible copy of EFTA00589279, Plan D promissory note, OCR mirror of a DOJ-released record. 13. Accessible copy of EFTA00589969, draft Highbridge agreement, OCR mirror of a DOJ-released record. 14. New York City Department of Finance, ACRIS, and Franklin County Recorder, Public Records Search, official property registries. 15. Accessible copies of EFTA00602408, EFTA00607612 and EFTA01119119, internal valuations dated July 31, 2012, January 2014 and 2009 · 2010, OCR mirrors of DOJ-released records. 16. Accessible copy of EFTA00300480, index of Epstein’s administrative and transaction files, including the Nine East 71st Street Corp. file, OCR mirror of a DOJ-released record. 17. Gabriel Sherman, Inside Jeffrey Epstein’s decades-long relationship with his biggest client, Vanity Fair, 2021, for the secondary description of the Nine East 71st Street Corp. sale records. 18. Giacomo Tognini, How Jeffrey Epstein Got So Rich, Forbes, 2025, used to cross-check Financial Trust aggregates, the estate inventory and the asset timeline. Limitations This investigation uses public records available as of August 7, 2026. Several decisive documents are not public or were not located in a complete official version: the full signed Wexner power of attorney, the 2007 · 2008 recovery settlement, Financial Trust’s customer ledger, unredacted tax schedules and the underlying bank statements for some Senate tables. Party filings and expert reports are identified as such. They may rely on contemporaneous evidence without becoming judicial findings. Dechert’s review was commissioned by Apollo; the Wexner report by the Wexner Foundation; the Senate document is a staff report, not a judgment. Yirah copies make DOJ-released EFTA records searchable and accessible, but the official facsimile should control if the OCR or mirror differs. Finally, each number is kept within its accounting category. Fees are gross revenue before expenses and tax; loans are receivables; asset valuations are not cash; repayments may overlap. None of these figures, standing alone, proves wrongdoing by the person or institution that made a payment. ============================================================================ ANALYSIS: CLARITY Act: the window before the midterms URL: https://l0g.fr/en/analysis/clarity-act-window-before-midterms/ Canonical French source: https://l0g.fr/posts/clarity-act-fenetre-avant-midterms/ Date: 2026-08-06 (reviewed 2026-08-06) Topics: clarity act, crypto, regulation, us politics, us senate ---------------------------------------------------------------------------- Status as of 7 August 2026. This article describes procedure and a calendar. Its scenarios are possible conditions for passage, not forecasts or probabilities. The CLARITY Act does not expire on 7 August. Nor has it become law. The official H.R. 3633 tracker still labels it Passed House, following a 294-134 vote on 17 July 2025, rather than Passed Senate. That distinction matters for markets and companies: a negotiation, a committee text or a promised vote creates neither new CFTC authority nor a new SEC registration regime. 7 August is therefore a calendar boundary, not a legal expiry date. The Senate's official schedule lists a state work period from 10 August through 11 September. It then leaves fourteen scheduled session days, from 14 September through 2 October, before another state work period from 5 October through 6 November. The federal election on 3 November falls in that second break. The Senate schedule is expressly tentative, but it is the visible constraint as of publication. A narrow window, not an absolute deadline The shorthand "no vote before August means the bill is dead" goes too far. Congress can still legislate after the election so long as both chambers and the president complete the necessary steps. But a delay beyond 7 August changes the sequence: the bill is no longer fighting for one last summer week; it is competing for a compressed September slot between the return and the campaign. That slot is measured in more than calendar days. A contested bill needs floor time and votes. The Senate explains that, where obstruction persists, a cloture motion requires three-fifths of senators, or 60 votes, to limit debate on legislation. Its procedural guide also explains that unanimous consent can shorten the route, while a single objection can prevent that fast track. The immediate constraint is political. On 6 August, Axios reported that Democrats were reluctant to grant an initial procedural vote, arguing it could reduce pressure on the negotiations. In late July, CoinDesk reported that open disputes included public-official ethics and stablecoin rewards. Those reports describe negotiations, not a final text enacted into law. Four steps are still required The House-passed text must not be conflated with the Senate's negotiated version. In May, Banking Committee leaders released text as the basis for their markup, saying it reflected talks with Democrats and input from regulators, banks, law enforcement and the industry. The committee release does not turn that text into law. For a federal framework to take effect, four locks remain: 1. A publicly identifiable Senate text and a procedural route to bring it to the chamber. 2. A cloture majority if opposition keeps debate open, followed by Senate passage. 3. House agreement to the same text, directly or after amendment exchanges or a conference. 4. Presidential signature, or the constitutional period expiring without a veto under the applicable conditions. The first lock is political; the other three are institutional. Treating a negotiated compromise as if it were enacted law gives it effects it does not have. Three scenarios for September and after Scenario A: a compromise enables September passage Under this scenario, negotiators release common language, leadership secures enough votes to organize debate, and the Senate passes text the House quickly accepts. Presidential signature would then be the final step. This is the only scenario that can start a rulemaking phase before the elections. It requires several successive decisions; none of them is established as of publication. Scenario B: September is used without producing a law Leadership could instead seek a procedural vote or force positions into the open without reaching final language or 60 votes. That would be useful political information about coalitions, but not a legal change. Companies would remain under the existing SEC, CFTC and state regimes. Scenario C: the file moves after the elections No agreement before 2 October would move the issue to the post-election period. That period is often called lame duck after the vote, but it is neither automatic nor synonymous with a vote. Two uncertainties remain: actual floor time and both chambers' willingness to settle their differences before the Congress ends. These scenarios are not probabilities. A calendar identifies a constraint; it does not reveal the conduct of senators, the House or the president. Signals that would change the analysis Four events would be more informative than a statement of intent or a token price: release of compromise text, filing of cloture, a recorded procedural or passage vote, and then formal House movement on a Senate version. Until those milestones occur, calling a "CLARITY framework" effective would be inaccurate. For the architecture of the bill and its potential sector-by-sector effects, read our text analysis. For the conflict-of-interest negotiation, read our earlier August countdown, now complemented by this procedural update. Limitations The Senate labels its calendar tentative and it can change. Press reporting on negotiations attributes positions to sources; it does not substitute for filed text or a vote. The procedural rules described here explain possible routes, not party-leader strategy or a vote result. This article is not investment or legal advice. Sources - Congress.gov, H.R. 3633 status and actions, accessed 7 August 2026: House passage, Passed House status and formal actions. - U.S. Senate, 2026 legislative schedule, accessed 7 August 2026: state work periods from 10 August through 11 September and from 5 October through 6 November. - U.S. Senate, how a session works, accessed 7 August 2026: debate, unanimous consent and 60-vote cloture. - Senate Banking Committee, markup working text, 12 May 2026: committee-text status. - Axios, end-of-summer negotiations and agenda, 6 August 2026: attributed political context. - CoinDesk, procedural calendar and unresolved issues, 27 July 2026: attributed negotiation context. ============================================================================ ANALYSIS: Hong Kong Life Insurance: Beijing Taxes a Breach in Its Financial Wall URL: https://l0g.fr/en/analysis/hong-kong-life-insurance-china-tax-enforcement/ Canonical French source: https://l0g.fr/posts/assurance-vie-hong-kong-taxe-chine-muraille-financiere/ Date: 2026-08-06 (reviewed 2026-08-06) Topics: china, hong kong, life insurance, capital controls, tax, dollar ---------------------------------------------------------------------------- A life-insurance policy can also be a long-term savings product. In Hong Kong, that distinction has made it a meeting point between mainland savings, dollar-denominated assets and Chinese tax law. The news is not a newly announced national tax. It is reported enforcement in Beijing and Hangzhou of an existing rule on income from certain policies bought in Hong Kong. Reuters reports that Chinese tax residents have received notices concerning interest and dividends from certain Hong Kong insurance policies. The 20% rate is the long-standing rate in China's individual income-tax law for interest, dividends and bonuses. At publication, l0g has identified neither a detailed national text specific to these policies nor a national tax announcement changing the rate. The verified event is therefore narrower, but material: reported enforcement in two cities, not a formally announced nationwide "new tax". Reuters, 6 August 2026 · China's Individual Income Tax Law An insurance policy that is not only insurance Two shortcuts should be avoided. A life-insurance policy is not a brokerage account: the policyholder does not directly own the insurer's portfolio. Nor is it, by itself, an illegal capital outflow. Depending on the policy, it can combine protection, savings, participation in profits and benefits denominated in US dollars or Hong Kong dollars. A claim on an insurer established in Hong Kong can therefore add non-renminbi exposure to a largely mainland portfolio. For a Chinese household, the appeal is not only the quoted return. The policy may offer another currency, an offshore issuer, Hong Kong contractual rules and a beneficiary designation. Those features guarantee neither performance, liquidity nor favourable tax treatment. They do, however, explain the appeal of such policies to customers seeking to diversify part of their wealth without directly buying a foreign portfolio. The foreign-exchange boundary remains real. China's State Administration of Foreign Exchange maintains an annual US$50,000 quota for individuals' current-account foreign-exchange purchases. An insurance policy does not, on its own, create an exemption from that rule. Calling it a wholly open door would be as wrong as reducing it to a simple death benefit. State Administration of Foreign Exchange An old rule, more visible data The rate itself is not new. China's law has long applied a flat 20% rate to interest, dividends and bonuses. It also taxes Chinese tax residents on income from Chinese and foreign sources. The crucial question is classification: which payment falls into which category, for which tax resident, at what time and after which possible foreign-tax credits. Answering that without the policy, residence facts and tax notice would be tax advice without a file. Automatic exchange of information changes the enforcement context, not the statute. In Hong Kong, the Common Reporting Standard requires a broad set of financial institutions, including specified insurance companies, to identify reportable accounts and transmit annual identity data, balance or value, interest, dividends and income from certain insurance products. Hong Kong's tax authority then transmits the information to the relevant residence jurisdiction where an exchange arrangement is in place. Hong Kong Inland Revenue Department That system makes more systematic detection of offshore income plausible. It does not establish that CRS triggered the notices reported by Reuters, or that every Hong Kong policy is reportable in the same way. An information-transparency tool is not, by itself, a tax-assessment instruction. A separate recent measure shows why the objects must be kept apart. In July, China's Ministry of Finance and tax administration issued formal guidance on the taxation of offshore trusts, including a 20% rate for certain income. That announcement concerns trusts, not life insurance. It cannot turn the reporting on Hong Kong policies into a published national insurance reform. China tax administration, 24 July 2026 A market too large to be anecdotal This is not only a story about a few wealthy policies. Hong Kong's Insurance Authority puts total gross premiums in 2025 at HK$827 billion, up 29.7%. New office premiums in long-term business, excluding retirement schemes, reached HK$330.9 billion, up 50.6%. Long-term insurance assets stood at HK$5.398 trillion at year-end. Those figures cover the entire Hong Kong market. They do not measure only mainland customers or only the policies potentially covered by reported tax notices. Hong Kong Insurance Authority AIA offers a more precise indication of the role of mainland Chinese visitors without giving a corresponding premium amount. In 2025, its Hong Kong value of new business rose 28% to US$2.256 billion. Value of new business from mainland Chinese visitors rose 35%, compared with 21% for domestic customers. Prudential says Hong Kong and its mainland joint venture each achieved double-digit new-business-profit growth in the first quarter of 2026, without disclosing market-by-market rates. These commercial measures do not quantify the impact of this week's reported enforcement. They show the value of a customer segment insurers will not want to lose. AIA, 2025 annual results · Prudential, Q1 2026 Possible consequences, not a closure already decided Three paths should be kept separate. - Local compliance enforcement. Notices remain confined to certain taxpayers and products. The tax and administrative cost rises, but policies continue to be bought where the wealth-planning value outweighs it. - More systematic enforcement. Foreign-income declarations become a visible part of the sales process. Insurers, brokers and banks must more clearly separate insurance, savings, tax and foreign exchange. Demand could weaken, but the result depends on the tax base ultimately applied and on available alternatives. - An explicit restriction on subscriptions or flows. That step would require an identifiable rule or instruction. No verified public evidence as of 6 August supports that conclusion for Hong Kong insurance policies. The third scenario explains investor sensitivity, but it remains a scenario. Saying today that Hong Kong has been closed to mainland savers, or that their policies will be confiscated, would go beyond the available facts. Is the financial wall really tightening? Political intent should not be invented. Without a national announcement on life insurance, it is not possible to attribute to Beijing an official objective of repatriating savings, defending the renminbi or widening the tax base through this exact channel. The economic mechanism can still be described: making foreign taxation more effective raises the relative cost of an offshore claim and reduces the advantage of opacity. The phrase "a breach in the financial wall" is therefore a metaphor, not a legal finding. Hong Kong retains its separate regime, insurance market and international funding links. But a policy bought in Hong Kong does not make its holder permanently invisible. That is the news in this episode: not the disappearance of a channel, but its return to China's tax perimeter. Read the French original: Assurance-vie à Hong Kong : Pékin taxe une brèche dans sa muraille financière. Sources 1. Reuters, "Hong Kong insurers' shares slide on report China to tax offshore insurance income", 6 August 2026: reported notices in Beijing and Hangzhou, products concerned and the absence of a detailed identified announcement. 2. China tax administration, Individual Income Tax Law of the People's Republic of China: income categories and the 20% rate for interest, dividends and bonuses. 3. Hong Kong Inland Revenue Department, Automatic Exchange of Information: CRS institutions and information, including certain insurance products, and the exchange logic. 4. Hong Kong Insurance Authority, 2025 provisional statistics, 24 April 2026: market premiums, new business and assets. 5. AIA, 2025 annual results, 19 March 2026: Hong Kong value of new business, domestic customers and mainland Chinese visitors. 6. Prudential, Q1 2026 business update, 29 April 2026: new-business-profit growth in Hong Kong and mainland China. 7. State Administration of Foreign Exchange, press conference on foreign-exchange receipts and payments, 19 January 2017: the US$50,000 annual individual purchase quota, described as unchanged. 8. China tax administration, offshore-trust tax clarification, 24 July 2026: a separate trust measure, used here to avoid conflating it with life insurance. Limitations The initial information on tax enforcement against Hong Kong insurance policies comes from Reuters' reporting. Without a published detailed national instruction, this article cannot establish the exact tax base by policy type, the geographical scope of notices, the precise reporting obligations, possible foreign-tax credits or the effect on policy sales. It is not tax, investment or insurance advice. ============================================================================ ANALYSIS: US Productivity: Real Pay Still Lags URL: https://l0g.fr/en/analysis/us-productivity-real-pay-lags/ Canonical French source: https://l0g.fr/posts/productivite-americaine-salaire-reel-derriere/ Date: 2026-08-06 (reviewed 2026-08-06) Topics: United States, productivity, wages, employment, inflation ---------------------------------------------------------------------------- In the second quarter of 2026, US businesses produced more for each hour worked. That gain did not translate into more purchasing power in hourly compensation. Both statements are true, but they do not measure the same thing. The Bureau of Labor Statistics (BLS) estimates that labour productivity in the nonfarm business sector rose at a 1.4% annual rate in Q2. Output rose 1.7% and hours worked 0.3%. At the same time, nominal hourly compensation increased 2.7%, but it fell at a 3.1% annual rate after inflation, using the consumer-price index applied by the BLS. Over four quarters, real hourly compensation was still down 0.1%. BLS, August 6, 2026 This is neither proof that all households are becoming poorer nor a measurement of AI's effect. It is a narrower signal: more output per hour does not automatically raise real labour income. The BLS also reports labour's share of output at 52.9%, the lowest level in a series beginning in 1947. Three figures with different jobs Productivity here means real output per hour worked in the nonfarm business sector. It is not a measure of one worker's performance or pay. Real hourly compensation includes wages and employer-paid benefits, then adjusts them for inflation. It is therefore broader than an hourly wage on a payslip. Its 3.1% Q2 fall is an annualised quarterly rate: it describes the pace observed between Q1 and Q2 as though it continued for a year. It does not mean a worker has already lost 3.1% of purchasing power since March. Unit labour costs divide hourly compensation by productivity. They rose at a 1.3% rate in Q2 and were up 1.4% over four quarters. For a company, this measure indicates whether labour cost per unit of output is accelerating or slowing. Here, productivity absorbed part of the rise in nominal compensation without turning it into a real gain for workers. BLS @media (max-width: 760px) { figure.productivity-chart { width: 100%; max-width: 100%; min-width: 0; overflow-x: auto; contain: inline-size; } figure.productivity-chart svg { width: 760px !important; max-width: none; } } An active economy with a less favourable split for labour The picture is not of an economy that has stopped. The Bureau of Economic Analysis (BEA) estimates that real US GDP rose at a 1.5% annual rate in Q2. Real final sales to private domestic purchasers, which combine consumption and private fixed investment, rose 3.9%. BEA, July 30, 2026 Those figures do not contradict the BLS release. They cover different perimeters. The BEA measures aggregate activity, and its private-demand measure does not tell us how output is divided among wages, profits and taxes. The BLS measures output and costs per hour in nonfarm business. Subtracting one rate from the other to infer a household "loss" would be wrong. The point of attention lies elsewhere. Output and demand can grow while real compensation stalls or falls. If that divergence persists, consumption becomes more dependent on other supports, including accumulated saving, credit or employment. Today's data cannot identify which support is at work, or establish that households are already under stress. They only show that productivity growth cannot settle the question of disposable income. Risk is not a proven recession The word "risk" needs precision. The BLS release does not demonstrate a consumer crisis, let alone a shock caused by AI, tariffs or the White House. It also does not measure the income distribution across household groups. It does identify three mistakes worth avoiding. - Treating a productivity gain as a living-standard gain. The first raises output per hour. The second also depends on prices and on how value added is shared. - Ignoring unit labour costs. Their 1.3% increase in Q2 remains moderate. That can ease pressure on margins and prices without improving real purchasing power. - Attributing the figure to AI. The BLS does not break productivity down by technology. Our review of AI and productivity evidence explains why one aggregate quarter cannot isolate a causal effect. This distinction also matters for markets. A company can report stronger margins because productivity improves. At economy level, sustained demand still requires households to have real income or another sustainable way to fund spending. The latest release cannot choose between those paths. It does rule out treating them as the same thing. The next tests Four releases or series will test this signal: - the July employment report, due August 7; - the second Q2 GDP estimate and corporate-profits release, due August 26; - the BLS revision to Q2 productivity, scheduled for September 3; - the four-quarter path of real compensation, unit labour costs and labour's share. One preliminary quarterly release does not make a trend. The four-quarter comparison is quieter, but points in the same direction: productivity +2.2%, real hourly compensation -0.1%, unit labour costs +1.4%. BLS Sources 1. Bureau of Labor Statistics, Productivity and Costs, Second Quarter 2026, Preliminary, August 6, 2026: productivity, output, hours, nominal and real hourly compensation, unit labour costs, labour share, revisions and release calendar. 2. Bureau of Economic Analysis, GDP, Advance Estimate, Second Quarter 2026, July 30, 2026: real GDP, real final sales to private domestic purchasers, price indexes and revision calendar. Limitations The BLS Q2 release is a first estimate and will be revised. Quarterly changes are stated at annual rates, while the four-quarter comparison spans a full year: they should not be read as the same horizon. The BLS deflates real hourly compensation with CPI-U, not the BEA's PCE price index. Finally, these aggregates cannot attribute the divergence to a technology, a policy, an industry or a household group. This article is not investment advice. ============================================================================ ANALYSIS: SoftBank: OpenAI’s Private Valuation Becomes a Liquidity Risk URL: https://l0g.fr/en/analysis/softbank-openai-liquidity-risk/ Canonical French source: https://l0g.fr/posts/softbank-openai-risque-liquidite/ Date: 2026-08-06 (reviewed 2026-08-06) Topics: AI, credit, liquidity, OpenAI, SoftBank ---------------------------------------------------------------------------- A financing contract does not settle OpenAI’s value. It sets out what happens if that value falls far enough to become a cash issue. On 5 August, SVF II TSUBAKI (DE) LLC, a wholly owned subsidiary of Vision Fund 2, one of SoftBank’s investment funds, entered into a $10 billion loan agreement. Drawdown was expected in August. SoftBank Group guarantees the facility, whose principal is due in August 2028. Its proceeds may be used for general corporate purposes of both SoftBank Group and Vision Fund 2, not just for OpenAI. SoftBank Financial Report, pp. 69-70 The material line is elsewhere. The financial report says that, under specified conditions in the agreement, a significant decrease in fair value of the OpenAI preferred shares held by Vision Fund 2 triggers cash-collateral provision and mandatory prepayment. Fair value is the estimate recorded in the accounts, not a continuously quoted stock-market price. The contract does not disclose the trigger threshold, haircut, or initial cash amount. Nor does it report that the clause has been activated. SoftBank Financial Report, p. 70 The risk is therefore not “OpenAI falls, SoftBank defaults.” It is narrower and more concrete: a private value, without a continuously quoted market price, has entered a debt mechanism that can require cash before maturity. Debt backed by cash, not shares for sale That wording changes the analysis. The report identifies a cash-collateral account of the borrower as collateral: money set aside to protect the lender. It does not say the lenders hold OpenAI shares and can sell them. The preferred shares nevertheless serve as a reference for deciding whether contractual protections must be reinforced. SoftBank Financial Report, p. 70 Such a structure does not make an unlisted share liquid. It is designed to protect the lender before a liquidation would be needed. If the condition is met, the pressure appears in cash management: more cash must be placed in the specified account or debt must be reduced. That is the difference between valuation risk, which first affects the balance sheet, and liquidity risk, which affects the capacity to pay on a given date. The disclosure does not support a more precise claim. It does not say how fair value is calculated, how often it is reviewed, who validates it, or what constitutes a “significant” decline. Saying the loan has already created a margin call, or will inevitably create one, would go beyond the public facts. A crowded funding calendar At 30 June, SoftBank said it had invested $44.6bn cumulatively in OpenAI and valued the stake at $89.6bn. SoftBank Financial Report, p. 4 It then completed a second $10bn tranche on 1 July. A third $10bn tranche is planned for 1 October, subject to the stated conditions. SoftBank’s 1 July 2026 release This sequence does not demonstrate a funding problem, but it makes any cash-locking covenant more material. The $40bn bridge facility was arranged primarily for the OpenAI investment and matures on 25 March 2027. The report says $20bn was drawn in April, $3.6bn was repaid during the quarter, and a further $10bn draw funded the July tranche. SoftBank Financial Report, pp. 4-5 The bridge and August loan have different structures. It would be unsafe to call the latter a refinancing of the former without a document saying so. A necessary correction to June’s signal In June, Bloomberg reported stalled talks over a separate $6bn margin loan backed by the OpenAI stake. Our analysis at the time treated this as a strong signal of market reluctance to monetise the private asset. The 6 August disclosure requires that conclusion to be narrowed. Mandated lead arrangers including Goldman Sachs, JPMorgan, Mizuho, Apollo Global Funding and SMBC participated in arranging a new $10bn loan for a Vision Fund 2 vehicle. SoftBank Financial Report, p. 70 The public documents do not say whether this facility is the same transaction as the one reported in June, or whether the negotiations, collateral and terms are comparable. They therefore do not support either “banks refused OpenAI” or “banks changed their mind.” The new fact is more specific: SoftBank has accepted financing under which a decline in private fair value can trigger a cash protection mechanism. That does not validate OpenAI’s valuation. Nor does it prove that valuation wrong. It makes the possible route from a balance-sheet mark to cash management visible. Evidence that would settle the question Four disclosures would make the risk measurable rather than rhetorical: - confirmation of actual drawdown and the initial cash-collateral amount; - the fair-value decline threshold and contractual haircut; - the method and frequency used to value the OpenAI preferred shares; - any cash deposit or mandatory prepayment reported in SoftBank’s next filings. Without those data, the analysis must remain conditional. The transmission mechanism is public. Its probability of activation and cost are not. Sources 1. SoftBank Group, Financial Report for the Three-Month Period Ended June 30, 2026, published 6 August 2026: OpenAI investment, $40bn bridge, $10bn SVF2 loan, borrower, guarantee, maturity, collateral and clauses, pp. 4-5 and 69-70. 2. SoftBank Group, Investment in OpenAI, 1 July 2026: second $10bn tranche, third tranche planned for 1 October subject to the stated conditions, financed by a $10bn bridge draw. Limitations The public report does not disclose the full agreement. It omits the trigger threshold, haircuts, initial cash amount, detailed valuation method and each bank’s share of the facility. It reports an August drawdown expectation, not confirmation of funding. This article establishes neither a default, a current margin call, nor a certain connection between the August loan and talks reported in June. It is not investment advice. ============================================================================ ANALYSIS: US Mortgages Face the Insurance Bill URL: https://l0g.fr/en/analysis/us-mortgages-face-the-insurance-bill/ Canonical French source: https://l0g.fr/posts/facture-assurance-risque-credit-immobilier-americain/ Date: 2026-08-05 (reviewed 2026-08-05) Topics: housing, credit, insurance, united states, climate ---------------------------------------------------------------------------- import InfographicBars from '../../components/infographic/InfographicBars.astro'; A fixed mortgage rate does not guarantee a fixed monthly payment. In the United States, homeowners insurance is generally required with a mortgage and its cost may be collected each month by the loan servicer. A higher premium can therefore reduce a household's available budget without changing either the interest rate or the outstanding principal. A Federal Reserve Bank of Dallas study, published in March 2026, links this mechanism to mortgage delinquency. It identifies a risk channel. It does not show that a national mortgage-credit crisis has already begun. The premium can move while the rate does not For many US mortgages, the loan servicer collects a monthly reserve called an escrow payment. It is used, among other things, to pay homeowners insurance and often property taxes. Fannie Mae notes that most lenders require insurance when there is a mortgage and that a servicer may collect the premium with the monthly payment.[^fannie-guide] The mechanics are straightforward. The loan agreement fixes the interest rate. The insurer reprices the policy at renewal. If the premium increases, the monthly reserve increases as well. That is neither a disguised rate hike nor a reset of the loan. It is a housing cost added to debt service. This matters because mortgage debt is long-lived. A household may have taken out a low-rate mortgage before rates rose, then face a larger insurance bill years later. Looking only at the mortgage rate leaves out a moving part of housing cost. Dallas Fed research documents the channel, within a defined scope The Dallas publication uses ICE McDash data, which record observed insurance payments for mortgage borrowers and cover roughly two-thirds of the US mortgage market. It reports that premiums rose by about 70 percent nationally from 2019 to 2025. In 2025, insurance represented 14 percent of the monthly payment measured in the study, which includes principal and interest, up from 10 percent in 2013.[^dallas] The more important figure is the associated credit finding. Related research, using 2015-23 loan data matched with credit records and moves, estimates that premium increases pushed roughly 31,000 mortgages into delinquency in 2022.[^dallas] This is an econometric research estimate, not an administrative count of defaults with insurance as their sole cause. The study also finds higher credit-card use and balances following premium increases. That offers a natural link to our analysis of deferred household debt: housing, insurance and revolving credit can pressure the same household budget. The available data do not allow their effects to be added borrower by borrower. Dallas also publishes a projection out to 2055 for additional potential mortgage delinquencies. We do not use it as a headline number because it relies on a premium path supplied by First Street, a private climate-risk modelling company. It is a conditional scenario, not an official forecast of defaults. Three states show pressure, not causation A 2026 Federal Reserve Bank of Philadelphia brief examines Pennsylvania, New Jersey and Delaware. From December 2021 to June 2025, it finds average real premium increases per loan of 28.9 percent, 26.1 percent and 25.4 percent respectively.[^philadelphia] In the three states, a 25 percent premium increase is associated with a 0.6 percentage point higher 30-day delinquency likelihood in Pennsylvania, 0.3 point in New Jersey and 0.7 point in Delaware. The authors explicitly state that this correlation cannot establish causation: a borrower already under pressure may also have a lower credit score and pay more for insurance.[^philadelphia] That caution matters. The two Federal Reserve Bank publications use ICE McDash data. They are complementary views, not two independent measurements of the same effect. Philadelphia extends the observed regional period to June 2025, but its brief alone cannot establish that insurance caused a regional change in arrears. The loss does not stop at the bank counter A missed payment does not yet identify the eventual loss bearer. Some mortgages remain on a bank balance sheet. Others are securitized into MBS, securities backed by mortgage loans. For conventional loans acquired or guaranteed by Fannie Mae, credit risk can then be shared with a private mortgage insurer or transferred to specialist investors. As of March 31, 2026, 46 percent of Fannie Mae's conventional single-family book had at least one form of credit enhancement or risk transfer. The maximum loss its credit-risk-transfer investors could absorb was $38 billion; its maximum potential recovery under mortgage insurance was about $200 billion.[^fannie-10q] These are not expected losses or blanket guarantees for every default. They describe a loss-sharing architecture. The economic path is therefore more complex than "the insurer raises the premium, the bank loses money." The homeowners insurer sets the cost of coverage. The household bears the increase first. If default follows, the loss depends on home value, loan-to-value, credit enhancement, recovery procedures and risk-transfer contracts. This distribution of exposure is the kind of mechanism examined in our article on the migration of credit risk. Current data do not describe a national mortgage crisis A risk channel is not a crisis diagnosis. In its May 2026 Financial Stability Report, the Federal Reserve says mortgage delinquency rates remain near the low end of their historical distribution and that very few homeowners have negative equity. The report's mortgage data run through December 2025.[^fsr] Fannie Mae reports a similar order of magnitude in its own perimeter. Its serious-delinquency rate for conventional single-family loans, defined as 90 days or more past due or in foreclosure, was 0.58 percent at March 31, 2026, unchanged from the prior quarter and near historical lows.[^fannie-10q] This book includes neither every US mortgage nor FHA and VA loans nor households without a mortgage. These data draw a clear boundary. It would be wrong to attribute national mortgage-default movements to homeowners insurance. It would also be wrong to conclude that the risk is absent because the aggregate remains calm. The channel shows up first among households with little budget room, then in particular places and credit profiles, long before it can be seen in a national rate. The data needed to measure the portfolio risk Moving from a plausible mechanism to portfolio analysis requires loan-level information on: - the premium change at renewal and its share of total monthly payment; - 30-day and 90-day delinquency by income, credit score, loan-to-value and geography; - cancellations, non-renewals and lender-placed coverage; - home value after a loss event, which determines loss severity after foreclosure; - the share actually absorbed by mortgage insurance and risk-transfer contracts. There is no public national database that joins all five. The US Treasury's insurance report provides a robust snapshot of the insurance market, with more than 246 million policies observed from 2018 to 2022, but not a mortgage-default file. It finds that average premiums per policy rose 8.7 percent faster than inflation over that period, with wide differences across ZIP Codes.[^fio] The signal is serious, but its scale remains to be measured. Home insurance becomes a credit variable because it changes payment capacity and can push constrained households toward arrears. It does not, at this stage, turn the US mortgage market into a new subprime crisis. --- Primary and institutional sources - Federal Reserve Bank of Dallas, "Home insurance premiums influence mortgage delinquencies, relocations," 24 March 2026: ICE McDash coverage, national premium trend, econometric estimate and projection limits. - Federal Reserve Bank of Philadelphia, "Homeowners insurance premiums and mortgage delinquency," 2026: regional data through June 2025, correlations and an explicit methodological caveat. - US Treasury, Federal Insurance Office, "Analyses of U.S. Homeowners Insurance Markets, 2018-2022," 2025: coverage, premiums, availability and the 246-million-policy scope. - Federal Reserve, Financial Stability Report, May 2026: aggregate mortgage delinquency and homeowner equity. - Fannie Mae, First Quarter 2026 Form 10-Q, pp. 23-28: serious delinquency, mortgage insurance and credit-risk transfer. - Fannie Mae, homeowners insurance coverage guide: insurance requirement and possible collection by the servicer. Reading limits. The 31,000-mortgage figure is a research estimate for 2022, not a count of defaults caused by insurance. The Dallas and Philadelphia series partly rely on the same data provider, ICE McDash. Treasury's report covers insurance policies from 2018 to 2022, not mortgages or defaults. Fannie Mae's 0.58 percent rate covers only its conventional book and uses a 90-day-or-foreclosure definition. Aggregate federal data do not isolate the share of 2026 mortgage arrears caused by insurance. Sources accessed on 5 August 2026. [^fannie-guide]: Fannie Mae, "Insurance coverage guide". [^dallas]: Federal Reserve Bank of Dallas, "Home insurance premiums influence mortgage delinquencies, relocations," 24 March 2026. [^philadelphia]: Federal Reserve Bank of Philadelphia, "Homeowners insurance premiums and mortgage delinquency," 2026, pp. 5-6. [^fannie-10q]: Fannie Mae, First Quarter 2026 Form 10-Q, pp. 23-28. [^fsr]: Federal Reserve, Financial Stability Report, May 2026, pp. 56-57. [^fio]: US Treasury, Federal Insurance Office, "Analyses of U.S. Homeowners Insurance Markets, 2018-2022," 2025. ============================================================================ ANALYSIS: When an Energy Rating Enters the Bank Balance Sheet URL: https://l0g.fr/en/analysis/energy-rating-enters-bank-balance-sheet/ Canonical French source: https://l0g.fr/posts/dpe-entre-bilan-banques/ Date: 2026-08-05 (reviewed 2026-08-05) Topics: housing, credit, climate, banks, europe ---------------------------------------------------------------------------- import InfographicBars from '../../components/infographic/InfographicBars.astro'; An energy certificate does not dictate a mortgage rate, and it cannot forecast the sale price of a home. But it is progressively entering bank credit judgement. In its July 2026 survey, the European Central Bank (ECB) found that euro-area banks had eased their internal credit standards for buildings that are already energy-efficient or will become so through the requested loan, while tightening them for persistently low-performing buildings. That is a real signal, with a narrow meaning: it records banks' internal approval criteria, not a public rate card or a map of mortgage rejections. A certificate is a legal document before it becomes a credit input France's diagnostic de performance énergétique (DPE) is a legally required energy-performance certificate that rates a home from A to G. Its registration number, issued after transmission to ADEME, is required for validity. Property listings for sale or rent must display the energy and climate labels and, where applicable, the excessive-energy-use warning. They must also show a range for a theoretical annual energy bill. The French ministry stresses that this estimate is based on standard use and cannot be compared with a household's actual bills, which vary with weather, occupancy and behaviour.[^dpe] That distinction is more than administrative. A DPE is not a household budget and it is not a credit score. It matters for lending because a property purchase combines repayment capacity, renovation cost, the ability to rent, the value of the property pledged as collateral and resale conditions. Energy performance can inform some of those inputs. It does not sum them up. French rules already make the information economically material. Rents on F and G homes have been frozen since August 2022. G-rated homes are progressively excluded from the rental market from 2025, F-rated homes from 2028 and E-rated homes from 2034, subject to the rules that apply to each case.[^dpe] That does not make a property unsellable or automatically unfinanceable. It adds a renovation and timing question for a landlord or buyer. The ECB measures a split in standards, not a split in rates The clearest source is the ECB's July 2026 bank lending survey, conducted from 15 to 30 June among 159 banks with a 100% response rate. It asks whether climate and energy performance affect banks' credit standards, meaning their internal guidelines or loan-approval criteria. They are not the contractual terms ultimately signed by every borrower.[^bls] For the twelve months from the third quarter of 2025 to the second quarter of 2026, the net balance of responding banks was -20 for buildings with good or high current or targeted energy performance. It was +14 for buildings with persistently low performance and no or limited planned improvement. In this survey, a negative number means that more banks reported easing than tightening; a positive number means the reverse. It is neither a 20% cut in mortgage rates nor a share of loans.[^bls] The ECB uses a Europe-wide classification that is broader than the French DPE alone. It defines good or high performance as an energy-performance certificate of A to C, current or targeted after the requested loan, generally for a new, modern or deeply renovated building. Low performance means D to G where the loan brings no material improvement. If no certificate is available, a bank may use the building's age, major renovation history or energy consumption as a proxy.[^bls] That matters for any French reading. F and G, often called thermal sieves, are in the ECB's low-performance group, but so are D and E. Conversely, a renovation project can place a building in the more favourable group before the work has been completed. The survey does not establish a French bank's rate differential by DPE class, and it does not prove that an F- or G-rated home is rejected. Loan demand points in the same direction, with the same survey limitation. The net balance of banks reporting a climate-related increase in mortgage demand was +28 for homes that are efficient or targeting good performance, and -15 for persistently low-performing homes. Banks expect +22 and -7 respectively over the following twelve months.[^bls] Physical and transition risk travel through different channels It would be too simple to reduce the finding to insulation. The ECB separates transition risk, linked to rules, energy costs and the investment needed to renovate, from physical risk, linked for example to weather events and the location of the property. The first can support lending for renovation; the second can weigh on repayment capacity or collateral value regardless of an energy label.[^bls] For household mortgage lending, physical risk was the largest climate-related tightening factor reported by banks, at +12 in net terms over the observed twelve months. Energy performance had a net easing effect of -5. Those figures do not offset each other home by home. They aggregate bank answers to different mechanisms. This caution also matters for readers looking for an effect on prices. In May, the ECB said that if credit-condition differentials widened and if they were reflected in property prices, collateral values would become more climate-sensitive.[^ecb-speech] That is a possible transmission mechanism, not an observed result for each local market. The data used here measure neither an average discount for poorly rated homes nor bank losses caused by a DPE class. Renovation is becoming a finance variable without yet solving affordability Europe's stock problem is large. The European Commission says 85% of EU buildings were built before 2000, 75% have poor energy performance and the annual renovation rate remains around 1%.[^epbd] Those figures describe the EU building stock, not the ECB mortgage sample. They show the gap between the stock that needs work and the annual capacity to renovate it. In the bank survey, investment in building energy performance is the main driver of climate-related mortgage demand, at +28 in net terms over the past twelve months. Lending rates aimed at raising real-estate sustainability (+17) and climate-related fiscal support (+9) also support this demand. Uncertainty over future climate rules is a drag (-6).[^bls] The tension is straightforward. A weakly performing home may need work to remain rentable, cheaper to occupy or easier to resell. But that initial investment increases financing needs for households that may not have extra borrowing capacity. The ECB observes credit direction, not the social affordability of the work. It does not report which household receives a grant, the remaining out-of-pocket cost or whether the renovation is actually completed. EU law does not resolve that uncertainty by itself. The revised Energy Performance of Buildings Directive had to be transposed by 29 May 2026. On 15 July, however, the Commission opened infringement procedures against all 27 member states for incomplete transposition, giving them two months to respond.[^epbd] The direction is European; the applicable rules, aid schemes and timing still vary nationally. A bank balance sheet starts with data, not a verdict on a home The most concrete shift is not yet a mortgage rate displayed at a branch. It is in bank measurement systems. In a 10 April 2026 consultation, the European Banking Authority (EBA) proposed collecting household mortgage exposures by real-estate energy-performance bucket. It calls the breakdown a relevant and proportionate proxy for transition vulnerabilities, and notes that efficiency standards, energy costs and renovation needs may affect both borrowers' repayment capacity and collateral value.[^eba] The status of the document matters: it is a consultation on supervisory reporting, not a final rule telling a bank to deny credit to a weakly rated home. Its significance lies elsewhere. Once energy performance is linked in reporting to exposures, impairments, collateral and risk parameters, it becomes a variable banks and supervisors can monitor across mortgage books. That new lens does not replace the traditional building blocks of bank analysis: capital, liquidity, credit losses and risk concentration, set out in our guide to reading bank health. It adds granularity to real-estate risk. Our earlier reading of the ECB credit survey places this signal in the wider movement of euro-area lending conditions. What the evidence establishes, and what it does not Three points are documented. First, euro-area banks already report treating buildings that are efficient or moving toward efficiency differently from persistently low-performing buildings. Second, physical risk is a distinct and more restrictive climate factor in net terms. Third, supervisors are preparing more granular measurement of those exposures. Three stronger claims are not supported by these data. They do not show that an F- or G-rated home causes a mortgage denial. They do not quantify a rate spread by energy class or a property-price discount. And they do not show that renovation is affordable for every household that needs it. The most concrete risk is therefore a gradual divergence. Households able to buy an efficient home or fund renovation may find credit easier to obtain, while those buying an older home with little room in their budget may have to carry the purchase price, renovation cost, energy expenditure and rental rules together. The ECB records early signs of that banking divergence. Its social scale, country by country and household by household, requires comparable loan, income, renovation and price data that are not yet publicly available. --- Primary and official sources - ECB, The euro area bank lending survey, second quarter 2026, 21 July 2026: Tables 19 and 20, definitions and survey method. - French Ministry for Ecological Transition, DPE: validity, listing requirements, theoretical cost estimate and the F/G timetable. - European Commission, Energy Performance of Buildings Directive and July 2026 enforcement action: building-stock figures, timetable and transposition status. - EBA, stress-testing module consultation, 10 April 2026, p. 6: energy-performance buckets in mortgage reporting, consultation status. Reading limits. The ECB survey aggregates bank opinions; it is not a loan-level file. Its net balances are not rates, volumes, default probabilities or rejection probabilities. The ECB energy categories may use a certificate, building age, renovation history or consumption, and cannot be reduced to the French DPE. ADEME's DPE data are useful for statistics, but ADEME says diagnosticians remain responsible for the quality of submitted records; a reweighting method is needed before using them as a national estimate.[^ademe] Data and legal texts checked on 5 August 2026. [^dpe]: French Ministry for Ecological Transition, DPE. [^bls]: ECB, The euro area bank lending survey - Second quarter of 2026, section 5.5, Tables 19 and 20, methodological notes. [^ecb-speech]: ECB, "Climate change and monetary policy", 5 May 2026. This article treats the relationship as the ECB's conditional hypothesis. [^epbd]: European Commission, Energy Performance of Buildings Directive; 15 July 2026 transposition action. [^eba]: EBA, Consultation Paper, Module on Stress testing, 10 April 2026, p. 6. [^ademe]: ADEME, DPE data warning. ============================================================================ ANALYSIS: The Golden Age of Buy Now, Pay Later URL: https://l0g.fr/en/analysis/the-golden-age-of-buy-now-pay-later/ Canonical French source: https://l0g.fr/posts/golden-age-buy-now-pay-later/ Date: 2026-08-04 (reviewed 2026-08-04) Topics: united-states, consumer, credit, bnpl, regulation, trump ---------------------------------------------------------------------------- The Golden Age of Buy Now, Pay Later is real, but the question is who benefits. In the first quarter of 2026, Klarna, Affirm and Sezzle reported volume growth above 30%. Across the US adult population, however, the share using BNPL rose by just one point, to 16%. The alarm is not a general rush into pay-in-four. It is more serious: card debt is rising mainly among people who already say they cannot make ends meet, while a growing share of workers taps retirement savings to avoid eviction or pay medical bills. Donald Trump created neither this divide nor BNPL. But his tariffs raised consumer prices and his administration reduced federal protection for instalment borrowers. He is not the sole cause of the household budget crisis. He is not irrelevant to it either. Card debt is changing in nature An American credit card can be a simple payment method. If the holder pays the full statement balance, there is generally no interest. But the holder can also carry part of the balance into the next month. This is revolving debt: it remains open, renews and incurs interest. In the first quarter of 2026, the average rate charged to accounts that actually paid interest was 21.52%, according to the Federal Reserve. At that price, debt used to finish the month can quickly feed on itself. The Fed's latest survey shows exactly who owns the additional debt. With participants' consent, researchers matched survey responses to anonymised credit records. They compare the same people's card balances in 2023 and 2025, then group them by how they described their financial situation in 2025. Among people who said they were finding it difficult to get by, the average balance rose from $6,735 to $9,265, an increase of $2,530 or 37%. Among those living comfortably, it rose by only $59, from $6,248 to $6,307, or 1%. In between, those who were “just getting by” saw their average balance climb from $6,782 to $8,581. That 65% is the central number. Rising balances no longer mainly describe broadly distributed, buoyant consumption. They are concentrating among households whose income does not comfortably cover expenses. Those living comfortably, by contrast, now explain just 3% of the increase, down from about one-quarter in the previous two surveys. The scope matters. The calculation covers the 63% of respondents who allowed their answers to be matched with credit records. The 2023 balances belong to the same people surveyed in 2025. It therefore does not directly measure the rise in all US card debt. But it answers the most important question: among the people followed, additional borrowing shifted sharply toward already fragile budgets. $1.252 trillion, with interest above 21% The national view confirms the weight of the problem. At the end of the first quarter of 2026, card balances stood at $1.252 trillion, $70 billion higher than a year earlier. The $25 billion decline from the Christmas quarter was seasonal and does not erase the annual increase. The same release says 7.10% of balances were moving into serious delinquency at an annualised rate, compared with 7.04% a year earlier. This does not mean 7.10% of all cards are already in default. The New York Fed starts with balances that were current, or less than 90 days late in the previous quarter, and measures the share that has just crossed 90 days past due. It then annualises that transition. The pace is nearly stable year on year, but remains high for debt charged at an average 21.52% on interest-bearing accounts. In May 2026, another New York Fed survey offered a signal closer to daily life. The average reported probability of missing a minimum debt payment in the next three months rose to 12.6%. The increase was driven mostly by households earning less than $100,000 and people with no education beyond high school. Respondents expected their income to grow 2.8%, but food prices by 5.8% and rents by 7.4%. These are expectations, not bills already paid. Their gap nonetheless explains why credit increasingly serves as a bridge between paycheques. The 401(k) is becoming an emergency fund The 401(k) is the main employer-sponsored retirement savings plan in the United States. Workers put part of their wages into it, often receive an employer contribution and gain a tax advantage. The money is intended to remain invested until retirement. A hardship withdrawal lets a worker take money out early for what the IRS calls an “immediate and heavy financial need.” It may pay certain medical bills, prevent eviction or foreclosure, cover funeral costs, repair a primary residence or meet certain education expenses. It is not a loan. The IRS explains that the money is not repaid to the plan, permanently reduces the account and may face tax and a penalty. Vanguard found that 6% of participants permitted to make such a withdrawal took at least one in 2025, up from 5% in 2024 and 2% in 2020. The median amount was $1,900. More importantly, 46% of those participants withdrew money more than once during the year and 21% did so at least three times. This is no longer only an exceptional bill: for nearly half of the workers concerned, the retirement plan is functioning as emergency savings. It would be wrong to attribute the entire rise to distress. Vanguard also says rules were relaxed and administration simplified. More lower-income workers are automatically enrolled in a plan, so more can access it. Fidelity still found in the first quarter of 2026 a record combined 401(k) saving rate of 14.4%, including employers. But that strong average can coexist with repeated withdrawals. Once again, it describes two different groups: those building wealth and those liquidating the future to pay for the present. BNPL is growing, but its user count is not exploding Buy Now, Pay Later generally splits a purchase into four instalments, often with no interest if each one is paid on time. In 2025, 16% of US adults said they had used it in the previous 12 months, compared with 15% in the previous survey and 10% in 2021. A one-point annual increase is not an explosion. Users' financial fragility is visible, however. Among them, 26% had paid at least one instalment late and 11% incurred an overdraft or nonsufficient-funds fee triggered by a BNPL debit. For 29%, BNPL was the only way to afford the purchase. That answer rose to 40% below $25,000 in family income. One user in five had financed groceries or food delivery; within that group, 45% said they could not otherwise afford the purchase. Platforms are accelerating faster than the population Company accounts explain the article's title. In the quarter ended March 31, 2026, Affirm processed $11.6 billion of GMV, up 35%. Klarna reported $33.7 billion, up 33% globally and 39% in the United States. Sezzle reached $1.1 billion, up 37.3%. GMV means gross merchandise volume. It is the amount spent through a platform during a period. It is not the platform's revenue, profit or the debt customers still owe. The three companies do not use exactly the same scope, so their GMV cannot be added together to obtain the size of the BNPL market. The gap with the 16% user share has a simple explanation. Platforms add merchants and products, while existing customers purchase more often. At Affirm, active customers increased 22%, while transactions per customer rose 20%. At Sezzle, average quarterly frequency rose from 6.1 to 7.1 purchases. Volume growth therefore comes not only from new users, but also from more intensive use. Adobe's 2026 recap confirms that “explosion” is the wrong word for the entire market. Adobe is not selling Photoshop here: Adobe Analytics measures purchases on retail websites that use its service. During the four days around Prime Day, June 23 to 26, it observed $2.1 billion in BNPL purchases, up 9.5%. Total online commerce rose 9.3%. During this major event, BNPL therefore grew almost in line with sales, far below the 30% to 37% reported by platforms. BNPL debt remains partly invisible Cards appear in national credit files. Pay-in-four is far less visible there. The CFPB's latest detailed report covers six large providers through 2023. It calculates an average $848 in annual loans per user per provider, 14% higher than in 2022 after inflation. That $848 is not the balance due on a given day. More importantly, it is not one person's total BNPL debt. A provider may not see purchases financed by competitors. The same borrower can therefore stack several payment schedules without any platform seeing all upcoming debits. Because loans are short, they can also disappear quickly from balance sheets while absorbing a large share of the next few paycheques. This burden per user, more than the number of users, is the reason for concern. The available buffer is thin. According to the BEA, the personal saving rate was only 2.7% of disposable income in June 2026. The average does not describe every household because saving is concentrated among the well-off. The l0g US Macro Dashboard tracks this PSAVERT series over time. Reading its decline alongside card debt, 401(k) withdrawals and BNPL late payments shows how little protection remains at the bottom of the distribution. Trump did not create the divide, but he is making it worse The timeline rules out blaming Donald Trump for the whole situation. The CFPB estimates that BNPL loans from six large providers had already risen from $2.7 billion to $45.2 billion between 2019 and 2023, in 2024 dollars. The card comparison begins in 2023 and therefore includes almost two years of the Biden presidency. High interest rates, cumulative post-pandemic inflation, rents, medical costs and easier access to 401(k) funds all predate January 20, 2025. But no single cause does not mean no responsibility. On April 2, 2025, Trump signed Executive Order 14257, imposing an additional 10% duty on a broad share of imports and higher rates on some countries. The details then changed repeatedly, but the policy choice and its author are unambiguous. In April 2026, Fed economists estimated that tariffs implemented in 2025 had raised core goods prices by 3.1% through February 2026 and lifted the broader core price level by 0.8%. Another Fed study published in June, using actual transactions, found in categories with average exposure prices 1% to 2% higher and spending roughly 4% lower. It concludes that the burden is proportionately heavier on low-income households. Part of the price increase since 2025 therefore comes from a policy chosen by Trump. This does not prove that a specific dollar of card debt or a 401(k) withdrawal was caused by a tariff. It proves that the administration added measurable pressure to the budgets of households already using credit to live. The second choice concerns consumer protection. In May 2024, the Biden CFPB had treated certain BNPL lenders as card issuers for statements and billing disputes. On May 12, 2025, under Trump, the Bureau withdrew that interpretation and cited presidential deregulation directives. Federal and state law did not disappear. But the signal is clear: as usage intensifies and one-quarter of users pays late, federal oversight has retreated. The risk is social before it is systemic The data do not yet describe a financial crisis comparable to 2008. The flow of card balances into serious delinquency is nearly stable year on year. Hardship withdrawals affect a minority of participants. BNPL use rose only one point. Wealthier households continue to save and spend, supporting the aggregates. The danger appears when we follow one family rather than the national average. A routine expense is split into instalments. A BNPL debit empties the account and triggers an overdraft. The card pays for the next month at more than 21%. A 401(k) withdrawal covers the emergency but destroys part of future retirement wealth. If income falls or a job disappears, no buffer remains. The risk is not only reduced consumption. It is the conversion of a temporary income shortfall into expensive debt and then a lasting loss of wealth. This is also why the average US consumer does not exist. Comfortable households keep the averages healthy while distress concentrates elsewhere. “Golden Age” accurately describes the BNPL industry. For fragile households, 2026 looks more like a golden age for products that postpone the bill by two weeks, one month or until retirement. --- Sources - Federal Reserve, Economic Well-Being of U.S. Households in 2025, Credit chapter, May 2026: matched 2023-2025 card balances, concentration of growth, BNPL use and late payments. - Federal Reserve Board, Consumer Credit G.19, July 8, 2026: average card rates and revolving credit. - New York Fed, Household Debt and Credit, Q1 2026, May 12, 2026: $1.252 trillion in card balances and a 7.10% annualised flow into 90-day delinquency. - New York Fed, Survey of Consumer Expectations, May 2026, June 8, 2026: perceived missed-payment risk and expected income, food and rent growth. - Vanguard, How America withstands financial hardships, March 2026: hardship withdrawals, frequency, income and reasons. - IRS, rules and consequences of 401(k) hardship distributions, updated February 26, 2026. - Fidelity, Q1 2026 Retirement Analysis: saving rates and plan coverage. - CFPB, The Buy Now, Pay Later Market, December 2025: six lenders, loans per user, 2019-2023 growth and visibility limits. - Federal Register, withdrawal of the BNPL interpretive rule, May 12, 2025. - White House, Executive Order 14257 on reciprocal tariffs, April 2, 2025. - Federal Reserve Board, Detecting Tariff Effects on Consumer Prices in Real Time, Part II, April 8, 2026, and Paying More and Buying Less, June 2026: estimated effects of tariffs on prices and spending. - BEA, Personal Income and Outlays, June 2026, July 30, 2026: 2.7% PSAVERT personal saving rate. - Affirm, 10-Q for March 31, 2026, Klarna, Q1 2026 results, Sezzle, Q1 2026 results filed with the SEC: GMV, users and frequency. - Adobe Analytics, Prime Day 2026: observed online purchases from June 23 to 26, 2026. Limitations Data cut off on August 4, 2026. The SHED surveys adults, and its card table covers respondents who allowed data matching. Vanguard and Fidelity describe their own participants, not every US worker. The CFPB's detailed user-level BNPL data end in 2023. Adobe covers only online purchases observed among its clients. Affirm, Klarna and Sezzle GMV combine different products, countries and accounting rules. The Fed studies identify an effect of tariffs on prices, not a precise share of card debt, BNPL late payments or 401(k) withdrawals. The article therefore separates measured effects, temporal overlap and links that remain unknown. ============================================================================ ANALYSIS: How American Is America's AI Boom? URL: https://l0g.fr/en/analysis/how-american-is-americas-ai-boom/ Canonical French source: https://l0g.fr/posts/boom-americain-ia-vraiment-americain/ Date: 2026-08-04 (reviewed 2026-08-04) Topics: ai, united-states, trade, investment ---------------------------------------------------------------------------- The boom is American in its centre of gravity: capital, hyperscalers, software, data centres and final demand. Its industrial chain is far less American. According to Federal Reserve staff estimates, roughly 90% of high-tech equipment used in the United States is sourced abroad. This dependence does not make investment illusory. It requires three figures that are often conflated to be kept separate: spending committed, output produced on US territory and value captured by US firms. Capex is not a measure of US production The Bureau of Economic Analysis provided the latest clue in its advance estimate for the second quarter. Investment rose, led by equipment and intellectual-property products. Within equipment, the BEA reported broad increases, including information-processing hardware. It also specified that the estimate was based primarily on import data. Capital-goods imports were led by telecommunications equipment, semiconductors and industrial equipment. This composition helps explain the gap between record capex and its domestic contribution. A US company may buy a server assembled in Mexico with processors fabricated in Taiwan and South Korean memory, install it in Virginia, then sell a global software service. The full bill enters the company's investment. Only the stages produced in the United States enter US GDP. The accounting does not say that imports destroy demand or mechanically "subtract growth". In the identity C + I + G + X - M, consumption and investment already contain imported goods. The BEA subtracts imports to avoid crediting the United States with production performed elsewhere. An import can therefore signal strong demand while widening the gap between gross investment and domestic output. Our balance-of-payments guide develops the other side of the relationship: financing the external deficit. The external adjustment changes the result Because the national accounts contain no AI line item, three Fed economists built a proxy combining software, data centres, power facilities, computers and peripherals. They then apply an adjustment based on net exports of computers, peripherals and parts. Their method is not an official AI statistic. It reveals what disappears when analysis stops at gross spending. In the first quarter of 2026, the proxy's four investment components together added 1.18 percentage points to annualised quarterly growth. Associated net exports subtracted 0.45 point, leaving a 0.73-point net effect. In the fourth quarter of 2025, the external adjustment absorbed 0.61 point of a 0.75-point gross contribution, leaving only 0.14 point net. The calculations come from the data file attached to the Fed note. @media (max-width: 760px) { figure.ai-origin-scroll { width: 100%; max-width: 100%; min-width: 0; overflow-x: auto; contain: inline-size; } figure.ai-origin-scroll svg { width: 820px !important; max-width: none; } } A supply chain with several geographies The answer depends on which layer is being observed. Demand and financing are largely American. The major platforms decide the investments, order computing capacity and bear the financial risk. US territory also hosts the buildings, grid connections, installation, part of the maintenance and associated jobs. This layer is why data-centre construction and software spending do create domestic value added. Hardware is far more international. The 14 July Fed note estimates that roughly 90% of high-tech equipment is sourced abroad. Another Fed note on AI trade traces the surge in exports of servers, graphics cards and parts notably to Taiwan, Mexico and Vietnam. It uses three customs codes, 8471.50, 8471.80 and 8473.30. Their precision is imperfect: some AI hardware is missing, while some included products serve other uses. Software and revenue can flow back to the United States. Foreign production of a server does not prevent a US company from later capturing cloud, software or intellectual-property margins. Conversely, US incorporation does not automatically turn all worldwide sales into US production. The BEA explains that value added excludes intermediate inputs and remunerates capital and labour. Its industry accounts show that information-sector capital growth was among the leading contributors to real GDP growth from 2021 to 2024, a result consistent with an AI effect but unable to isolate it. A transfer of demand, not a total loss Importing equipment does not mean that all project value leaves the United States. Land, the building, part of the power grid, installation, software, operations and financing may be domestic. Nor does it mean that foreign suppliers capture the full final price: a value chain distributes revenue, wages and profits across several jurisdictions. The macro point is narrower. The boom immediately stimulates demand for foreign capital goods before any productivity gains diffuse. Using earlier technology episodes, the 14 July Fed note estimates that an investment shock can widen the current-account deficit by about 0.2 percentage point of GDP for several quarters. This is a model response, not a measurement of the deficit caused by AI in 2026. The authors also find stronger spillovers to Taiwan, Korea and Mexico than their aggregate trade links alone would suggest. This asymmetry may persist: the United States mainly imports capital-intensive goods and exports more knowledge-intensive services. Software revenue may offset part of the hardware bill, but not necessarily at the same time or in the same account. The semiconductor cycle and the power constraint on data centres therefore become macro variables, not merely technology topics. The answer in three propositions The boom is American by decision: capital, risk, demand and a major share of software assets are concentrated in the United States. It is international by hardware production: high-tech equipment depends heavily on Asian and Mexican supply chains. The net-export adjustment shows that this layer materially reduces gross capex's contribution to US GDP in some quarters. It remains unsettled by future outcomes: productivity gains cannot be inferred from spending. Our review of the available evidence on AI and productivity finds task-level gains but incomplete macroeconomic diffusion. The debt financing the infrastructure adds a second test: who retains the return if revenue arrives more slowly than liabilities mature? A falsifiable diagnosis The thesis of a US boom with high import content would weaken under three observable developments: sustained growth in US semiconductor and electronic-equipment production faster than capex, a decline in the net import share of high-tech equipment and faster US computer-service exports sufficient to offset hardware. It would strengthen if investment, equipment imports and the current-account deficit rose together while domestic industrial production plateaued. The next step is therefore not to total capex announcements. It is to track production, net imports, industry value added and productivity at the same time. Sources 1. Bureau of Economic Analysis, GDP, Advance Estimate, Second Quarter 2026, 30 July 2026. Investment, equipment, intellectual property and import composition. 2. Bureau of Economic Analysis, The Expenditures Approach to Measuring GDP, 3 June 2025. Accounting treatment of imports in GDP. 3. Federal Reserve, Soto, Thieu and Allen, The AI Buildout and the Economy, 17 July 2026, and data file. Growth-contribution proxy and methodological limits. 4. Federal Reserve, Fiori, Lipa and Nuenninghoff, Technology Shocks, the AI Boom, and the U.S. Current Account, 14 July 2026. Import share, historical model and international spillovers. 5. Federal Reserve, de Soyres, Haag, Liu and Van Leemput, The Global Trade Effects of the AI Infrastructure Boom, 13 February 2026. Customs codes, supplier economies and measurement limits. 6. Bureau of Economic Analysis, How Does AI Drive Growth?, 8 June 2026. KLEMS framework, value added and information-sector capital. Limitations There is no isolated AI sector or exhaustive AI-capex series in the national accounts. The Fed proxy components include non-AI spending. Its external adjustment assigns goods between investment and consumption using a weight because BEA accounts cannot directly trace the destination of each import. The 90% estimate concerns high-tech equipment as defined by the authors, not an entire data centre. FEDS Notes express their authors' analysis, not an official position of the Board. Data cut off at 10:35 CEST on 4 August 2026. The June 2026 US trade report, scheduled for 14:30 CEST, is not yet public and is not incorporated into this analysis. This is not investment advice. ============================================================================ ANALYSIS: Do strip clubs really predict recessions? URL: https://l0g.fr/en/analysis/do-strip-clubs-really-predict-recessions/ Canonical French source: https://l0g.fr/posts/strip-clubs-annoncent-ils-recessions/ Date: 2026-08-02 (reviewed 2026-08-02) Topics: recession, economic indicators, consumption, employment, united states ---------------------------------------------------------------------------- No, strip clubs do not possess a secret recession detector. The intuition behind the "Stripper Index" is nonetheless reasonable: optional leisure spending may fall as soon as customers feel poorer. A Texas study even confirms that club revenues moved with the local economy before and during the 2008 crisis. But no work we found shows that dancers' tips or club attendance consistently lead US recessions. The strongest signal in this odd selection comes from somewhere with far less neon: temporary staffing agencies. A sound intuition, but no public national series found The "Stripper Index" spread widely in May 2022. A dancer using the name Botticelli Bimbo said on Twitter that clubs were a leading indicator and that the United States was already in recession. KQED traced the origin of the meme. A few weeks later, ABC7 interviewed dancers and a California agency reporting empty clubs and bookings far below pre-pandemic levels. The mechanism is economically unsurprising. A night at a club is discretionary spending. It can be cancelled immediately, unlike rent or an electricity bill. Workers paid in tips can observe that retreat without waiting for a quarterly report. Their experience may therefore provide early, local information. The essentials required for an index are missing. There is no shared definition of the tip being observed, no stable sample of clubs, no adjustment for seasons, tourism, openings or closures, and no denominator separating the number of customers from their average spend. A decline may come from a weaker economy, a new competitor, a change in payment methods or different nightlife habits. The viral 2022 prediction illustrates the problem of false positives. The NBER chronology currently records no US recession after the one from February to April 2020. The NBER does not reduce a recession to two quarters of negative GDP. It looks for a significant, broad and lasting decline across the economy. This does not mean the 2022 testimonies were false. They may have accurately described struggling businesses and incomes. They were simply insufficient to diagnose a national recession. Texas measures the cycle, not the lead The strongest source we found on clubs validates their sensitivity to the economy, not their forecasting power. In 2009, a team at the University of Texas at Austin delivered a 202-page study of the adult entertainment industry to the state legislature. It reconstructed quarterly revenues from 2005 through the third quarter of 2008, including mixed-beverage sales for 123 clubs. Aggregate revenues moved closely with personal income, gross state product, a coincident business-cycle index and nonfarm payrolls. They fell as unemployment rose. The authors explicitly warn that the correlations do not establish causation. More importantly, their test compares series in the same quarter. It does not establish whether club revenues turn three, six or twelve months before the economy. @media (max-width: 760px) { figure.odd-indicator-scroll { width: 100%; max-width: 100%; min-width: 0; overflow-x: auto; contain: inline-size; } figure.odd-indicator-scroll svg { width: 760px !important; max-width: none; } } Since 2008, Texas has collected an even more direct measure. Covered businesses must record admissions every day and pay a fee per customer. The Comptroller still requires those records. The system could support a serious analysis if aggregate attendance, changing coverage and seasonal factors were published as one consistent series. It would still measure neither tips nor workers' income. Underwear survives only slightly better The Men's Underwear Index rests on an idea attributed to Alan Greenspan: men's underwear sales normally change very little, then fall when households become constrained enough to postpone even this private purchase. The line became famous through journalist Robert Krulwich's account, later retold by NPR. Unlike club tips, underwear has been subjected to a published test. A 2012 study covering 57 countries found limited evidence of a relationship in the United States, but called it unclear and recommended extreme caution. Across the other 56 countries, underwear sales appeared unrelated to the aggregate economy. The reasonable conclusion is neither "Greenspan was right" nor "underwear says nothing." A deferrable purchase may reveal pressure on household budgets. But a fragile relationship in one country, based on commercial data that are difficult to reproduce freely, is not a universal economic gauge. Cardboard suffers from a price-volume mix-up The cardboard mechanism sounds stronger still. Consumer goods, industrial parts and parcels travel in boxes. Falling shipments could therefore foreshadow weaker production and trade. The trap lies in the chosen series. FRED publishes a monthly BLS index for corrugated boxes, but it measures producer prices, not the number of boxes shipped. It may rise because of paper, energy or margins with no increase in volume. The public sectoral output series does measure activity, but it is annual and ends in 2021. It cannot identify a 2026 turning point in real time. Professional shipment data may be useful to subscribers. For a reader trying to reproduce the signal for free, the most visible official series mainly presents a classic risk: mistaking price for quantity. Temporary staffing wins the unlikely-signal contest Temporary help is the only candidate here with a clear mechanism, a monthly series and a history of turning points. A company can eliminate a temporary position before dismissing a permanent employee. It can also bring in temps before committing to lasting hires. The BLS found that temporary-help employment declines preceded those in the broader labor market by six to twelve months during the 1990-91, 2001 and 2007-09 recessions. Before the Great Recession, temp employment peaked in December 2006, a full year before total nonfarm employment. The series is far from infallible. The BLS counted a loss of 624,000 jobs between the March 2022 peak and December 2024, yet no new recession has since appeared in the NBER chronology. Changes in recruitment, the exit from the pandemic and substitution between types of contracts can alter the historical relationship. Recent data do not create a fresh alarm on their own. The monthly BLS series carried by FRED recorded 2.499 million jobs in June 2026, 9,300 more than in May and 27,300 more than in February. These figures are revisable and cover too short a period to establish a trend. They show why any indicator must be read over several months and alongside the rest of the US employment report. Four checks for any strange indicator An alternative signal deserves attention if it passes four simple checks. 1. A stable definition. Are we counting admissions, dollars spent, units sold or prices? 2. Useful frequency. Annual data released two years late cannot predict the next quarter. 3. A demonstrated lead. The series must turn before activity across several cycles, not merely fall during a known crisis. 4. False positives. Alerts not followed by recession must be counted alongside successes remembered after the fact. The verdict is straightforward. Dancers may detect a genuine decline in discretionary spending very early, but their observations do not form a verifiable national index. Underwear has one study with fragile results. Cardboard says something about the movement of goods, provided price is not confused with quantity. Temporary help has the best historical record, without being an oracle. Among neon lights, boxer shorts, pallets and short-term contracts, the most serious candidate is therefore the least glamorous. In economics, that is often a good sign. Sources 1. NBER, Business Cycle Dating procedure and FAQ: recession definition, indicators considered and latest available chronology. 2. KQED, "Recession Indicator Memes Are Getting Too Real", 2025: documented origin of the viral 2022 "Stripper Index." 3. ABC7, "Are strip clubs a good predictor of a recession?", 30 June 2022: testimony from workers and data from a California agency. 4. Busch-Armendariz et al., University of Texas at Austin, "An Assessment of the Adult Entertainment Industry in Texas", 2009: revenue, sample, seasonality, correlations and limitations. 5. Texas Comptroller, Sexually Oriented Business Fee FAQ: current daily admissions record and $10-per-entry fee requirements. 6. Phil Smith, "Do Sales of Men's Underwear Really Predict the State of the Economy?", International Journal of Technology, 2012: test across 57 countries and cautious conclusion. 7. BLS via FRED, monthly corrugated-box prices and annual sectoral output: price, volume and data-freshness distinction. 8. BLS, "What happened to temps?", 2021, and 2024 temporary-help employment review: mechanism, historical lead and decline since 2022. 9. BLS via FRED, All Employees, Temporary Help Services: seasonally adjusted monthly series and latest June 2026 observation. Limitations This research cannot prove that no private or local database on tips exists. The Texas study covers a short period, one state and establishment revenue rather than dancers' income. Historical relationships can change with digital payments, independent work, online commerce and recruitment practices. None of these indicators can date or predict a recession on its own. Data cut off on 2 August 2026. This is not investment advice. ============================================================================ ANALYSIS: US GDP: 1.5% on the surface, 3.9% in private demand URL: https://l0g.fr/en/analysis/us-gdp-private-demand-q2-2026/ Canonical French source: https://l0g.fr/posts/pib-americain-demande-privee-t2-2026/ Date: 2026-08-02 (reviewed 2026-08-02) Topics: united-states, gdp, inflation, fed ---------------------------------------------------------------------------- The same official release offers two nearly opposite readings of the US economy. Real GDP grew by only 1.5% in the second quarter of 2026. Yet private domestic demand, restricted to consumption and private fixed investment, accelerated to 3.9%. There is no contradiction. The first figure measures total domestic production, including trade, inventories and government. The second isolates the private domestic engine. The gap shows why a GDP headline alone can tell the wrong economic story. Two speeds in one release On 30 July, the Bureau of Economic Analysis reported real growth of 1.5% at a seasonally adjusted annual rate, down from 2.1% in the first quarter. Without annualisation, the second-quarter increase was 0.4%. The US convention scales up the pace observed over three months as if it continued for a year. It does not mean GDP has already gained 1.5% since March. In the same release, real final sales to private domestic purchasers rose from 1.7% in the first quarter to 3.9% in the second. The BEA defines the measure as consumer spending plus gross private fixed investment. It therefore excludes inventory changes, government spending and net exports. The contrast extends to prices. The gross domestic purchases price index accelerated from 3.6% to 5.7%. The headline PCE price index moved from 4.6% to 5.1%, while core PCE slowed from 4.4% to 3.4%. All four measures below use the same annualised quarterly convention. @media (max-width: 760px) { figure.gdp-scroll { width: 100%; max-width: 100%; min-width: 0; overflow-x: auto; contain: inline-size; } figure.gdp-scroll svg { width: 760px !important; max-width: none; } } GDP accounting creates the gap GDP answers a precise question: how much value was produced inside the United States? An import is purchased in the United States but produced somewhere else. It is therefore subtracted so that foreign output is not attributed to the US economy. That accounting subtraction does not, by itself, mean American demand contracted. In the second quarter, the BEA recorded faster consumer spending, a decline in government spending, and slower investment and exports. Imports increased more than in the previous quarter. Within investment, equipment and intellectual property products increased, while private inventories and non-residential structures declined. The 1.5% figure combines all these movements. The 3.9% measure deliberately removes the components most likely to blur final private demand. Solid consumption, a thin cushion The BEA's monthly release completes the picture. In June, nominal personal income and disposable income each rose 0.2%, while consumer spending increased 0.3%. Adjusted for prices, spending gained another 0.4%. Consumption therefore does not yet show the drop implied by a superficial reading of 1.5% GDP. The fragility lies in how that resilience is financed. The personal saving rate was only 2.7% in June. This aggregate ratio proves neither imminent exhaustion nor generalised over-indebtedness. It only shows that a small share of disposable income remains as a buffer. Distribution matters too. Our analysis of the K-shaped US consumer explains why a resilient average can coexist with households already under pressure. The Fed does not receive a recession signal One day before the GDP release, the FOMC kept its target range at 3.50% to 3.75% by a nine-to-three vote. All three dissenters preferred a 25-basis-point increase. The statement described economic activity as expanding at a solid pace and inflation as remaining above the 2% objective. The figures published the next day cannot establish that the Fed would have acted differently had they been available. They do reinforce the difficulty described in our analysis of the July FOMC: aggregate growth is modest, but private demand does not provide the contraction signal that would justify easing on its own. Keeping time horizons separate prevents another error. In the second quarter, headline PCE prices increased 5.1% at an annualised rate and core prices rose 3.4%. In June alone, headline PCE fell 0.1% and core PCE increased 0.1%. Over twelve months, they were still up 3.7% and 3.3%. Placing these rates side by side without their time periods creates a statistical disagreement that does not exist. A falsifiable diagnosis The 1.5% figure does not signal an ongoing recession because consumption and private fixed investment accelerated. The 3.9% measure does not guarantee a durable expansion either, because saving is low and prices remain elevated. The restrained reading is an economy whose private engine remains robust but whose household cushion and disinflation are still fragile. Three releases will test that reading. The July employment report arrives on 7 August. The BEA will publish its second GDP estimate and the July income and spending accounts on 26 August. The annual update of the national accounts begins on 30 September. A material revision to consumption or private fixed investment would directly change the diagnosis. Sources 1. Bureau of Economic Analysis, GDP, Advance Estimate, Second Quarter 2026, 30 July 2026. GDP, private final sales, price indexes, composition and revision calendar. 2. Bureau of Economic Analysis, Personal Income and Outlays, June 2026, 30 July 2026. Income, real consumption, PCE inflation and the saving rate. 3. Federal Reserve, FOMC statement, 29 July 2026. Target range, vote and assessment of activity and inflation. 4. Federal Reserve, Monetary Policy Report, July 2026, 10 July 2026. Macroeconomic context and inflation risks before the July meeting. Limitations The second-quarter figure is an advance estimate based partly on assumptions for source data that remain incomplete. Quarterly rates are annualised, while monthly and twelve-month rates cover different horizons. Private final sales better isolate domestic demand, but they measure neither its distribution across households nor its financial sustainability. Data cut off on 2 August 2026. This is not investment advice. ============================================================================ ANALYSIS: Yen: 163,412 net short contracts before the shock URL: https://l0g.fr/en/analysis/yen-163412-net-short-contracts-before-shock/ Canonical French source: https://l0g.fr/posts/yen-163412-contrats-vendeurs-avant-choc/ Date: 2026-08-02 (reviewed 2026-08-02) Topics: yen, carry trade, cftc, cot, boj, mof, positioning, markets ---------------------------------------------------------------------------- The chart was published after the shock, but its data stop before it. The latest available Commitments of Traders report describes positions held at the close on Tuesday, 28 July. It therefore cannot yet see the foreign-exchange moves of 30 and 31 July. Its value lies elsewhere: it measures the quantity of short-yen bets already in place before the market was hit. A snapshot taken before the move The official CFTC row for the Chicago Mercantile Exchange yen future, contract code 097741, records 101,271 long contracts and 264,683 short contracts among non-commercial traders on 28 July. The difference is a net position of -163,412 contracts. The figure was released on Friday, 31 July, but it remains dated Tuesday. The CFTC states that the COT is generally released on Friday using positions from the preceding Tuesday. Publishing a number after an event does not turn it into a post-event measurement. That distinction is decisive this week. Reuters reported that the Japanese government intervened in the market on 30 July, according to market participants. The Ministry of Finance register, however, reports zero intervention between 29 June and 29 July and does not yet cover 30 July. The official total for the period from 30 July to 26 August is due on 28 August, according to the MoF calendar. Until then, the 30 July operation must remain attributed to Reuters and its market sources. The institutional distinction matters too: the MoF decides Japanese foreign-exchange interventions and the BoJ executes them as its agent. The central bank explains this division in its own overview of intervention operations. It also kept its policy rate at 1.00% on 31 July in its latest monetary policy decision. @media (max-width: 760px) { figure.yen-scroll { width: 100%; max-width: 100%; min-width: 0; overflow-x: auto; contain: inline-size; } figure.yen-scroll svg { width: 720px !important; max-width: none; } } Twelve billion dollars, but not twelve billion of global carry Each standard contract represents 12.5 million yen. The ECB reference rates for 31 July give 184.03 yen and 1.1485 dollars per euro, producing a calculated USD/JPY rate of 160.235. Applied to the 163,412 net contracts, the conversion gives about $12.75 billion: 163,412 × ¥12,500,000 ÷ 160.235 = $12.75bn The calculation is correct, but its perimeter must remain visible. This is the net notional of the relevant yen futures, not the size of the global yen carry trade. The series comes from the Legacy Futures Only report. It excludes, among other things, over-the-counter FX forwards and swaps, bank yen funding and foreign assets bought with that funding. The trader label requires the same care. The Legacy report's “non-commercial” category is not identical to hedge funds, CTAs or macro funds alone. The CFTC explains that categories are based on traders' reported primary business purpose and that it does not know the specific reason for each position. In the more detailed Traders in Financial Futures report, Leveraged Funds were net short 101,990 contracts on 28 July. That is a narrower population than the Legacy report's -163,412. The rigorous description is therefore: the net position of non-commercial traders in CME yen futures. It reveals a market loaded against the yen, not the precise identity of every carry holder. Price and positioning answer different questions Looking only at USD/JPY shows the aggregate result of currency buying and selling. Spot reacts quickly and incorporates new information immediately. Its perimeter is far wider than one Chicago-listed contract: the BIS 2025 Triennial Survey covers spot, forwards, swaps and FX options, and places the yen on one side of 16.8% of global turnover. Price still does not reveal how many futures positions were already crowded on the same side. The COT supplies that second piece of information, with a three-day lag and a perimeter limited to reported futures. It measures crowding, not the unwind in real time. The two indicators are complementary: price describes the move, while positioning describes part of the fuel that may amplify it. Our guide to reading the CFTC COT explains the categories and their limits. The signal available today The chart does not prove that the carry has already unwound. It establishes a more restrained and useful initial condition: the shock met a market almost as short yen as at the July 2024 peak. I infer greater potential sensitivity to forced covering, without claiming that such covering occurred. The next question is falsifiable. If the 4 August COT shows a sharp reduction in the net short position, the result will be consistent with a futures unwind around the shock without proving its cause on its own. If the position remains near its current level, the yen move will have mostly reflected other channels or an adjustment still invisible in this report. The Yen Carry Monitor tracks this series alongside the exchange rate and the policy-rate differential. It should be read according to its methodology: as a risk-monitoring tool, not a comprehensive measure of global carry. For transmission into other assets, see our analysis of dollar-yen and unwind risk and the Japanese bond channel. Sources 1. CFTC, Legacy Futures Only report for 28 July 2026, Japanese Yen CME contract 097741, and the reproducible 170-observation historical query. 2. CFTC, Commitments of Traders overview and limits and 2026 release calendar, with the 7 August release scheduled. 3. CFTC, Traders in Financial Futures, Futures Only, Leveraged Funds category on 28 July 2026. 4. European Central Bank, 90-day reference exchange-rate feed, 31 July 2026. 5. Ministry of Finance Japan, Foreign Exchange Intervention Operations, 29 June to 29 July 2026, zero total. 6. Ministry of Finance Japan, intervention release calendar, next monthly release announced for 28 August 2026. 7. Bank of Japan, MoF authority and the BoJ's operational role. 8. Bank of Japan, Statement on Monetary Policy, 31 July 2026. 9. Bank for International Settlements, OTC foreign exchange turnover in April 2025, FX instrument perimeter and the yen's share of global turnover. 10. Reuters, Japan carries out yen-buying intervention as US executes rate check, 31 July 2026. The intervention is attributed to market participants pending the MoF's official release. Data cut off on 2 August 2026. Ratios and conversions explicitly identified as such are l0g calculations from official data. This is not investment advice. ============================================================================ ANALYSIS: European risk, national supervision URL: https://l0g.fr/en/analysis/european-risk-national-fund-supervision/ Canonical French source: https://l0g.fr/posts/risque-europeen-supervision-nationale-fonds/ Date: 2026-08-01 (reviewed 2026-08-01) Topics: investment funds, European supervision, systemic risk, LDI funds, Ireland and Luxembourg ---------------------------------------------------------------------------- The problem is not an absence of supervisors. An Irish fund answers to the Central Bank of Ireland, a Luxembourg manager to the CSSF, ESMA promotes supervisory convergence and the ESRB monitors systemic risk. The problem emerges when the vehicle, manager, investor and threatened market fall under different countries. Each authority may then see an accurate slice of the case without any one of them necessarily holding both the complete map and the power to act on every link. The UK liability-driven investment crisis of 2022 made that fragmentation visible. This article concludes the “€12 trillion in transit” series. The first instalment separated domicile, manager, investor, currency and issuer. The second followed portfolios into US assets. The third described how a shock could return through liquidity and banks. This final instalment asks which authority can see the whole chain. The fund, its manager and the market do not answer to the same authority European regulation provides a common foundation, but authorisation and day-to-day supervision of funds and managers remain largely national. The Undertakings for Collective Investment in Transferable Securities framework, or UCITS, covers funds marketed mainly to retail investors. The Alternative Investment Fund Managers Directive, or AIFMD, applies principally to alternative fund managers and governs their reporting, leverage and European passport. In both frameworks, the competent authorities in the home country retain the central operational role. The European Securities and Markets Authority, or ESMA, is therefore not the equivalent of the European Central Bank for significant banks. It develops standards, centralises some data, runs peer reviews, settles certain disagreements and promotes convergence. It directly supervises specific infrastructures and categories of firms, but not all European funds or their managers. The European Systemic Risk Board, or ESRB, adds the macroprudential view: it looks for collective behaviour capable of amplifying a shock. It can issue warnings and recommendations without becoming the vehicle's day-to-day supervisor. The European Commission's legislative proposal of 4 December 2025 describes this limitation itself. It argues that the current frameworks do not enable ESMA to address every divergence in national practices or every disagreement over cross-border fund and manager operations effectively. This is an institutional finding, not evidence that national authorities do no work. @media (max-width: 760px) { figure.supervision-scroll { width: 100%; max-width: 100%; min-width: 0; overflow-x: auto; contain: inline-size; } figure.supervision-scroll svg { width: 760px !important; max-width: none; } } The LDI precedent separated risk, domicile and intervention After the UK “mini-budget”, the 30-year gilt yield rose by 140 basis points in four trading days, from 21 to 26 September 2022. The Bank of England says this move was more than twice the previous record observed since 2000. Falling bond prices triggered collateral calls on the repo financing and derivatives used by liability-driven investment, or LDI, strategies. The central bank estimates that margin and collateral calls on LDI funds and pension schemes exceeded £70 billion. From 23 September to 14 October, LDI funds sold about £23 billion of gilts and pension schemes about £14 billion. Those figures do not cover every cash need, which was also met through other asset sales and existing buffers. Domicile complicated the response. The Central Bank of Ireland estimates that Irish-authorised funds accounted for 30% of net gilt sales by LDI funds and their investors during the crisis. This share measures neither the whole gilt market nor all EU-domiciled funds. It nevertheless shows that a national European supervisor oversaw a significant portion of the vehicles amplifying stress in the United Kingdom. The Bank of England had to buy gilts temporarily to break the feedback loop between prices, collateral calls and forced sales. Irish and Luxembourg authorities subsequently coordinated their requirements with ESMA. Since July 2024, the GBP LDI funds concerned must withstand a rise of at least 300 basis points in yields before their net asset value becomes negative. ESMA endorsed the national measures under Article 25 of AIFMD and called on other relevant authorities to adopt similar measures. The case also shows that cooperation can work. Its weakness was timing: the common permanent response was codified after the event, while market intervention had to be decided within days. Regulatory labels sometimes obscure common risks Fragmentation is not only geographical. It begins in the data. An ESRB study published on 4 May 2026 examines alternative funds reported as “other” or “none” under AIFMD. “Other” accounted for €3.6 trillion, or 50% of the net asset value of EU-domiciled alternative investment funds in the fourth quarter of 2024. “None” added €306 billion, or 4%. The authors group more than 10,000 funds holding €3.7 trillion in assets using 70 exposures and eight regions. They identify 12 economically interpretable cohorts, including GBP LDI, European private credit and US private assets. These cohorts explain 16 percentage points more of the variance in returns than traditional AIFMD classifications. The paper is an ESRB authors' study, not an official decision by its General Board. It nevertheless documents two verifiable limitations. Funds resembling funds of funds or private equity funds appear in the residual category despite the existence of dedicated AIFMD types. More importantly, most cohorts are dispersed across several countries, so a national authority may see only part of a European risk. The study also finds that Ireland and Luxembourg manage the vast majority of private equity and private credit funds in its extended sample. That concentration helps the two financial centres build supervisory expertise. It also concentrates information about illiquid and opaque exposures in two jurisdictions while investors and potential losses may sit elsewhere. A small number of groups have a pan-European footprint Fund-by-fund supervision can also miss group structure. An analysis by ECB researchers published in February 2026 identifies roughly 10 to 15 asset management groups ranking highly by size, cross-border activity and interconnectedness. They account for about €6.3 trillion in the dataset and domicile a large share of their funds in Ireland and Luxembourg. That figure has a material limitation. The commercial Lipper dataset covered only about 60% of the euro-area fund sector in October 2025, representing €13 trillion against €21.5 trillion in ECB statistics. It mainly covers UCITS and underrepresents alternative funds. The “10 to 15” are therefore neither a definitive regulatory list nor a comprehensive risk measure. They show that a small core of groups already has a European footprint comparable in reach to centrally monitored banking groups. Size is not enough, however. Many modest funds holding the same assets may sell together. European oversight of large groups would not replace the analysis of cohorts, liquidity and leverage described in the third instalment. The proposed reform does not yet create a single fund supervisor The Commission's December 2025 package proposes that ESMA identify the largest asset management groups by net asset value and cross-border activity, then conduct a review with national authorities at least annually. The text would also strengthen ESMA's ability to address supervisory divergence and, in some cases, suspend a cross-border activity. The scope is deliberately limited. The proposal states that the review concerns managers' operations, not the authorisation or direct supervision of the funds they manage. In July 2026, the Council was still examining whether to retain the annual review, its scope and frequency, and supervisory colleges as an alternative. This is a proposal under negotiation, not a power already in force. The ECB goes further on macroprudential policy. Its May 2026 Financial Stability Review supports reciprocity mechanisms and complementary European “top-up” powers that would let ESMA strengthen a national measure with the authorities concerned and after consulting the ESRB. It also calls for faster cross-border access to granular data and tools addressing liquidity mismatch in open-ended funds. Four building blocks would reduce the blind spot A proportionate architecture does not require transferring every routine check to Paris. Four functions do, however, need to operate at the scale of the risk. 1. Comparable and accessible data. Exposures, leverage, liquidity, investors and delegation chains must be linkable beyond the first vehicle and shared rapidly between authorities. 2. Cohort-based monitoring. Legal classifications need to be complemented by groups of comparable exposures and behaviours, without treating an algorithm as regulatory truth. 3. Operational reciprocity. A measure adopted in one domicile should be reproducible elsewhere before funds can move around it. The Irish and Luxembourg LDI buffer offers a useful precedent. 4. A consolidated view of large groups. ESMA-coordinated colleges or more integrated oversight could bring together supervisors of managers, banks and insurers tied to the same group. Two questions remain open. No threshold for making an asset management group systemic has been settled. More importantly, entity supervision alone cannot solve collective selling by thousands of independent funds. Reform therefore needs to cover both the largest firms and the common activities that transmit stress. The series ends on this distinction. Europe does not domicile €12 trillion without rules or supervisors. It applies common law through mainly national authorities to vehicles connecting global investors, asset management groups and markets. The system can supervise the parts. Its challenge is to see the movement of the whole early enough. Sources 1. Bank of England, Financial Stability Report, December 2022: yield shock, collateral calls, gilt sales and intervention. 2. Central Bank of Ireland, framework for LDI funds: share of sales, coordination and minimum buffer. 3. ESMA, opinion on restrictions for GBP LDI funds, 29 April 2024: AIFMD Article 25 and the 300-basis-point measure. 4. ESRB, “No labels, no problem”, Occasional Paper No 30, 4 May 2026: AIFMD classifications, cohorts, geographical concentration and methodological limitations. 5. ECB, “Why we need an EU perspective in the supervision of large asset managers”, 13 February 2026: group concentration and Lipper coverage limitations. 6. European Commission, proposal COM(2025) 942: proposed role for ESMA and direct fund supervision remaining national. 7. ECB, Financial Stability Review, May 2026: data sharing, reciprocity and top-up powers. This is not investment advice. ============================================================================ ANALYSIS: A US shock returns through European financial plumbing URL: https://l0g.fr/en/analysis/us-shock-european-financial-plumbing/ Canonical French source: https://l0g.fr/posts/choc-americain-plomberie-financiere-europeenne/ Date: 2026-08-01 (reviewed 2026-08-01) Topics: investment funds, nbfi, liquidity, systemic risk, banks, united states, euro area ---------------------------------------------------------------------------- A US asset can lose value without causing a European crisis. Transmission starts only when the loss becomes a need for cash: fund redemptions, collateral to deliver, a derivative to margin or bank funding to roll. It intensifies if several funds sell at the same time and the banks financing them shrink their balance sheets. The risk does not lie in an Irish or Luxembourg address by itself. It lies in the chain connecting a global portfolio, a vehicle domiciled in Europe and European banks. That chain is documented. Its full activation is not. This is the third instalment in the "€12 trillion in transit" series. The first mapped the distinction between domicile, manager, investor, currency and issuer. The second followed capital into US securities. This article takes the route in reverse. A valuation loss is not yet a liquidity crisis In the second quarter of 2025, euro area residents held about €6.1 trillion in US securities, according to the ECB's November 2025 Financial Stability Review. Investment funds accounted for 75% of this population's US equities, almost 50% of its US sovereign debt and about 60% of its other US debt. A correction in those assets lowers their euro value and the NAV of the funds holding them. If investors stay put, the fund has no leverage and its derivatives require no collateral, the loss remains primarily a portfolio loss. A second event is needed for systemic risk: a cash outflow. In an open-ended fund, redemption requests require the manager to mobilise liquidity. It can use cash, sell money-market instruments, reduce a repo position or sell securities. The order depends on the mandate, the assets and market conditions. There is no universal liquidation waterfall. The May 2026 Financial Stability Review estimates that cash represented about 4% of euro area bond fund assets in the first quarter of 2026, compared with about 2% for equity funds. During the April 2025 tariff turmoil, outflows from high-yield funds exceeded their average cash buffers. The ECB says this indicates that they may have resorted to precautionary or forced sales. The observation applies neither to every fund nor to every US equity. It documents the link from redemption to sale. Our analysis of bond funds' liquidity buffer examines that first line of defence. @media (max-width: 760px) { figure.plumbing-scroll { width: 100%; max-width: 100%; min-width: 0; overflow-x: auto; contain: inline-size; } figure.plumbing-scroll svg { width: 720px !important; max-width: none; } } FX hedging changes both the loss and its cash timing The dollar adds a second axis to the shock. The ECB estimates that euro area investment funds and insurance corporations hedge only about one-third of the currency risk in their US dollar bond portfolios. For funds, gross FX derivative notional represented less than 10% of dollar assets in equity funds and 55% in bond funds in the second quarter of 2025. The ECB warns that gross notionals mix long and short positions and cannot reliably reconstruct net hedging. A simultaneous decline in a US asset and the dollar amplifies the euro loss of an unhedged investor. For a fund hedged by selling dollars forward, the dollar's decline instead creates a gain on the hedge that offsets some or all of the FX loss. A margin call is therefore not automatic in this scenario. The cash strain can emerge in the opposite configuration. If the dollar rises, the asset's euro value cushions part of the correction, but a short-dollar hedge can lose and require cash or collateral. The outcome depends on the contract, clearing, netting agreements and margin thresholds. The margin call definition preserves the distinction between the final economic loss and cash that must be delivered immediately. Maturity creates a separate vulnerability regardless of the currency's direction. Long-lived foreign-currency assets are often hedged with shorter FX derivatives that must be rolled. The ECB identifies a liquidity mismatch: when FX markets are strained, rolling becomes more expensive and a fund may have to choose between selling a foreign asset and retaining more currency risk. This is the mechanism behind the cross-currency basis and our guide to dollar liquidity. Bank balance sheets form the European bridge Funds and other non-bank financial intermediaries (NBFIs) deposit cash with banks, lend to them through repo, buy their bonds and borrow to obtain leverage or market access. The joint ECB-ESRB report released on 12 February 2026 isolates two systemic channels. The first runs through bank liabilities. The ECB's detailed study estimates that euro area banks fund, on average, 15% of their assets through liabilities to NBFIs. About 60% of those liabilities are very short-term deposits and repos. A fund facing redemptions or margin calls may withdraw a deposit or decline to roll a repo. The bank then loses funding that may be hard to replace quickly. The second runs through bank assets. Euro area banks' exposures to NBFIs amount to about 10% of their total assets. Much of this credit is collateralised and short-dated, which reduces direct credit risk. But a bank can raise a haircut, refuse to roll funding or reduce the leverage supplied to a counterparty. If the fund then sells into a falling market, collateral values decline further and deleveraging can reinforce itself. This is the secured-funding logic explained in our repo and SOFR guide. The relationship also operates at longer maturities. In June 2025, euro area NBFIs held about €1.5 trillion in bank bonds, close to one-third of the amount outstanding. The ECB notes that this funding, held prominently by insurers and pension funds and spread over long maturities, presents limited immediate liquidity risk. A prolonged loss of bond-market access would matter more than a one-day sale. These figures cannot be added together. They use different denominators, instruments and populations. Nor do they imply that 10% of bank balance sheets finances Irish or Luxembourg funds. They establish the interconnection between euro area banks and the broader non-bank sector. Three conditions make the loop procyclical The scenario becomes systemic only if three conditions combine. 1. Outflows are synchronised. Several funds face redemptions or seek the same collateral at the same time. 2. Buffers cannot absorb the need. Cash, liquid assets and available lines are insufficient, forcing sales. 3. Banks reduce intermediation together. They replace less withdrawn funding, tighten repo, provide less leverage or shrink market-making capacity. The ESRB's EU Non-bank Financial Intermediation Risk Monitor 2025 describes precisely this combination of liquidity mismatch, leverage and interconnectedness. It does not find every open-ended fund fragile. It identifies pockets of risk, notably among funds making heavy use of derivatives, some absolute-value-at-risk UCITS and hedge funds. NBFI must never be treated as one homogeneous balance sheet. Buffers keep the scenario from becoming a forecast Listed equities and Treasuries are generally easier to sell than private credit or thinly traded bonds. Collateralised bank exposures limit losses in default. Long-term bank bonds held by insurers and pension funds do not all flee on day one. Funds can hold cash, stagger sales, use liquidity-management tools and reduce hedges instead of immediately liquidating assets. In February 2026, the ECB and ESRB wrote that bank-NBFI linkages did not then pose acute risks to financial stability, while creating vulnerabilities that could amplify stress. That is the appropriate conclusion here. The channel exists, its scale is significant and some connections are concentrated. Nothing in the public evidence reviewed supports a claim that a crash is imminent. The thesis is falsifiable. A US correction accompanied by contained redemptions, absorbed margin calls, stable NBFI deposits and a functioning bank repo market would remain mainly a portfolio loss. Synchronised outflows, forced sales, withdrawals of short-term funding and rising haircuts would instead show that the plumbing is transmitting the shock. The fourth instalment examines the fragmented supervision of this chain when the fund, manager, investor, asset and bank fall under several jurisdictions. Sources 1. ECB, "What safe haven after the April US tariff announcement?", Financial Stability Review, November 2025: US holdings, FX hedging and the maturity mismatch in derivatives. 2. ECB, Financial Stability Review, May 2026: cash buffers, redemptions and procyclical fund sales. 3. ECB, "Systemic risks in linkages between banks and the non-bank financial sector", November 2025: bank funding, exposures and bank bonds held by NBFIs. 4. ECB and ESRB, press release on bank-NBFI linkages, 12 February 2026: two transmission channels and the current risk assessment. 5. ESRB, EU Non-bank Financial Intermediation Risk Monitor 2025: liquidity, leverage, margin calls and fund heterogeneity. 6. FSB, Liquidity Preparedness for Margin and Collateral Calls, 10 December 2024: cash management, stress testing and collateral availability. This is not investment advice. ============================================================================ ANALYSIS: Two small states, €12 trillion in transit URL: https://l0g.fr/en/analysis/two-small-states-12-trillion-in-transit/ Canonical French source: https://l0g.fr/posts/deux-petits-etats-12000-milliards-transit/ Date: 2026-08-01 (reviewed 2026-08-01) Topics: ireland, luxembourg, investment funds, europe, capital flows, regulation ---------------------------------------------------------------------------- At 31 March 2026, funds domiciled in Ireland reported €5.667 trillion in net asset value, including money market funds. Luxembourg undertakings for collective investment reported €6.2078 trillion. At the same date, the total was €11.8748 trillion, close to €12 trillion. Replacing Luxembourg's March figure with its latest May observation, €6.6344 trillion, while Ireland's latest release remains at March produces €12.3014 trillion, but the two clocks no longer match. This precision does not make the phenomenon smaller. It identifies it correctly: Ireland and Luxembourg are less two vaults filled with domestic savings than two major addresses in the circulation of global capital. This article opens the "€12 trillion in transit" series. The first instalment establishes the map. The second follows the portfolios from European funds into US securities. The next two will examine the return channels for an external shock and the fragmentation of European supervision. The sum works under strict conditions The Central Bank of Ireland publishes two separate populations. At the end of March 2026, the net asset value of investment funds excluding money market funds was €4.718 trillion. Money market fund NAV was €949 billion. Together they equal €5.667 trillion. The release also reports gross assets under management of €5.595 trillion for the first group and €972 billion for the second. Adding those gross amounts to Luxembourg net assets would create an inconsistent aggregate. NAV measures a fund's assets minus its liabilities. AUM in the Irish release measures assets under management before this subtraction. This article therefore compares net values only. In Luxembourg, the CSSF reported €6.207822 trillion in net assets at 31 March. Its perimeter includes undertakings subject to the 2010 law, specialised investment funds and SICARs. It also includes the money market categories within those undertakings, which must not be added a second time. The dates also matter for interpretation. Between end-March and end-May, Luxembourg net assets rose by €426.571 billion. CSSF releases for April and May attribute only €36.174 billion to net investment. The remaining €390.397 billion came from market movements. Crossing the threshold did not correspond to €426 billion of fresh money. One fund has five economic addresses The published domicile answers one precise question: under which law is the fund constituted, and which authority supervises its operation? It does not answer the four other questions needed to locate savings and risk. 1. The fund's legal domicile. An Irish or Luxembourg UCITS is subject to its home state's rules for incorporation, valuation, and the issuance and redemption of shares. The EU UCITS Directive explicitly distinguishes the fund's home state from that of its management company. 2. The manager's country. The same directive allows a management company authorised in one Member State to manage a fund established in another. Portfolio management may also be delegated under the responsibility of the regulated manager. The fund's domicile therefore locates neither every team nor necessarily the group making investment decisions. 3. The investor's residence. Issued shares are the fund's liabilities. Their holder can be a household, insurer, pension fund, another fund, or an intermediary acting for clients. An ECB study of the geography of capital allocation uses Irish and Luxembourg administrative data to look through immediate counterparties, often financial intermediaries, towards underlying owners. As a benchmark, Irish households held €11.2 billion in fund shares in Q3 2025, including €6.1 billion in Irish-domiciled funds. That figure does not include all Irish capital, notably pension and insurance assets. It is nevertheless enough to rule out the idea that the trillions domiciled in Dublin are the direct portfolio of local households. 4. The asset's currency. A bond can be denominated in dollars and issued by a European company. A US equity can sit in a euro-denominated fund share with its currency risk hedged. Currency identifies the unit of contractual cash flows, not the issuer's nationality or the post-hedging currency risk. The Irish Q1 release separately reports purchases of US securities and valuation effects on dollar positions. 5. The final issuer's country. It locates the company, government or bank financed by the security purchase. When a fund owns shares in another fund, the first observed issuer is still a vehicle. One or more layers must be looked through. The ECB publishes experimental data designed to reconstruct those underlying assets. The Central Bank of Ireland also explains the limit: the initial look-through reveals asset categories but not always the characteristics of every final security. A transit infrastructure, not a €12 trillion economy The transit description comes from external accounts, not from an assumption about tax. In its June 2026 report on the international role of the euro, the ECB estimates that foreign investors made more than €850 billion in net purchases of euro area securities in 2025. About €470 billion went into fund shares. A share usually close to three-quarters of those inflows is then reinvested outside the euro area by funds established mainly in Ireland and Luxembourg. This does not mean that 75% of the €12 trillion follows that route. The ECB ratio applies to a specific flow, foreign inflows into euro area funds, not to the total stock of their portfolios. It does establish the international intermediation role of the two centres. The ECB supplies a broader benchmark in its 2026 Financial Integration and Structure report: euro area investment funds held €22.1 trillion in gross assets at the end of 2025, and about half of their securities were issued outside the euro area. That gross total cannot be added to national NAVs. It describes portfolio destination, not the net value due to fund investors. Four limits stop the total from becoming a risk measure Funds of funds create layers. When a Luxembourg fund holds a share in an Irish fund, each vehicle reports its own NAV. The ECB study cited above must unwind cross-holdings before assigning assets to underlying investors. Adding both industries correctly measures legal activity in two domiciles, but it can count the same capital at several levels. The sum is not a consolidated measure of money reaching final issuers. Statistical residence does not always identify the ultimate owner. A distributor, custodian or omnibus account can appear as the holder for many clients elsewhere. The data correctly describe the first declared counterparty. They do not always pierce the full ownership chain. Currency is not enough to locate risk. A dollar asset is not necessarily American, and a hedge can transfer currency risk to a bank in a third country. Issuer residence, currency and derivative counterparty must remain separate columns. The regulatory perimeter is not the whole universe. The Luxembourg release specifies which categories it aggregates. The CSSF publishes a separate framework for funds it does not directly authorise. The €6.6344 trillion figure is therefore an official and reproducible measure of covered undertakings, not an exhaustive estimate of every private structure linked to Luxembourg. The solid conclusion before tracing the assets The €12 trillion figure proves neither systemic fragility, nor a flight of European savings, nor a failure of supervision. It establishes something else: two small states provide the legal envelope and part of the operating infrastructure for a mass of funds whose owners, decision-makers, currencies and assets are largely cross-border. This separation avoids two equally false stories. The first would treat the trillions as the wealth of Ireland and Luxembourg. The second would treat every asset domiciled there as European financing. Domicile data support neither conclusion. The second instalment follows the portfolios, separating dollar-denominated securities from US issuers, European from non-European capital, subscriptions from transactions, and purchases from valuation effects. Primary sources: Central Bank of Ireland, Q1 2026 investment fund statistics; CSSF, UCI net assets at 31 March 2026 and 31 May 2026; ECB, The international role of the euro, June 2026; ECB, Financial Integration and Structure in the Euro Area, May 2026; ECB, The geography of capital allocation in the euro area, 2024; Directive 2009/65/EC on UCITS. l0g calculations use unrounded values where authorities publish them. This is not investment advice. ============================================================================ ANALYSIS: European funds, American portfolios URL: https://l0g.fr/en/analysis/european-funds-american-portfolios/ Canonical French source: https://l0g.fr/posts/fonds-europeens-portefeuilles-americains/ Date: 2026-08-01 (reviewed 2026-08-01) Topics: investment funds, united states, europe, capital flows, nbfi, savings ---------------------------------------------------------------------------- The channel to the United States is measurable. Its owner is harder to identify. In the first quarter of 2026, investment funds domiciled in Ireland made €135 billion of net purchases in equities and debt securities. US securities absorbed €66 billion, almost half. But that €66 billion is not automatically European savings. It identifies the country of the issuer bought by an Irish fund, not the residence of the investor holding the fund's shares. A second statistic is needed. ECB data show that euro area residents are indeed heavily exposed to the United States and that funds are their main route into US equities. This is the second instalment in the "€12 trillion in transit" series. The first separated the fund's domicile, manager, investor, currency and final issuer. This one follows the portfolios. Dublin was a net buyer of US securities The Central Bank of Ireland counts funds resident and authorised in Ireland, excluding money market funds in this part of its release. At the end of March 2026, their assets under management reached €5.595 trillion. Equities represented 55%, debt securities 29%, and cash, deposits, loans and other assets the remaining 16%. During the quarter, net transactions reached €88 billion in equities and €48 billion in debt securities. Those two rounded figures add to €136 billion, while the Central Bank reports an unallocated total of €135 billion. Within that total, net purchases of US securities reached €66 billion, or 48.9%. This is a quarterly flow, not the US share of the whole portfolio. The change in the stock cannot be equated with purchases either. Equity holdings rose from about €3.0 trillion to €3.1 trillion. They received €88 billion of positive transactions but suffered €33 billion of negative revaluations. Debt securities combined €48 billion in purchases with a €5 billion positive revaluation. A valuation effect is a change caused by prices or exchange rates rather than a purchase or sale. Fund investors reside elsewhere The liability side of a fund identifies who holds its shares. A more detailed Central Bank release for the third quarter of 2024 found that the United Kingdom held 41% of the shares in Irish funds excluding money market funds, ahead of the Netherlands at 12%, Luxembourg at 10% and Ireland at 9%. These countries identify the immediate holders on record, not necessarily the ultimate owners behind an intermediary or omnibus account. The snapshot also predates the flow under review by eighteen months. It nevertheless sets the boundary: "Irish fund" does not mean "Irish savings", or even "European Union savings". The leading disclosed holder country was outside the EU. Assigning the €66 billion solely to European savers would therefore be an unsupported extrapolation. European capital appears in a different population To isolate European capital, the starting point must be euro area resident investors, followed by their foreign assets. The ECB's November 2025 Financial Stability Review does exactly that. In the second quarter of 2025, those residents held more than €12 trillion in foreign portfolio assets, about half of them issued in the United States: €3.8 trillion in equities, €0.8 trillion in sovereign debt and €1.5 trillion in other debt securities. The three rounded items add to €6.1 trillion. Investment funds accounted for 75% of the US equities held by euro area residents, almost 50% of their US sovereign debt and about 60% of their other US debt. Here, the statistical holder is a euro area resident. The ECB nevertheless notes that its dataset does not comprehensively cover foreign securities held outside the euro area and that the amounts are reported at market value. Allocation shifted, but prices built most of the stock The recent change is not solely the result of Wall Street's rise. In an analysis published on the ECB Blog on 15 May 2026, five economists calculate that transactions by euro area non-bank financial institutions increased the portfolio share of US corporate equities by 2.7 percentage points between the fourth quarter of 2023 and the fourth quarter of 2025. Over the same period, the share of euro area corporate equities fell by 1.5 points. The authors estimate that a one-point increase in the US equity share is associated with a 0.3-point decline in the euro area equity share. This is an econometric association, not proof that every euro withdrawn in Europe directly finances one euro in the United States. The post also says that its views do not necessarily represent those of the ECB or the Eurosystem. Over ten years, however, the market effect dominates. A May 2026 Financial Stability Review analysis estimates that euro area investors' US equity holdings quadrupled between 2015 and 2025. About 70% of the increase came from valuation effects, mainly prices, and 30% from net transactions. Exposure grew for two separate reasons: investors bought, then the relative performance of US equities magnified their weight. A savings union must track destination, not address The European Commission presents the Savings and Investments Union as a way to connect savings with productive investment and finance the EU's strategic objectives. Market integration can cut costs, improve diversification and offer better vehicles to savers. By itself, it cannot guarantee that additional capital will buy European securities. The available evidence therefore supports a two-part answer. Euro area residents do finance US markets on a large scale, and funds are the main vehicle for their US equity exposure. The recent allocation of euro area NBFIs has also shifted towards US equities at the relative expense of euro area equities. But Ireland's €66 billion flow combines capital from several origins and cannot be used as a meter of European savings leaving the continent. The right policy indicator is not the volume domiciled in Dublin or Luxembourg. It would cross, for every period, the residence of the fund investor, the residence of the issuer and net transactions, while separating price and currency effects. Without that matrix, Europe can improve its financial plumbing without knowing precisely which economy receives the capital flowing through it. The third instalment now follows the reverse path: how a fall in US assets can travel back into European funds, investors and markets through redemptions, margin and bank balance sheets. Primary sources: Central Bank of Ireland, Q1 2026 investment fund statistics and Q3 2024 holder distribution; ECB, US holdings of euro area residents, November 2025; ECB Blog, NBFI portfolio reallocation, 15 May 2026; ECB, drivers of flows into US equities, May 2026; European Commission, Savings and Investments Union, updated 17 July 2026. Additions and ratios explicitly identified as such are l0g calculations based on rounded official figures. This is not investment advice. ============================================================================ ANALYSIS: After the fines: the black box inside JPMorgan market surveillance URL: https://l0g.fr/en/analysis/jpmorgan-market-surveillance-black-box/ Canonical French source: https://l0g.fr/posts/jpmorgan-boite-noire-controle-marches/ Date: 2026-07-31 (reviewed 2026-07-31) Topics: JPMorgan, sponsored access, trade surveillance, CFTC, Federal Reserve, OCC, governance, operational risk ---------------------------------------------------------------------------- Part one established the sequence: manipulation admitted in 2020, surveillance gaps discovered in 2021 and coordinated sanctions in 2024. It also set a necessary limit. Billions of order messages missing from JPMorgan's systems are not billions of abuses. They are billions of objects never tested by the scenarios intended to detect abuse. Part two starts after the fine. Orders from the Federal Reserve, Office of the Comptroller of the Currency and Commodity Futures Trading Commission required a retrospective review, a full list of trading venues, independent assessment, a corrective plan, progress reports and, for the CFTC, a final certification. As of 31 July 2026, the public sources consulted for this investigation disclose the obligations but not the work produced under them. The issue is larger than an IT fault. A global dealer buys software, combines feeds from many markets and allows algorithmic clients to reach venues through different contractual arrangements. Control works only if every expected message arrives, every venue appears in the inventory and each detection test covers the relevant behaviour. Risk grows at the junctions. Sponsored access, delegated trading and retained control On the venue called "DCM-1", JPMorgan attributed most missing messages to sponsored access trading by three significant algorithmic firms. The CFTC records this explanation without identifying the venue or the firms. Sponsored access lets a client or intermediary send orders to a venue through a market member's access. Economically, the client originates the order. Routing, clearing, checks before trading and monitoring after trading can fall to different entities depending on the market and contract. The rules reflect that detail. In securities markets overseen by the SEC, Rule 15c3-5 requires the broker with market access to maintain financial and regulatory controls under its direct and exclusive control, subject to limited exceptions. It must also review their effectiveness regularly. Futures rules are different. In a 2013 interpretation, the CFTC said a futures broker providing sponsored access to an executing firm is not, solely because it provides access, required under Regulation 1.73(a)(2)(iv) to screen the executing firm's customer orders. One shortcut therefore fails: sponsorship does not create universal responsibility for every control. The distinction does not weaken the 2024 case. The CFTC sanctioned J.P. Morgan Securities under Regulation 166.3 for failing to supervise diligently. The proven failure involved ingesting and monitoring order messages. The client's economic identity did not make the data feeding JPMorgan's own surveillance optional. Every alert needs a complete chain Electronic market surveillance is a chain, not one piece of software: 1. the venue creates messages for new orders, changes, cancellations and executions; 2. connectors transport those messages and put them into a common format; 3. an inventory links each venue, product, trading team and client to the correct control rules; 4. reconciliation compares the amount expected with the amount received; 5. detection tests search for suspicious patterns; 6. analysts examine alerts and record their decisions; 7. serious cases reach compliance staff, managers and, where appropriate, regulators. In 2020, JPMorgan told the CFTC that it used three main alert types in the SMARTS software for spoofing and layering. Order 20-69 also describes quality checks and monthly reporting by trader, team, supervisor and region. Those controls came after data entry. A well-designed test never sees a message that failed to arrive. The golden-source assumption JPMorgan reconciled some data every quarter but excluded feeds received directly from venues. The firm assumed exchange data were a golden source and did not need the same test. The assumption mixed up two properties: - accuracy of the data produced by the venue; - completeness of the data arriving in JPMorgan's tool. The first can be excellent while the second falls to zero. A wrong setting, an unrecognised product code, an incomplete connector or a rejected transformation can be enough. The CFTC identified feed-configuration problems as a cause of the gaps. It did not publish a breakdown of each type of failure. Operational risk then becomes circular. The control system assumes its own input is complete. Without a separate test of that assumption, an empty alert dashboard looks reassuring. It can also mean no data arrived. Five CFTC reports and steps The 23 May 2024 CFTC order did more than impose a penalty. It required a precise sequence: 1. a JPMorgan report listing each affected venue and activity, the period, the volume not surveilled and any related market misconduct; 2. an independent consultant's report on policies, the venue inventory, reconciliation, detection tests, testing and previously unsurveilled activity; 3. a remediation plan responding to the consultant's findings and recommendations; 4. quarterly progress reports describing work completed, status and timing; 5. a completion certification signed by the chief compliance officer and another senior business executive. The Commission may extend deadlines for good cause. Quarterly reporting ends only after the certification is submitted and accepted by the Division of Enforcement. The Fed order follows a similar structure: an internal report, an independent party, a report to the board and Federal Reserve Bank of New York, an approved plan and quarterly updates. The review must cover the firm's own trading and client activity, board oversight, the venue inventory, automated reconciliation, detection tests and periodic testing. The OCC order also requires a lookback, a retrospective search through previously unsurveilled activity for misconduct not identified earlier. Its penalty order expressly preserves the possibility of an additional penalty based on the lookback results. JPMorgan's conclusion and no public audit In its second-quarter 2024 Form 10-Q, JPMorgan said it had completed improvements to its venue inventory and data-completeness controls. Other remediation remained underway. The firm had retained the required independent consultant and paid approximately $450 million in coordinated penalties. The filing also says its review of previously unsurveilled data identified no employee misconduct, no harm to clients and no harm to the market. The wording matters and should remain intact. It is JPMorgan's published conclusion. It is not a published retrospective report or independent assessment. As of 31 July 2026, our searches of public CFTC, Fed and OCC pages and JPMorgan's SEC reports did not locate the content of those documents. The public therefore cannot compare: - the method applied to billions of messages; - the detailed venues, products and periods; - the thresholds used to rebuild alerts; - the consultant's recommendations; - exceptions, limitations and validation tests; - the status of any final certification accepted by the CFTC. Non-publication does not mean the reports were not delivered to regulators. The orders require delivery. It means JPMorgan's conclusion cannot be reproduced from public material. DCM-1 remains unnamed The CFTC uses "DCM-1", shorthand for designated contract market, a regulated US futures venue. No exchange name appears in the order. Assigning one would be speculation. The anonymity blocks several external checks. Without the venue, readers cannot compare the period with its rules, technical notices, feed incidents or disciplinary data. Without the products, concentration by asset class remains unknown. Without the algorithmic firms, their regulatory histories cannot be checked. Silence may protect commercial information, clients or investigations. The order gives no specific public explanation. This investigation therefore retains DCM-1 and refuses to guess. Three unnamed algorithmic firms JPMorgan described three "significant" algorithmic firms behind most sponsored activity on DCM-1. The adjective gives no volume, market share or risk measure. Concentration among three clients nevertheless creates a clear mechanism. A misconfigured feed for a few very active producers can generate billions of missing messages. Automation explains the scale. It proves neither fraud nor loss, but makes a retrospective reconstruction harder and magnifies mistakes in the control perimeter. Monitoring bank employees can also differ from monitoring high-frequency clients. Detection tests, identifiers, noise thresholds and referral methods need not be the same. The independent review was supposed to assess both proprietary and client trading, as well as detection thresholds. Its absence from public material prevents an external assessment of this distinction. No alert is not proof of no abuse The teaching point has three parts: 1. a complete system can generate no alerts because no suspicious behaviour exists; 2. an incomplete system can generate no alerts because the necessary messages are missing; 3. alert counts become meaningful only after data completeness has been established. This logic does not turn an unknown into suspicion. It fixes the order of proof. Completeness comes first, then threshold settings, human review and attribution of intent. A 2025 thematic review by the International Organization of Securities Commissions states the principle for market authorities: access to orders, trades and cancellations is necessary for effective surveillance and market reconstruction. The report is not about JPMorgan. It confirms the general control logic. Academic research reaches the same data requirement. Bao Linh Do and Tālis Putniņš identify order-book imbalances, order activity, abnormal cancellations and cyclical patterns among useful inputs in their study of spoofing detection. The paper proposes a method and makes no finding about DCM-1. Without complete order messages, those inputs cannot be reconstructed reliably. Four layers of risk The case presents four different risks. Combining them produces either an excessive allegation or false reassurance. Conduct risk. Manipulative behaviour can escape detection when its messages never enter the system. The public record does not show such behaviour in the missing feeds. Regulatory risk. The CFTC, Fed and OCC have already imposed sanctions. The OCC order allows another penalty based on the lookback. Any new action would depend on new facts or inadequate remediation, neither established here. Operational risk. An inventory, connector or reconciliation error can neutralise sophisticated detection tools. The problem lies in the path taken by data, not only in the alert model. Governance risk. In 2020, the board and regulators received a detailed description of improvements. In 2021, the firm discovered a massive missing perimeter. The governance test is independent validation of coverage, not the number of written procedures. This risk does not appear in quarterly earnings like a loan-loss provision. It resembles the market plumbing in our analysis of repo and collateral: infrastructure looks secondary until a break exposes every connection. Our guide to bank earnings and risk likewise separates accounting performance, conduct and operational exposure. Five markers for follow-up A long-running investigation needs falsifiable updates. Five events would change the assessment: 1. a public lifting of new-venue onboarding restrictions by the Fed or OCC; 2. a completion certification accepted by the CFTC, if the agency publishes it; 3. a new action based on the lookback, a possibility expressly preserved by the OCC; 4. a more detailed JPMorgan disclosure on reconstruction methods, the consultant or remediation status; 5. a court decision or regulatory case connecting specific conduct to an unsurveilled period and venue. Until such evidence appears, only three conclusions are firm: the gap was massive, JPMorgan says it identified no harm, and the public lacks the reports needed to verify that conclusion independently. Limits of the public record This investigation does not identify DCM-1 or try to infer its identity. It names none of the three algorithmic firms. It does not treat every cancellation as fraud. It does not turn a surveillance failure into manipulation. It also does not add coordinated penalties as independent payments. The effective $448.168 million paid in 2024 is separate from the coordinated $920.204 million resolution in 2020, but each group has its own credits between agencies. Finally, reports produced for regulators and consultants may contain confidential information. Their non-public status does not necessarily violate the orders. It limits the ability of readers, investors and researchers to audit the published conclusion. Primary sources 1. CFTC, Order 24-07 on surveillance gaps, 23 May 2024. 2. Federal Reserve, Orders 24-007-B-HC and 24-007-CMP-HC, 14 March 2024. 3. OCC, Order AA-EC-2023-50, 14 March 2024. 4. OCC, penalty Order AA-EC-2023-49, 14 March 2024. 5. JPMorgan Chase, Form 10-Q for 30 June 2024, Trading Venues Investigations note. 6. CFTC, Order 20-69 on surveillance and spoofing, 29 September 2020. 7. SEC, Rule 15c3-5 on market access, 3 November 2010. 8. CFTC, Interpretative Letter 13-27 on sponsored access and Regulation 1.73, 29 April 2013. 9. IOSCO, Thematic Review on Technological Challenges to Effective Market Surveillance, 2025. Additional academic source: Bao Linh Do and Tālis J. Putniņš, “Detecting Layering and Spoofing in Markets”, version dated 3 November 2023. The paper is used only to explain the data needed for detection. It does not study JPMorgan. Method and limit: research closed on 31 July 2026 across public CFTC, Fed, OCC and SEC orders, releases and databases, then JPMorgan regulatory filings. "Not found" describes the public documents searched. It proves neither the absence of a confidentially submitted report nor a breach of an order. ============================================================================ ANALYSIS: The market beyond the screen: billions of order messages outside JPMorgan surveillance URL: https://l0g.fr/en/analysis/jpmorgan-market-beyond-screen-order-surveillance/ Canonical French source: https://l0g.fr/posts/jpmorgan-marche-hors-radar-surveillance-ordres/ Date: 2026-07-31 (reviewed 2026-07-31) Topics: JPMorgan, spoofing, trade surveillance, CFTC, Federal Reserve, OCC, regulation, operational risk ---------------------------------------------------------------------------- In June 2021, onboarding a new trading venue exposed an anomaly at JPMorgan. Order and trade feeds were not properly reaching its surveillance tools. The internal review then expanded across the world: at least 30 venues were affected, several products were involved and some gaps dated back to 2014. On one US venue identified only as "DCM-1", more than 99% of order messages escaped surveillance from 2014 to 2021. The volume ran into billions. The Commodity Futures Trading Commission did not find billions of fraudulent trades. It established a surveillance failure on an exceptional scale. The distinction defines this investigation. Missing data do not prove manipulation. They prevent the very control designed to detect it. The problem becomes more serious in light of the past: in September 2020, JPMorgan had admitted manipulation in precious metals and US Treasuries, promised remediation and described a strengthened system to the CFTC. Part of the market nevertheless remained invisible to it. This two-part investigation reconstructs the record from orders issued by the CFTC, Federal Reserve and Office of the Comptroller of the Currency, the Department of Justice criminal case, Securities and Exchange Commission action and JPMorgan's own SEC filings. This part establishes chronology and risk. Part two examines sponsored access, the surveillance pipeline and the material missing from the public record. An established scandal The starting point is neither rumour nor extrapolation. In September 2020, JPMorgan entered a deferred prosecution agreement with the Department of Justice. The bank admitted two separate wire-fraud schemes involving market manipulation. The DOJ case record first covers gold, silver, platinum and palladium futures. From March 2008 to August 2016, traders and salespeople on trading teams in New York, London and Hong Kong placed orders intended for cancellation before execution on tens of thousands of occasions. The second scheme involved Treasury futures and the cash market for Treasury notes and bonds. From April 2008 to January 2016, the DOJ records thousands of deceptive sequences. Liability did not remain solely corporate. In August 2023, Gregg Smith and Michael Nowak received prison sentences. Smith was sentenced to two years, Nowak to one year and one day. The DOJ put losses to market participants in the scheme tried before the jury at more than $10 million. Those convictions concerned precious metals and individual conduct proven at trial. They establish nothing about the messages missing from surveillance between 2014 and 2021. Spoofing, a false order in the book A limit order book displays intentions to buy and sell. Price, quantity and visible depth help participants and algorithms estimate supply and demand. The Commodity Exchange Act defines spoofing as bidding or offering with the intent to cancel before execution. The mechanism used by JPMorgan traders combined two sides: 1. a genuine order intended for execution; 2. one or more deceptive orders on the opposite side, intended for cancellation. The false orders created the appearance of stronger buying or selling pressure. Once the genuine order traded at a more favourable price, the deceptive orders disappeared. A rapid cancellation alone does not prove spoofing. Intent to cancel before execution is the decisive element. Criminal cases established it from trading sequences, communications and other evidence. Eight years, two teams, three markets The 2020 CFTC order goes further than the criminal summary. It describes hundreds of thousands of deceptive orders in precious metals and Treasury futures from 2008 to 2016. JPMorgan traders created artificial prices in many instances. The order also found a supervision failure at J.P. Morgan Securities. Warning signs existed. The CFTC cites internal alerts, inquiries from CME and the Commission, and internal allegations from a JPMorgan trader. Before 2014, the surveillance system could not effectively identify spoofing. A newer tool followed, yet the firm still failed to identify, investigate and stop the conduct during the relevant period. The SEC separately documented manipulation in cash Treasuries from April 2015 to January 2016. J.P. Morgan Securities admitted the findings. Genuine orders were accompanied almost simultaneously by non-bona-fide orders on the opposite side to improve execution prices. The conduct was not confined to metals or futures. A strengthened system presented in 2020 At settlement, JPMorgan described a significant change in control. The CFTC order records the firm's representations: hundreds of new compliance officers, larger budgets, specific training, surveillance of more than 80 equity exchanges and more than 40 futures and options exchanges. JPMorgan also said it used three primary alert types in the SMARTS software for spoofing and layering. Quality testing covered alerts referred to a higher review level and alerts closed without such referral. Monthly reports aggregated alerts by trader, team, supervisor and region. The firm further represented that its communications platforms processed about 100 million electronic messages each month and analysts reviewed every alert generated by that communications surveillance. The CFTC did not frame those statements as an end-to-end audit of every market feed. It recorded them among JPMorgan's remediation representations. The order required the firm to maintain and update a programme designed to detect and deter violations. The $920.2 million commonly attached to the case was one coordinated resolution, not three independent sums. The DOJ breakdown was a $436,431,811 criminal penalty, $311,737,008 in victim compensation and $172,034,790 in disgorgement. Credits for CFTC and SEC payments prevented double counting of the same components. A new venue exposes the gap in June 2021 Nine months after the settlement, an ordinary event prompted a major discovery. JPMorgan was preparing to onboard a new venue. In June 2021, it identified significant gaps in the order and trade data reaching its surveillance systems. The review went global. According to the May 2024 CFTC order, gaps affected at least 30 venues, multiple products and periods dating back to at least 2014. JPMorgan disclosed them to the Commission in 2021 and represented that it had not known about them. They therefore had not been discussed during the 2020 settlement. The CFTC states the consequence directly: at the time of the spoofing resolution, JPMorgan was not surveilling certain order messages. Improvements described in 2020 could have been genuine for data present in the system. They did not cover absent data. DCM-1 and more than 99% missing The most severe case involved a regulated US futures venue anonymised by the CFTC as "DCM-1". From 2014 to 2021, JPMorgan failed to feed billions of order messages into surveillance. More than 99% of the venue's messages went unsurveilled. Three qualifications prevent a false reading: - an order message can create, modify or cancel an order and is not necessarily an executed trade; - the message count gives neither a total dollar value nor the size of any position; - the CFTC sanctioned a supervision failure, not new manipulation across every missing message. JPMorgan said most activity came from sponsored access trading for three significant algorithmic firms. The order names neither the venue nor the firms. Part two examines this architecture because it changes the origin of orders without removing the need for complete surveillance data. A golden source without reconciliation The technical cause described by the CFTC was a data-governance error. JPMorgan used feeds received directly from exchanges. It had a quarterly process for reconciling the completeness of some data sent to surveillance tools, but direct-from-exchange feeds were excluded. The assumption sounded reassuring and became the central defect: exchange data were treated as a golden source and therefore not tested by the same reconciliation. The content could be accurate at origin and still fail during configuration, transport or entry into the system. The CFTC identifies configuration problems that kept feeds from entering a third-party surveillance system. Risk did not live solely in data quality. It appeared between systems. A file could be reliable and an alert engine functional while the end-to-end chain remained blind because nobody compared messages received with messages expected. Three penalties and one data failure On 14 March 2024, the Federal Reserve and OCC acted simultaneously. The Fed found gaps from 2014 to 2023 on at least 30 global venues and inadequate controls over data and reconciliation. It classified the practices as unsafe or unsound and imposed $98,167,980. The OCC reached the same unsafe-or-unsound finding. Its order concerned billions of trading instances, at least 30 venues, data governance and venue coverage. It imposed $250 million, paid to the Treasury according to the agency. On 23 May 2024, the CFTC added a nominal $200 million obligation. Its order granted two $50 million credits for payments under the OCC and Fed actions. The CFTC-specific payment therefore became $100 million if both credits applied, producing a coordinated total of $448,167,980. In its 30 June 2024 Form 10-Q, JPMorgan rounded the figure to $450 million and said it had paid it. A limited finding in the second case The 2024 actions did not find a repetition of the 2008 to 2016 manipulation. The CFTC sanctioned J.P. Morgan Securities for failure to supervise. JPMorgan admitted the facts concerning the scope and causes of the gaps and acknowledged a violation of CFTC Regulation 166.3. Under the settlement formula, it neither admitted nor denied the other findings. In its second-quarter 2024 report, JPMorgan said it had reviewed the previously unsurveilled data and identified no employee misconduct, harm to clients or harm to the market. This was the firm's published conclusion. Regulatory orders also required detailed reports and independent review. Their contents do not appear in the public documents consulted for this investigation. The DOJ, meanwhile, closed the 2020 criminal case. The three-year DPA term expired on 29 September 2023. On 29 March 2024, the Department moved to dismiss with prejudice on the ground that JPMorgan had fully met its obligations; the court granted the motion the same day. The decision followed the Fed and OCC orders but concerned compliance with the earlier criminal agreement. It did not erase the 2024 findings or certify every surveillance feed. Part two and the control black box Part one supports a narrow conclusion. JPMorgan admitted a vast historical manipulation scheme. During part of the remediation period, billions of messages did not reach surveillance. Regulators established a supervision failure, not a new fraud across those messages. The central risk is therefore one of knowledge. A bank can count alerts, calibrate scenarios and hire control staff while remaining unaware of a massive upstream absence. Without end-to-end reconciliation, the dashboard measures only data received. The investigation continues in “After the fines: the black box inside JPMorgan market surveillance”: sponsored access, venue inventories, mandated reports, independent review, public unknowns and tests for remediation. Primary sources 1. CFTC, Order 20-69 on manipulation and supervision failures, 29 September 2020. 2. Department of Justice, case 20-CR-175 and deferred prosecution agreement, updated 27 August 2024. 3. Department of Justice, sentencing of Gregg Smith and Michael Nowak, 22 August 2023. 4. SEC, cash Treasury manipulation action, 29 September 2020. 5. Federal Reserve, Orders 24-007-B-HC and 24-007-CMP-HC, 14 March 2024. 6. OCC, trade-surveillance order and penalty, 14 March 2024. 7. CFTC, Order 24-07 on surveillance gaps, 23 May 2024. 8. JPMorgan Chase, Form 10-Q for 30 June 2024, Trading Venues Investigations note. Method and limit: each penalty was reconciled to its order and applicable credits. “Billions” refers to order messages, not dollar value. This investigation attributes no misconduct to missing feeds beyond conduct already admitted or tried. JPMorgan's conclusions are labelled as such. No internal report, non-public consultant report or non-public remediation report was used. ============================================================================ ANALYSIS: When the warehouse does not clear: Atlas risk inside the Apollo-Athene machine URL: https://l0g.fr/en/analysis/when-credit-warehouse-does-not-clear-atlas-apollo-athene/ Canonical French source: https://l0g.fr/posts/atlas-entrepot-credit-risque-apollo-athene/ Date: 2026-07-30 (reviewed 2026-07-30) Topics: apollo, athene, atlas, private credit, securitisation, insurance, liquidity, risk ---------------------------------------------------------------------------- Atlas SP Partners calls itself a "finance company to finance companies". Part of its business is to advance money against loans and receivables, then turn those temporary warehouses into longer capital-markets financing. While the exit works, the warehouse revolves. If investors step back, assets stay longer, funding must be extended and contractual protections become decisive. Apollo describes this scenario in one of its own fund filings. At the same time, group insurer Athene held $6.146 billion of securities issued by Atlas or its affiliates at 31 March 2026, with another $1.343 billion of commitments. Two group entities also guarantee a $2.5 billion deferred obligation to Credit Suisse. These figures must not be added. They map three different channels in the same system: investment, future funding and contingent support. The warehouse before securitisation Warehouse financing is a temporary facility secured by a pool of loans or receivables. A consumer lender, mortgage originator or specialty-finance platform produces the assets. The warehouse finances them while they accumulate. Once the portfolio is large and documented enough, it can be sold or transferred to a vehicle that issues asset-backed securities to investors. The Federal Reserve's report to Congress on risk retention provides the useful definition: warehouse lines are short-term loans, usually collateralised by the assets awaiting securitisation. The capital-markets exit repays the line and allows the cycle to begin again. Atlas occupies this position. In an Apollo presentation, its chief executive says Atlas provides asset-based warehouses and loans, followed by capital-markets services that term out those warehouses. The cited activities include commercial and residential real estate, corporate and consumer debt, and more specialised asset classes. The description comes from Atlas and Apollo; it explains their model, not an independent assessment of its quality. Credit Suisse transfers a platform and an obligation Atlas took its current form in February 2023. Apollo and Credit Suisse announced the first close of the sale of a significant part of the Securitized Products Group. A majority of the associated assets and professionals became part of or managed by Atlas. Athene's 2022 Form 10-K, filed after the first closes, described the initial consideration: approximately $400 million in cash and a portfolio of senior-secured warehouse assets, subject to debt, with approximately $1 billion of tangible equity value. The consideration was not paid immediately. Atlas accepted a $3.3 billion deferred purchase obligation to Credit Suisse. In March 2024, the management agreement for Credit Suisse's retained portfolio ended. Apollo's 2025 Form 10-K says Atlas then gave up $800 million of future fees and the obligation was reduced by the same amount to $2.5 billion. Strategic investors made equity commitments covering part of that obligation. Apollo says the assets received were senior secured, carried industry-standard loan-to-value ratios and were structured to investment-grade-equivalent criteria. Those features can reduce expected loss. They do not disclose which assets remained at March 2026, the haircuts applied to them, their maturities or the debt funding them. Apollo's own filing writes the stress scenario The most direct document is the Apollo IG Core Replacement Form 10 filed with the SEC on 1 August 2025. It is not outside criticism. It is the risk disclosure of a fund managed by Apollo. The filing says Atlas provides shorter-duration warehouse facilities whose exit depends on periodic securitisations and sufficient investor demand, including demand from Athene. Under capital-markets dislocations, the fund's ability to exit those short-term financings could be adversely affected. It also says committed backstops may be provided by Apollo clients. The same passage discloses the potential conflict. Apollo has an ownership interest in Atlas and may be incentivised to support its revenue growth, source additional investment-grade opportunities for Athene and Athora, and increase fees. An Apollo fund could receive a material part of its portfolio from Atlas. As majority shareholder, Apollo may receive a substantial share of Atlas compensation without that amount offsetting the management fees charged to the fund. These warnings prove neither a poor allocation nor a loss. They establish three dependencies: 1. the warehouse exit relies on buyers of longer-term securities; 2. entities in the same ecosystem can be originator, arranger, investor or backstop provider; 3. Apollo can earn revenue at several stages. The 2025 10-K provides another signal without isolating an Atlas dollar amount. Apollo attributes part of an increase in management fees to several vehicles and strategies, including Atlas. It only says the Atlas increase was driven by higher fee-generating assets after warehouse facilities expanded. The $342 million increase in that discussion covers a group of strategies, not Atlas alone. Assigning the entire amount to Atlas would be false. Three measures and no honest total Athene's accounts show a substantial Atlas exposure under several different definitions. At 31 December 2025, Athene held $5.679 billion of available-for-sale securities issued by Atlas or its affiliates and had $1.833 billion of additional commitments. Its Form 10-Q at 31 March 2026 put the two figures at $6.146 billion and $1.343 billion respectively. The concentration table in the same 10-Q separately lists $3.325 billion of investments in Atlas Securitized Products Holdings and $1.964 billion in Atlas Secured Advance Funding. Its footnote says each line measures single-issuer risk and may represent only part of total related-party exposure. Adding $6.146 billion, $1.343 billion, $3.325 billion and $1.964 billion would be tempting and wrong. Atlas and Atlas Secured Advance Funding securities may be included in the "Atlas or affiliates" aggregate; commitments are not assets already funded; and the tables serve different accounting purposes. AFS securities increased by $467 million between December and March. Commitments fell by $490 million. These movements could reflect funding, repayments, purchases, sales or reclassifications. The public tables do not provide a complete bridge between the dates. We therefore do not turn the lower commitments into a certain purchase or the higher securities balance into certain new production. The guarantee reaches the insurer The deferred obligation to Credit Suisse follows a disclosed chain. It is an obligation first of Atlas, then Apollo/Athene Dedicated Investment Program, known as AAA, Apollo Asset Management, Athene Holding and finally Athene Annuity Re. Apollo Asset Management and Athene Annuity Re each issued an assurance letter guaranteeing the full $2.5 billion. At 31 March 2026, Athene judged payment under its guarantees not probable and therefore recorded no liability. That accounting conclusion is important evidence against an alarmist reading. It does not cancel the contractual guarantee; it says the threshold for recognising a liability was not met at that date. Most importantly, the guarantee has a precise perimeter. It covers the deferred purchase price owed to Credit Suisse. It does not guarantee every Atlas loan, every warehouse or every security Athene holds. Treating it as a general umbrella would misstate the mechanism. Four risks in transmission order 1. Exit risk. A delayed securitisation extends the warehouse. This is not yet a credit loss, but capital and funding remain tied up. The Apollo IG Core prospectus explicitly identifies demand for securities, including Athene demand, as an exit condition. 2. Funding risk. The Federal Reserve's interagency statement on funding and liquidity risk lists disruption of warehouse funding among sources of unexpected funding needs. It notes that collateral deterioration can trigger higher margin or collateral requirements and that collateral values should be stress-tested. 3. Valuation risk. If an asset cannot be sold at the expected price, the question moves from "when will it exit?" to "what is it worth?" Senior protection depends on initial haircuts, overcollateralisation, position in the capital structure and first-loss capital held elsewhere. The public does not have a consolidated inventory of Atlas loans, valuations, haircuts and first-loss positions. 4. Circularity risk. Atlas originates and structures; Apollo funds can provide a warehouse or backstop; Athene and Athora seek investment-grade assets; Apollo earns fees. Each role can be economically rational. Their combination increases the importance of transfer pricing, independent allocation and the real ability of an affiliated buyer to say no. The IMF's October 2025 Global Financial Stability Report frames the sector issue: a growing part of insurer private-credit exposure takes the form of asset-backed structures, fund financing and private placements, while origination through affiliated managers requires special attention to conflicts and transparency. The finding is not specific to Atlas and proves no Athene loss. It explains why the architecture requires more scrutiny than a credit rating alone. A teaching scenario, not a forecast The following is an illustrative mechanism with no Atlas amount. A finance platform produces receivables and funds them in a warehouse. The securitisation market temporarily closes. The receivables continue to pay, but the facility reaches renewal before they can be sold. The funder agrees to extend at a higher haircut. More capital must be posted or the pool reduced. If some loans deteriorate at the same time, collateral value falls and junior protection starts to be consumed. An affiliated fund or insurer may buy a senior tranche or provide a backstop, if its limits allow. The first event in this scenario is a liquidity strain. It becomes an economic loss only if carrying cost, haircut or defaults exceed the protections. It reaches Athene according to the specific securities and ranks it holds. It activates the Credit Suisse guarantee only if the deferred purchase obligation is not met under its terms. These paths can occur separately. Evidence against an alarmist reading Several facts weaken the claim that a crisis is already under way. Atlas says it focuses on senior-secured assets, industry-standard loan-to-value ratios and investment-grade-equivalent criteria. Strategic investors, including an Abu Dhabi Investment Authority subsidiary announced in June 2023, diversify the capital beyond the Apollo group alone. Athene judges guarantee payment not probable. Finally, none of the documents reviewed says Atlas currently cannot sell or refinance its assets, or that Athene has suffered an Atlas loss. The structure also answers a real economic need. Warehouse funding gives nonbank originators access to financing before they have assembled a securitisable portfolio. Apollo says its sixteen platforms originated approximately $309 billion of assets in 2025. That number covers Apollo's entire origination ecosystem, not Atlas alone. These counterpoints rule out a conclusion that the structure is a time bomb. They do not eliminate risk: investment-grade status guarantees neither permanent liquidity, absence of conflicts nor stable valuation. Unknowns that keep the investigation open Public disclosures do not provide: - an aggregate inventory of assets held or financed by Atlas by class, vintage and performance; - a consolidated maturity schedule for warehouse facilities and the debt funding them; - transaction-level haircuts, margin calls, triggers and backstop commitments; - the exact rank of Athene's Atlas securities and the subordinated protection beneath each exposure; - a reconciliation between the $6.146 billion of Atlas or affiliate securities and the single-issuer concentration lines; - Atlas's share of Apollo's $309 billion of origination; - an isolated Atlas fee amount; - the pricing terms used when one Apollo entity sells, finances, structures or allocates an asset to another. The absence of these data proves neither concealment nor poor quality. It prevents an independent stress-loss calculation and a certain identification of the first-loss holder. A dashboard for the next dislocation The investigation can be updated each quarter from public indicators: 1. Atlas or affiliate AFS securities held by Athene; 2. Athene's additional commitments to Atlas; 3. single-issuer concentrations, including Atlas Secured Advance Funding; 4. status, amount and accounting treatment of the Credit Suisse guarantee; 5. Atlas fee-generating assets and Apollo's fee commentary; 6. references to backstops, extended financing or securitisation demand in Apollo fund filings; 7. impairments, expected credit losses, downgrades and collateral changes affecting the relevant securities. A functioning warehouse need not produce a spectacular signal. The first sign may be quieter: commitments becoming securities, duration extending, a new financing entity appearing in the concentration table, or a change in the accounting assessment of a guarantee. The public record's conclusion Atlas is not an accounting black hole. Athene discloses securities, commitments and the Credit Suisse guarantee. Apollo discloses conflicts and the risk of a delayed securitisation exit. The problem is fragmentation: each piece appears in a different note and answers a different definition. The risk can be explained without drama. Atlas advances liquidity before a final buyer exists. If that buyer is late, time becomes a funding need. If collateral falls, the need becomes a call for capital. If protections run out, it becomes a loss. In a system where originator, manager, some funders and some buyers are affiliated, the central question is then: who can reject the price, and who bears the first dollar of loss? Public filings cannot yet answer that question transaction by transaction. They can identify the channel, the disclosed amounts and the variables to monitor. That is enough to establish a risk. It is not enough to announce a crisis. This investigation complements The balance sheet Apollo does not consolidate, which reconstructs the Athora relationship, Apollo, the triangular domino, on the bridges between manager, insurer and bank, and From the credit card to the annuity, on the path of consumer credit. For method, read the guides to private-credit risk, CLOs and leveraged loans and life-insurer soundness. Method and scope This investigation uses disclosures available and verified as of 31 July 2026. Atlas data is kept within its original date and definition. AFS securities, commitments, concentrations and guarantees are never added. The cited arithmetic is limited to simple differences between Athene's published amounts: 6.146 minus 5.679 equals $0.467 billion; 1.833 minus 1.343 equals $0.490 billion. Commercial statements by Apollo and Atlas are attributed to them. Prospectus risk factors describe possible events, not realised events. The teaching scenario is neither a forecast nor a loss estimate. No Apollo market price or valuation ratio is used. Primary sources 1. Athene Holding Ltd., 2022 Form 10-K, Atlas note: first closes, assets received, tangible value and initial deferred obligation. 2. Apollo Global Management, Atlas launch announcement dated 8 February 2023: purchase of part of Securitized Products Group and platform description. 3. Apollo Global Management, ADIA commitment announcement dated 7 June 2023: majority ownership by Apollo affiliates, strategic capital and warehouse capacity announced at that date. 4. Apollo Global Management, Atlas SP Partners presentation and Origination page: warehouse model, asset classes, sixteen platforms and 2025 origination. Commercial sources attributed to Apollo. 5. Apollo Global Management, Form 10-K for 2025, Atlas, Related Party Transactions and management-fee disclosures: obligation reduced to $2.5 billion, securities, commitments and facility expansion. 6. Athene Holding Ltd., Form 10-K for 2025, Investments, Related Party Transactions and Commitments and Contingencies: concentrations, Atlas or affiliate securities, commitments and guarantee. 7. Athene Holding Ltd., Form 10-Q at 31 March 2026, same notes: latest available snapshot of securities, commitments, concentrations and assurance letter. 8. Apollo IG Core Replacement, Form 10 filed 1 August 2025, Atlas risk factors: securitisation dependence, backstops, Athene demand, compensation and potential conflicts. 9. Federal Reserve, Report to the Congress on Risk Retention, 2010, sections on securitisation mechanics and warehouse lines. 10. Federal Reserve and federal banking agencies, Interagency Policy Statement on Funding and Liquidity Risk Management, updated after the 2023 stress: warehouse disruption, collateral and contingency funding plans. 11. International Monetary Fund, Global Financial Stability Report, October 2025, chapter 1, Insurance Companies: insurer private credit, asset-backed structures, concentration and conflicts involving affiliated managers. Limitations The disclosed amounts are accounting snapshots, not loss measures. A single-issuer concentration is not necessarily a first-loss exposure. An investment-grade security may become illiquid without defaulting. An unrecognised guarantee may remain legally effective. The analysis alleges neither fraud, insolvency nor imminent default. The investigation does not have individual contracts, collateral inventories, valuation models, allocation minutes or a consolidated Atlas stress test. It therefore maps the transmission channel and its unknowns without inventing the missing result. This analysis is not investment advice. Original l0g text, CC BY 4.0 licence. The three charts are l0g representations built exclusively from the primary sources cited. ============================================================================ ANALYSIS: The balance sheet Apollo does not consolidate URL: https://l0g.fr/en/analysis/the-balance-sheet-apollo-does-not-consolidate/ Canonical French source: https://l0g.fr/posts/bilan-apollo-non-consolide/ Date: 2026-07-30 (reviewed 2026-07-30) Topics: apollo, athora, insurance, private credit, governance, risk, united kingdom ---------------------------------------------------------------------------- On 1 January 2018, $6.3 billion of assets left Athene's consolidated accounts. The US insurer had just diluted its European subsidiary, renamed Athora, among private investors. Eight years later, the public record shows a separation that is legally real but economically incomplete. Apollo and its subsidiary Athene together hold 26% of Athora's shares, have five representatives on its board, Apollo manages or advises $57.2 billion of its assets, and Athora recorded €139 million of expenses with Apollo in 2025. Since March 2026, the acquisition of Pension Insurance Corporation has taken the Athora group to €139 billion of assets under management and administration. The balance sheet was not hidden. It was fragmented across companies, currencies, jurisdictions and reports. This investigation reconstructs the relationship document by document, and stops exactly where the documents stop. Deconsolidation, not disappearance Athora began as an Athene construction. The company was incorporated in Bermuda on 1 December 2014 as AGER Bermuda Holding Ltd. to hold the group's European operations. In 2017, it secured €2.2 billion of equity commitments from investors. On 1 January 2018, the capital increase closed, Athene fell to 10% of the voting power and less than 50% of the economic interest, and AGER became a related-party investment instead of a consolidated subsidiary. The company adopted the Athora name days later. The accounting effect is measurable. Athene's 2018 annual report filed with the SEC, in Athora Deconsolidation, says that $6.3 billion of total assets and $6.0 billion of invested assets left the consolidated balance sheet. Its third-quarter 2018 Form 10-Q states the new treatment: Athora became an alternative investment in a related party. This did not mean that the assets disappeared or that Athene ceased to be exposed. It changed the consolidation perimeter, the boundary inside which assets, liabilities, income and expenses are combined line by line. Outside that boundary, stakes, commitments and transactions remain disclosed, but in separate notes. This is where the investigation starts: reconnecting what accounting legitimately separated. Four links survived the separation The present relationship does not rest on a single shareholding. It is a bundle of four ties: capital, governance, asset management and insurance contracts. Capital. At 31 December 2025, Athora's 2025 Financial Condition Report, page 7, attributed to Apollo, including Athene, 24.51% of the economic interest and 26.00% of the voting power in the common share capital. After the capital raise for the PIC acquisition, Athora's 24 June 2026 offering memorandum, numbered page 122, provides a new snapshot: Apollo-managed funds excluding Athene hold 8% of the shares, Athene 18% and Abu Dhabi Investment Authority 16%, excluding treasury shares. Apollo and Athene therefore total 26% of shares. That measure cannot automatically be converted into voting power or economic interest, because the document does not do so. Governance. The same memorandum says Apollo has four elected directors and Athene one. At the end of 2025, the board had eleven members, including five independent directors, according to pages 15 to 19 of the condition report. Five seats document influence. They do not, on their own, establish that Apollo legally controls Athora. Asset management. Apollo's 2025 Form 10-K, under Athora, says its subsidiaries managed or advised $57.2 billion of Athora assets at 31 December, of which $55.2 billion was fee-generating. Within that total, $34.7 billion was classified as Athora Non-Sub-Advised Assets. The label can suggest the absence of a mandate. Apollo's definition says the opposite: these assets are managed by Apollo, but are neither explicitly sub-advised nor invested in Apollo funds or vehicles. Flows and commitments. Athora recorded €139 million of expenses with Apollo in 2025, with €35 million payable, against €148 million of expenses in 2024. The table appears on page 15 of the condition report. It aggregates expenses involving Apollo. It does not establish that the entire amount was asset-management fees. On Athene's side, Apollo's 10-K reports $1.487 billion of investments in Athora at the end of 2025 and $2.7 billion of additional commitments, mainly related to conditional support for the PIC purchase. The same filing mentions a conditional commitment of up to $2 billion made by Apollo Asset Management in July 2025. Public disclosures do not reconcile it sufficiently with Athene's $2.7 billion of commitments. Adding both figures would produce a precise-looking number that the evidence does not support. We do not add them. The 2018 alignment contract The separation came with a contract designed to preserve common interests. The preamble to the January 2020 amendment filed with the SEC says so explicitly: the cooperation agreement dated 1 January 2018 was entered into to maintain alignment between Athora and Athene following deconsolidation. The original agreement organised several possible exchanges. In the final public description, Apollo's 2025 10-K refers among other things to first-offer or first-refusal rights over certain reinsurance liabilities ceded to Athene, an overall cap equal to 20% of Athora liabilities for certain third-party flows, and potential purchases of Athene funding agreements, generally limited to 3% of the assets of each relevant subsidiary. Apollo says these rights had never been exercised when they ended on 5 August 2025. Their termination therefore matters: the contractual alignment created in 2018 no longer exists in that form. It does not remove the shareholding, board seats or asset-management mandates. A new contract appears at the reporting date. On page 15 of its condition report, Athora says Athora Life Re entered into a tail-risk retrocession treaty with Athene Annuity Re on a block of US-dollar whole-life business, effective 31 December 2025 and conducted on normal commercial terms. The documents reviewed disclose neither its notional amount nor its pricing formula. They also do not establish that it is funded reinsurance. Treating it as one of the British structures targeted by the PRA would be an extrapolation. Apollo's own description of the vehicle The most revealing words come from neither a critic nor a journalist. They appear in the Form 10 of Apollo IG Core Replacement filed with the SEC on 1 August 2025, on pages 211 and 212. This Apollo-managed fund is disclosing conflicts that could affect its own investors. It explains that Apollo provides asset-management services to Athene and Athora, allocates a significant part of their assets among its clients and often characterises them, in relation to its business, as "captive permanent capital vehicles". The passage adds that overlapping ownership and voting power mean Apollo is, or could be perceived to be, able to exercise significant influence over major decisions: corporate transactions, appointments, elections of directors, termination of investment-management agreements and corporate policies. This is not a judicial finding on control of Athora. It is a broadly drafted risk warning for an Apollo fund. Its documentary value lies elsewhere: it describes the economic purpose of the system from the manager's perspective. Long-dated insurance liabilities supply stable capital, Apollo teams allocate it across strategies and assets, and the mandates generate recurring revenue. The filing also lists possible conflicts: preferential terms for Athene or Athora, co-investments, cross-trades, allocation of opportunities, ownership of different tranches in the same structure and the possible substitution of insurer capital for a direct Apollo commitment. It does not prove that any particular transaction harmed Athora. It proves that the manager itself identifies these channels as conflicts requiring controls. Athora's balance sheet under the lens Athora's 2025 annual report shows the assets financing promises to policyholders. The group reports €75.547 billion of assets under management and administration, including €51.475 billion of general-account assets under management. The allocation on page 24 includes €14.6 billion of sovereign and supranational debt, €10.9 billion of traded corporate bonds, €9.3 billion of private credit, €7.2 billion of mortgages and savings mortgages, €4.8 billion of net derivatives and cash, €3.9 billion of alternatives and other assets, and €0.8 billion of investment property. The strategy is explicit. On pages 19 and 22, Athora says it seeks an illiquidity and complexity premium, notably through Apollo's origination capabilities, with an illustrative 25% to 35% allocation to private assets. This does not mean Apollo issued or owns every such asset. The June 2026 memorandum gives one useful boundary: debt and equity investments in Apollo-owned entities represent less than 2% of AuMA. That is a narrow measure. It does not disclose the share of assets originated, selected or managed by Apollo. Valuation difficulty appears in the IFRS hierarchy. On page 121 of the 2025 annual report, recurring assets at fair value total €87.969 billion: €53.719 billion in Level 1, €13.236 billion in Level 2 and €21.014 billion in Level 3. Our calculation, 21.014 divided by 87.969, gives 23.9% of assets whose valuation uses significant unobservable inputs. A Level 3 asset is not a hidden loss. It is an asset whose price requires more judgement. EY made the valuation of these €21.0 billion a key audit matter. On page 87, the auditor describes tests of models, assumptions, yields, spreads and samples, then concludes that the valuations were reasonable. The audit reduces the risk of material error. It does not create a continuous market price where none exists. Evidence against an alarmist reading A serious investigation must test its thesis against the evidence that weakens it. Such evidence exists. The first item is prudential. Athora's 2025 report shows €6.392 billion of available statutory capital against an enhanced capital requirement, or ECR, of €3.280 billion, a solvency ratio of 195%. This is a regulatory snapshot, sensitive to models, rates and management actions. It nonetheless represents a substantial buffer at the reporting date. The second concerns published quality. Athora says 98% of its traded corporate-bond portfolio is investment grade and 87% of its government debt is rated A or better. For private credit, the June 2026 memorandum says 99% is senior debt and reports €129 million of cumulative realised losses on €18.493 billion of gross deployments since 2018. The simple ratio between both amounts is 0.70%. It is not an annual default rate or a total economic loss: the denominator cumulates gross flows, while the numerator excludes, among other things, unrealised losses. The figure is nevertheless inconsistent with a portfolio that has already crystallised massive losses. The third item is institutional. The offering memorandum describes a conflicts committee made up of the five independent directors and the ADIA appointee, plus a related-party transactions policy requiring arm's-length terms and approval procedures. The same document candidly warns that these mechanisms may be insufficient and that an unmanaged conflict could materially damage the business or its reputation. These protections do not disprove the conflicts. They show that Athora identifies them and has designed a process to handle them. The public record does not show deliberations, pricing comparables or transaction-level votes. Whether the countervailing power works remains a question to test, not an available conclusion. PIC changes the scale and the supervisor On 27 March 2026, Athora completed the acquisition of Pension Insurance Corporation Group for an announced price of about £5.7 billion. PIC insures the defined-benefit pensions of British companies. The transaction took Athora to €139 billion of AuMA and 3.1 million policyholders, according to the completion release. PIC represents about 45% of the new group. The financing combines equity and debt. Athora announced on 6 March 2026 €3.5 billion of new common-equity commitments from existing and new investors, including Apollo and Athene. Its offering memorandum also discloses a term facility of up to £2.2 billion, of which £1.6 billion was drawn at 31 March, and a €1.635 billion revolving credit facility, of which €255 million was drawn. The prudential centre of gravity is meant to follow. Athora plans to move its headquarters to the United Kingdom by late 2027, subject to approvals, and says the PRA is ultimately expected to become the group supervisor after a transition with the Bermuda Monetary Authority. This shift arrives as the PRA sharpens its view of funded reinsurance. In its CP8/26 consultation dated 29 April 2026, the authority estimates that about 15% of recent new UK pension risk-transfer business has been ceded through that channel. It says the current treatment can understate risk and create a capital advantage, particularly where counterparties are credit-focused and collateral includes private credit. This sector context does not prove that PIC uses Athora or Athene in this way. It identifies the prudential test the enlarged group will face. The exact perimeter of PIC assets entrusted to Apollo, the associated fee schedule and the future share of related-party reinsurance are not detailed in the public documents reviewed. They are the principal unknowns after the acquisition. EIOPA asks the precise independence question The issue is no longer confined to the United States or Bermuda. On 3 February 2026, EIOPA opened a consultation on the supervision of insurers related to private equity. The consultation closed on 30 April, so the text is not a final rule. The draft supervisory statement, pages 8 to 10, asks authorities to monitor asset-management agreements, reinsurance and outsourcing, check the fairness of fees and verify that investment decisions remain independent when an affiliated asset manager is also a shareholder. It also recommends examining significant influence regardless of the amount of equity or voting rights, including influence exercised through special rights. Applied to Athora, this framework changes the question. Asking whether Apollo has a majority is not enough. The relevant test is whether Athora has effective countervailing power over assets, fees, reinsurance contracts and cross-transactions. The reports establish that committees and independent directors exist. They do not disclose the material needed to evaluate their operational independence. Seven indicators for a continuing investigation The architecture is observable. Its development can therefore be tracked without speculation. 1. The PIC mandate. What share of PIC assets will Apollo manage, advise or sub-advise, and at what price? 2. Expenses with Apollo. Do the €139 million recorded in 2025 rise after the acquisition, and faster or slower than the related assets? 3. Origination of private assets. Will Athora publish the share of its private credit originated by Apollo, separately from investments in Apollo-owned companies? 4. Transactions with Athene. What volumes, collateral, prices and capital effects will attach to future reinsurance or funding contracts between both insurers? 5. Level 3. Does the €21.014 billion increase, and how do impairments, defaults and disposals compare with previous valuations? 6. PIC financing. Do acquisition debt and drawn credit lines reduce holding-company flexibility or its ability to support subsidiaries under stress? 7. Countervailing power. Do future reports disclose more about conflicts-committee decisions, fee-comparison methods and transactions rejected or modified? These questions make the investigation falsifiable over time. Detailed disclosure of mandates, pricing, asset origination and independent decisions would reduce opacity risk. A rise in affiliated flows without equivalent disclosure would increase it. The public record's conclusion The 2018 deconsolidation is real. Athora owns its companies, publishes its accounts, raises its own capital and answers to its supervisors. The documents reviewed prove neither fraud, an abusive transaction nor hidden legal control by Apollo. They establish something else with enough precision to matter: Athora remains linked to Apollo through a combined 26% shareholding with Athene, five board seats, $57.2 billion of assets managed or advised, €139 million of published annual expenses, Athene investments and commitments, and related-party insurance contracts. Apollo itself characterises Athora, relative to its business, as a captive permanent capital vehicle and acknowledges the conflicts this proximity can create. The distinction between consolidation and influence is the heart of the investigation. Consolidation is a binary accounting rule. Influence runs by degree through mandates, seats, contractual rights, origination and financing. With PIC, this architecture now carries a significant part of the British pensions market. The relevant risk is not that a balance sheet vanished. It is that a reader, policyholder or investor stops at the accounting boundary and never reconstructs the economic system continuing beyond it. For the wider l0g corpus, read Apollo, the triangular domino, the investigation into Atlas risk when the warehouse does not clear, our work on life insurers, private credit and Bermuda, the lender of next-to-last resort and Private credit, one asset, two prices. The methods are set out in the guides to reading life-insurer health and reading private-credit risk. Method and scope This investigation relies on public documents reviewed or downloaded on 30 July 2026. Athora data is presented in the currency, date and perimeter of its source. Apollo AUM in dollars is not divided by Athora AuMA in euros because the definitions and scopes do not coincide. l0g calculations are limited to two simple ratios disclosed in the text: 21.014 / 87.969 for the Level 3 share and 129 / 18,493 for the cumulative realised-loss share announced against gross deployments. A statement by Athora or Apollo is attributed to its issuer. The audit report establishes the auditor's procedures and conclusion, not a guarantee of future value. The EIOPA and PRA texts are consultations, not final decisions. The argument requires no market prices or listed-company ratios. Primary sources 1. Athene Holding Ltd., 2018 Form 10-K, Athora Deconsolidation: history and the removal of $6.3 billion of total assets and $6.0 billion of invested assets. 2. Athene Holding Ltd., third-quarter 2018 Form 10-Q, Deconsolidation note: 10% of voting power, less than 50% economic interest and treatment as a related-party investment. 3. Athora and Athene, Cooperation Agreement dated 1 January 2018 and amendment dated 7 January 2020: funding, reinsurance, cooperation and post-deconsolidation alignment. 4. Apollo Global Management, Form 10-K for 2025, definitions of Athora and Athora Non-Sub-Advised Assets, note 18 Related Party Transactions: AUM, fees, capital, commitments and termination of the cooperation agreement. 5. Apollo IG Core Replacement, Form 10 filed on 1 August 2025, pages 211-212, Strategic Relationship with Insurance Businesses: influence, allocation, permanent capital vehicles and potential conflicts. 6. Athora Holding Ltd., 2025 Financial Condition Report, pages 5, 7 and 15-19: solvency, ownership, board, Apollo expenses, termination of the cooperation agreement and retrocession treaty. 7. Athora Holding Ltd., 2025 Annual Report, pages 19, 22-26, 87 and 121-124: strategy, allocation, private credit, audit and fair-value hierarchy. 8. Athora Holding Ltd., Offering Memorandum dated 24 June 2026, sections Risk Factors, Shareholders, The PIC Transaction and Private credit; the ownership snapshot is on numbered page 122: PIC financing, ownership, governance, conflicts and private credit. 9. Athora Holding Ltd., €3.5 billion capital raise and regulatory-approval announcement, 6 March 2026. 10. Athora Holding Ltd., PIC acquisition completion and planned headquarters move, 27 March 2026. 11. EIOPA, consultation page and draft supervisory statement, pages 8-10, consultation opened on 3 February and closed on 30 April 2026. 12. Prudential Regulation Authority, CP8/26, Funded reinsurance, 29 April 2026: share of new BPA business ceded, counterparty risk, private collateral and proposed prudential treatment. Limitations The public documents do not disclose the exact share of Athora assets originated by Apollo, the breakdown of the €139 million of expenses, the future Apollo mandate over PIC, the full economics of the December 2025 retrocession treaty or minutes of conflicts-committee decisions. The absence of these data does not prove an anomaly. It prevents a complete test. Solvency and portfolio-quality figures are point-in-time data published by Athora and, for the accounts, audited within the described perimeter. They do not prejudge resilience to every scenario. Conversely, Level 3 assets and related-party transactions do not prove overvaluation or harm. This analysis is not investment advice. Original l0g text, CC BY 4.0 licence. The three charts are l0g representations built exclusively from the primary sources cited. ============================================================================ ANALYSIS: The Treasury toll: who controls access to clearing? URL: https://l0g.fr/en/analysis/the-treasury-toll-access-to-mandatory-clearing/ Canonical French source: https://l0g.fr/posts/le-peage-du-tresor-acces-clearing-obligatoire/ Date: 2026-07-30 (reviewed 2026-07-30) Topics: treasuries, clearing, liquidity ---------------------------------------------------------------------------- Beginning December 31, 2026, an additional share of cash Treasury transactions must pass through a clearing house. Repo follows on June 30, 2027. The reform promises less bilateral risk and more netting. It also creates a new gateway: a client that is not a direct member needs an intermediary able to carry its margin, liquidity obligations and, depending on the model, its guarantee. A survey published by DTCC on July 27 describes an industry that is largely prepared, but says only about one-third of responding members expect to offer clearing to clients. That figure does not prove there will be a bottleneck. It identifies where to look for one. The Treasury market is familiar ground for l0g. We have examined the leverage in the basis trade, the creation of liquidity in repo and the chains of collateral and rehypothecation. Mandatory clearing is often presented as the regulatory answer to those vulnerabilities. This article examines the answer itself. The question is not whether central clearing is wholly good or bad. It is more concrete: who provides access to the system, who ties up the collateral, and who must find the cash when volatility triggers margin calls? The reform does not remove intermediaries The rule adopted by the SEC in December 2023 requires direct participants of a covered clearing agency to submit their eligible Treasury transactions for central clearing. After a one-year delay, compliance dates are December 31, 2026 for the cash market and June 30, 2027 for repo. The SEC's Treasury clearing implementation hub, updated on July 24, 2026, now lists three clearing agencies registered for Treasuries: FICC, CME Securities Clearing and ICE Clear Credit. A central counterparty, or CCP, interposes itself between both sides. It becomes the buyer to every seller and the seller to every buyer. That novation makes it possible to net offsetting positions and reduce direct counterparty exposures. It does not give every fund, foreign bank or asset manager direct access. At FICC, a firm that cannot become a full-service member may use a Sponsoring Member or an Agent Clearing Member. The first sponsors a Sponsored Member. The second submits trades for an Executing Firm Customer. In both cases, a regulated intermediary remains between the client and the clearing house. The market says the market is ready The immediate trigger is the report published by DTCC on July 27. In June, FICC surveyed all full-service Netting Members of its Government Securities Division. The response rate was 92%. The results describe an advanced transition: - more than $1.2 trillion in cash Treasury activity is already centrally cleared at FICC each day; - respondents report another $300 billion to $400 billion in daily par value not currently submitted for clearing; - 79% say they already have the necessary FICC account setups; - approximately one-third expect to offer cash Treasury clearing to clients. Those four figures do not measure the same thing. The 79% describes respondents' readiness for their own requirements. The one-third concerns firms expecting to become access providers for clients. The proportions cannot be subtracted, and they do not imply that two-thirds of the market will be excluded. The report has useful limitations. DTCC operates FICC and is presenting its own readiness. It discloses neither the raw number nor the identities of respondents, nor the amount of client capacity each firm will provide. It does not report prices, commercial requirements, rejected onboarding requests or the future concentration of volumes. A 92% response rate makes the survey informative. It does not turn an infrastructure operator's survey of its members into a complete picture of competition. The one-third figure is therefore a signal of possible concentration, not proof of an oligopoly. The client clearer carries the hidden bill FICC documents make the responsibility chain traceable. Under the Agent Clearing Service, the Agent Clearing Member is responsible to FICC for fees, settlement, margin, liquidity obligations and any loss allocation attributed to submitted activity, including client trades. Under the Sponsored Service, the Sponsoring Member guarantees to FICC the obligations of its Sponsored Members. It also carries Clearing Fund deposits associated with the omnibus account, calculated twice daily on a gross basis. The client retains its own legal obligations, but FICC can look to the sponsor if those obligations are not fulfilled. This shift has three consequences. First, access has a price. An intermediary mobilizes capital, systems, staff, collateral and liquidity lines. The amount charged to the client may reflect those costs even when netting reduces balance-sheet use elsewhere. Second, capacity is not unlimited. A dealer can be ready for its proprietary activity without accepting every fund seeking client access. It must set limits, manage default risk and forecast stressed margin needs. Third, competition cannot be read from account counts. Three clearing agencies are registered and FICC offers several models. Effective competition will depend on volumes, interoperability, carrying costs and the ability to transfer a client's positions if its intermediary defaults. Credit risk becomes a liquidity clock Central clearing reduces bilateral exposure but imposes a timetable. Initial margin protects against a future price move during the closeout of a default. Settlement payments and margin calls require cash at the specified time. FICC's Disclosure Framework for the first quarter of 2026 says FICC does not rely on routine access to central bank credit in its liquidity planning. If a net-buying member defaults, FICC must still receive the securities and pay the corresponding cash. It draws on liquid resources and can redistribute securities to members through repo under the Capped Contingency Liquidity Facility, or CCLF. An Office of Financial Research paper explains why this matters especially for physically settled securities and repo. Those CCPs tend to need more liquid resources than some derivatives clearing houses because the full settlement value must move. Credit lines supplied by members distribute that funding requirement, but they also reconnect the resilience of the clearing house to participant liquidity. On July 1, FICC lowered from 30% to 10% the buffer parameter used to size the aggregate GSD CCLF. The same notice says individual caps were reset using needs observed between January and June and all other parameters remained unchanged. The notice does not show that FICC's total liquidity fell by 20%, nor that the clearing house became less resilient. It does not disclose the aggregate dollar amount before and after recalibration, and the underlying requirement may change. It does show that member-provided liquidity capacity is an active variable in the transition, not a detail settled once and for all. The Federal Reserve's May 2026 Financial Stability Report provides a reassuring counterpoint. During volatility linked to the conflict with Iran, CCPs raised margins significantly on energy products, with no observed difficulty for participants. The Fed described prefunded resources as high. That is evidence of resilience during that episode, not a test of the future Treasury migration. The balance-sheet gain exists, but its size is disputed The main economic benefit expected from clearing is multilateral netting. A dealer lending cash on one side and borrowing it on another can offset more positions when the same CCP is its counterparty. Using data across all repo segments, the OFR estimates that 77% of daily repo would have been centrally cleared during the first eight months of 2025 had the rule applied, compared with an observed 45%. For six U.S. global systemically important banks, the counterfactual reduces non-netted repo and reverse repo positions by $207 billion, or $34.5 billion per bank on average. The calculation assumes transactions and behavior do not change. It is neither a volume forecast nor a profit estimate. A Federal Reserve research paper, using a different dataset to answer a different question, finds that the effect of clearing on the supplementary leverage ratio should be relatively limited. Some transactions are already structured to net outside a CCP, while others would not automatically become nettable. Client margin adds another qualification. A Federal Reserve note on repo explains that the CCP charges the direct member, not necessarily the end client. The member then sets its own requirements for that client. Clearing may standardize margin at the clearing-house level without immediately standardizing the haircut or price paid by each fund. The toll risk is a falsifiable hypothesis The central hypothesis can be stated without drama: if a limited group of intermediaries concentrates client access, dispersed bilateral risk could be replaced by common dependence on a small number of clearing capacities. Those intermediaries could then influence the price, limits and terms of access to the financial system's most important market. Several facts prevent that scenario from being presented as established. FICC reports more than 2,850 Sponsored Members across 66 jurisdictions and more than $2.5 trillion in daily Sponsored Service volume. The Agent Clearing Service is growing. CME Securities Clearing and ICE Clear Credit provide additional options. Finally, one-third of respondents offering the service may be enough capacity if those firms are large, diversified and genuinely competitive. Five pieces of evidence matter more than the narrative: 1. The effective concentration of client volume, by clearing house and intermediary, rather than the number of accounts opened. 2. Access prices and terms, including fees, haircuts, margin calls, minimum thresholds and rejected onboarding. 3. Liquid resources and the CCLF, including aggregate amounts, tiered member contributions and intraday calls during volatility. 4. Porting capacity, meaning whether client positions can actually move quickly after a sponsor or agent defaults. 5. Market quality around the deadlines, measured through bid-ask spreads, settlement fails, depth and repo behavior. The first deadline will test more than whether the software works. It will show whether central clearing produced a more open infrastructure or a more concentrated toll gate. Reducing one risk can create a new dependency Central clearing addresses a real problem. It makes exposures more visible, imposes structured margining and nets flows that currently consume balance sheet. In a market where the Treasury is issuing at record scale and hedge funds rely heavily on repo, those gains can improve resilience. But the outcome cannot be read from cleared volume alone. A reform can reduce counterparty risk while concentrating operational risk, liquidity and access power. It can free balance sheet in normal conditions and demand more cash at the worst moment. It can protect the clearing house while sending the bill back to the client through its clearing intermediary. Mandatory clearing is not the end of the collateral story. It is a change of address. Beginning in December, the market must prove that the new vault has enough doors, enough liquidity and enough competitors. Sources 1. DTCC/FICC, Industry Readiness for U.S. Treasury Cash Clearing: A Survey of FICC Membership, June survey, 92% response rate, volumes, readiness and client offering, July 27, 2026: 2. DTCC, Market Participant Firms Making Significant Progress Toward U.S. Treasury Cash Clearing Deadline, release accompanying the survey, July 27, 2026: 3. SEC, Treasury Clearing Implementation, rule, compliance dates, guidance, FICC actions and clearing-agency registrations, updated July 24, 2026: 4. SEC, SEC Adopts Rules to Improve Risk Management in Clearance and Settlement and Facilitate Additional Central Clearing for the U.S. Treasury Market, rule adoption, December 13, 2023: 5. SEC, SEC Extends Compliance Dates and Provides Temporary Exemption for Rule Related to Clearing of U.S. Treasury Securities, dates moved to December 31, 2026 and June 30, 2027, February 25, 2025: 6. FICC, Client Clearing Capabilities for Treasury Market Activity, Sponsored and Agent Clearing models and margin, settlement, guarantee and liquidity responsibilities: 7. FICC, Disclosure Framework for Covered Clearing Agencies and Financial Market Infrastructures, first quarter of 2026, access, collateral, liquidity, defaults and CCLF: 8. FICC, CCLF Liquidity Buffer Parameter Adjustment, notice GOV2174-26, parameter lowered from 30% to 10% on July 1, 2026: 9. Office of Financial Research, How Will Central Clearing Impact the Repo Market?, repo data, clearing counterfactual and estimated balance-sheet effect, January 29, 2026: 10. Office of Financial Research, John Heilbron and Nick Schwartz, Central Counterparty Management of Liquid and Prefunded Resources, liquid resources at securities and repo CCPs, March 5, 2026: 11. Federal Reserve, Sriya Anbil, Mark Carlson, Christopher Han and John Wang, Balance-Sheet Netting in U.S. Treasury Markets and Central Clearing, FEDS 2024-057, July 2024: 12. Federal Reserve, Sebastian Infante, R. Jay Kahn, Luke M. Olson and Mary-Frances Styczynski, Proportionate margining for repo transactions, February 14, 2025: 13. Federal Reserve, Financial Stability Report, CCP margins and prefunded resources during the energy shock, May 8, 2026: 14. CPMI-IOSCO, Streamlining variation margin in centrally cleared markets: examples of effective practices, liquidity and predictability of margin calls, January 15, 2025: ============================================================================ ANALYSIS: Warsh removes the compass: the cost of a Fed without guidance URL: https://l0g.fr/en/analysis/warsh-fed-without-guidance-uncertainty-cost-july-fomc/ Canonical French source: https://l0g.fr/posts/warsh-fed-sans-guidance-cout-incertitude-fomc-juillet-2026/ Date: 2026-07-29 (reviewed 2026-07-29) Topics: fed, warsh, fomc, rates, bonds, inflation ---------------------------------------------------------------------------- The Federal Reserve did not raise its policy rate on July 29. Financial conditions nevertheless tightened between its last two meetings, especially at longer maturities. Kevin Warsh sees this partly as the desirable result of a central bank that guides markets less. The proposition is coherent, but not yet demonstrated. With three members now calling for a hike, reduced guidance may also make the reaction function of a divided committee harder to read. The decision itself is straightforward. The FOMC held the federal funds target range at 3.50-3.75% and maintained an ample-reserves regime. The statement passed by 9 votes to 3. Beth Hammack, Neel Kashkari and Lorie Logan preferred a 25-basis-point increase. Six weeks earlier, the same hold had passed 12-0. This break extends our analysis of Warsh's first FOMC, but it moves the subject forward. The balance-sheet regime examined in Warsh's first battlefield remains operationally intact. The energy shock described before the meeting in the Fed trapped by the barrel has not disappeared. The new fact on July 29 lies elsewhere: the Fed chair is treating uncertainty about the path as a way to make markets less dependent on the committee's words. An unchanged rate conceals a broken consensus The rate did not move, but the vote changed character. In the June minutes, all participants still supported the hold. A few already saw a case for raising the range, yet stopped short of voting against the decision. In July, three regional-bank presidents crossed that threshold. A dissent is neither a forecast nor a commitment for September. It records a preference based on the information available on the meeting date. The three-vote shift nevertheless shows that June's consensus was conditional. The disagreement is less about the diagnosis than the action threshold: how many more months of inflation above 2% can the committee tolerate while it regards the labor market as stable? The pre-meeting record explains the divide without requiring speculation. On July 13, Governor Christopher Waller said another high core-inflation reading could warrant near-term tightening, while asking for several observations before reaching a conclusion. The next day's CPI fell 0.4% in June; it was still up 3.5% from a year earlier, with core at 2.6%. Two days after that release, Lorie Logan continued to argue for "modestly higher" rates, judging that underlying inflation was heading toward the mid-2s rather than all the way to target. The July vote turns this difference in thresholds into policy dissent. Warsh's doctrine: forecast less, observe more In his opening statement, Warsh describes the policy release as a statement of facts that deliberately avoids forecasting the path. His argument is that markets should react directly to economic information rather than wait for a rolling Fed scenario or constant validation. Market participants, in his phrase, should "play the ball, not the referee." The retreat is not improvised. On July 9, the Fed established five task forces, including one on communication under uncertainty and another on balance-sheet policy. In Congress, Warsh also stressed the need to revisit the institution's models, tools and methods. The FOMC retains its 2% goal, its statement, recorded votes and quarterly Summary of Economic Projections. It is not going silent. It is offering less information about the likely rate path. That distinction matters. Forward guidance is not necessarily a promise about future rates. In its careful form, it maps a reaction function: if inflation, employment or risks move in a given direction, policy should respond in a given way. Withdrawing it protects the central bank from false precision and makes investors do more work. It can also make the threshold for committee action harder to identify. The market tightened without a policy-rate hike Warsh offers a measurable fact in support of his doctrine: nominal and real yields increased between the two meetings. He says reduced guidance may have contributed and presents the move as an improvement in price discovery. Treasury data confirm the move, with an important qualification. Between the June 17 and July 29 closes, the nominal 2-year yield rose only 2 basis points, from 4.20% to 4.22%. The increase then grows with maturity: 10 basis points at 5 years, 18 at 10 years, 26 at 20 years and 27 at 30 years. On the real curve derived from inflation-protected securities, increases range from 16 to 25 basis points at common maturities. The comparison between nominal and real curves is revealing. At 10 years, both rise by 18 basis points. At 30 years, the nominal yield gains 27 points and the real yield 25. In these data, the long-rate increase is not principally a story of exploding inflation compensation. It is consistent with higher expected real rates, a larger term premium, stronger expected real growth, greater debt supply, or some combination of these forces. Subtracting a real par curve from a nominal par curve gives only an approximation of inflation compensation: the instruments, cash flows and liquidity differ. The comparison is sufficient here to show that the real move is similar in size to the nominal move, not to calculate an exact breakeven rate. The curves cannot isolate the communication effect. Investors also received employment and inflation releases, news about the Middle East conflict, fiscal information and new signals on artificial-intelligence investment between the two meetings. Warsh himself describes reduced guidance as a possible factor, not the sole cause. Assigning 18 or 27 basis points to his communication would go beyond the evidence. The possible price of uncertainty Economic research gives reasons to take the communication channel seriously without assigning it false precision. A Bank for International Settlements working paper covering several central banks estimates that a change in forward guidance moves professional rate forecasts by 5 basis points on average in the intended direction. A study published in the American Economic Review finds that the type of FOMC language changes how the private sector interprets a lower expected policy path: as news about the economy or as policy inclination. These results do not imply that a central bank should always provide more guidance. Guidance that looks too precise can create false certainty, be mistaken for a commitment and weaken price discovery. They establish only that saying less also changes prices. No signal is not neutral. A structural model published by Federal Reserve researchers attributes about half of the long-horizon variance of long-term nominal yields in its sample to uncertainty shocks, operating through risk premia and expected future real rates. This is a model result for a historical period, not an estimate for July 2026. It establishes a possible channel, not the cost of Warsh's strategy. July's specific risk comes from combining less path information with more internal dispersion. With a unanimous committee, markets may reconstruct a likely response from the data. With three votes for a hike, an oil supply shock, inflation still above target and conflicting views on how restrictive policy is, the same release can lead members toward different decisions. Markets must estimate not only the economy, but also the coalition that will prevail. That uncertainty can raise the compensation investors demand at longer maturities and tighten financial conditions without a formal rate hike. Warsh appears willing to accept this effect. The unanswered question is whether its cost rewards better economic information or a less legible reaction function. Three tests for Warsh's wager The argument can be made falsifiable. Three observations can help separate better price discovery from an uncertainty tax. First, the dispersion of expectations. If professional rate forecasts and scenario probabilities remain more dispersed for comparable data, reduced guidance will have increased policy uncertainty. If dispersion stays stable, the hypothesis weakens. Second, the composition of yield increases. A move concentrated in real yields and the term premium, without a lasting increase in inflation expectations, would indicate a broader tightening of financial conditions than a simple inflation scare. The l0g guide to reading the Treasury market explains that decomposition. Third, the September 15-16 meeting. It will bring a new Summary of Economic Projections and dot plot. Converging projections and votes would make July's fracture episodic. Wider dispersion with further dissents would show that Warsh leads a committee whose path can no longer be summarized by the chair's words alone. The most important change on July 29 is therefore not a rate left at 3.50-3.75%. It is the coexistence of three features: a reaffirmed inflation target, a more divided committee and less prescriptive communication. Warsh wants markets to watch the data rather than the Fed. Markets now have to watch both, because the data alone do not reveal which threshold will produce a majority. The wager can work. A less omnipresent central bank can make prices more informative and prevent every sentence from becoming a guarantee. It can also tighten through uncertainty, in a way that is more diffuse and less controllable than a 25-basis-point hike. Official yields prove that tightening occurred. They do not yet show that it was progress. Sources 1. Federal Reserve, Federal Reserve issues FOMC statement, 9-3 vote, 3.50-3.75% hold and three preferences for a 25-basis-point hike, July 29, 2026: 2. Federal Reserve, Transcript of Chairman Warsh's Press Conference Opening Statement, communication doctrine, higher yields and questions discussed by the committee, July 29, 2026: 3. Federal Reserve, Implementation Note issued July 29, 2026, reserve rate at 3.65%, repo operations and reinvestments: 4. Federal Reserve, Federal Reserve issues FOMC statement, 12-0 vote and the same target range, June 17, 2026: 5. Federal Reserve, Minutes of the Federal Open Market Committee, June 16-17, 2026, support for the hold, tightening scenarios and communication debate: 6. U.S. Department of the Treasury, Daily Treasury Par Yield Curve Rates, nominal yields on June 17 and July 29, 2026: 7. U.S. Department of the Treasury, Daily Treasury Par Real Yield Curve Rates, real yields on June 17 and July 29, 2026: 8. Bureau of Labor Statistics, Consumer Price Index, June 2026, 0.4% monthly decline, headline at 3.5% and core at 2.6% year over year, July 14, 2026: 9. Federal Reserve, Kevin Warsh, Semiannual Monetary Policy Report to the Congress, mandate, inflation and five reform projects, July 14, 2026: 10. Federal Reserve, Federal Reserve announces the leadership and objectives of its task forces to advance the conduct of monetary policy, communication, balance sheet, data, productivity and inflation, July 9, 2026: 11. Federal Reserve, Christopher J. Waller, Monetary Policy at a Crossroads, conditions for possible tightening and inflation analysis, July 13, 2026: 12. Federal Reserve Bank of Dallas, Lorie K. Logan, Remarks on inflation, employment and monetary policy, case for modestly higher rates, July 16, 2026: 13. Bank for International Settlements, Christopher S. Sutherland, Forward guidance and expectation formation: A narrative approach, Working Paper no. 1024, June 14, 2022: 14. Kurt G. Lunsford, Policy Language and Information Effects in the Early Days of Federal Reserve Forward Guidance, American Economic Review, vol. 110, no. 9, September 2020: 15. Federal Reserve, Vaishali Garga et al., Monetary Policy, Uncertainty, and Communications, Finance and Economics Discussion Series 2025-074, August 2025: 16. Federal Reserve, Gianni Amisano and Oreste Tristani, Uncertainty shocks, monetary policy and long-term interest rates, Finance and Economics Discussion Series 2019-024, April 2019: 17. Federal Reserve, Meeting calendars, statements, and minutes, 2026 calendar and September 15-16 meeting with projections: ============================================================================ ANALYSIS: The synthetic patient: the diagnosis that pays URL: https://l0g.fr/en/analysis/synthetic-patient-medicare-advantage/ Canonical French source: https://l0g.fr/posts/patient-synthetique-medicare-advantage/ Date: 2026-07-29 (reviewed 2026-07-29) Topics: Medicare Advantage, health insurance, United States, regulation, fraud, data ---------------------------------------------------------------------------- import InfographicBars from '../../components/infographic/InfographicBars.astro'; In Medicare Advantage, a patient exists twice. There is the person who visits a doctor, receives care and lives with an illness. Then there is the administrative double: a collection of diagnoses converted into a risk score. This second patient determines part of the monthly cheque the federal government sends to the insurer. The sicker that patient looks, the larger the cheque. The most disturbing financial engineering in American health insurance therefore sits not in a derivative, but in the medical record. The principle is legitimate. Medicare must pay more to a plan covering genuinely sicker people, otherwise every insurer would have an incentive to avoid costly patients. The problem begins when a diagnosis becomes not only clinical information but a unit of revenue. In March 2026, the independent commission advising Congress on Medicare, MedPAC, estimated that the programme would pay $615 billion to Medicare Advantage plans this year, $76 billion, or 14%, more than if the same beneficiaries had remained in traditional Medicare. MedPAC does not call the $76 billion fraud and states that the estimate measures neither plan profits nor administrative expenses. Its calculation nevertheless reveals a peculiar mechanism. Before selection and coding effects, spending would have been $3 billion below traditional Medicare. Favourable selection adds $57 billion; more intensive coding, after the regulatory adjustment, adds $22 billion. An initial saving disappears and becomes a large additional cost. A medical record becomes a payment formula Medicare Advantage, or Medicare Part C, allows a beneficiary to receive coverage from a private insurer instead of remaining in the traditional fee-for-service programme. CMS, the federal agency administering Medicare, then pays the plan a fixed monthly amount for every enrollee. That capitated payment is adjusted for age, certain demographic characteristics and diagnoses submitted by the plan. The calculation is explicit. CMS groups diagnoses into Hierarchical Condition Categories, or HCCs. Each category carries a coefficient representing the expected cost of the condition. The risk score adds those coefficients and demographic factors. The final payment equals the plan's base rate multiplied by the score, as the Department of Health and Human Services inspector general explains. The mechanism is not marginal. In 2025, 55% of eligible beneficiaries had chosen Medicare Advantage, about 34.9 million people. A small difference in scores, multiplied across millions of enrollees and twelve monthly payments, becomes a significant revenue line. MedPAC puts it plainly: documenting one additional HCC can materially increase the payment for an enrollee. The first arbitrage selects the right enrollee The first gain requires no change to the medical record. It comes from favourable selection. MedPAC defines this as beneficiaries whose scores overpredict future spending enrolling in Medicare Advantage more often than beneficiaries whose scores underpredict it. In other words, a plan may be paid for more statistical risk than the care its members actually consume. This selection is not necessarily organised. Provider networks, benefit design, individual preferences about how to receive care and health differences that the model fails to capture can be sufficient. The financial effect remains. MedPAC attributes $57 billion of the estimated $76 billion payment gap in 2026 to favourable selection. A federal case shows how the incentive can move beyond passive selection. In May 2025, the Department of Justice filed a complaint against Aetna, Elevance, Humana and three brokers. The government alleges that insurers paid hundreds of millions of dollars to intermediaries for enrolments and that Aetna and Humana pressured them to steer away disabled beneficiaries considered less profitable. These remain allegations to be tested in court: no liability has been determined. Their economic logic nevertheless matches MedPAC's measurement problem. Risk adjustment is supposed to neutralise the cost of illness, yet actors may still have an incentive to choose who enters. The second arbitrage manufactures the right score Once a beneficiary is enrolled, the second lever is to make the profile more complete and therefore often sicker on paper. Plans can arrange in-home health risk assessments, then retrospectively review medical records to find diagnoses that the physician did not submit. These tools can correct incomplete information and improve follow-up. They can also produce a payable code without producing care. The HHS inspector general isolated the risk in 2022 data. Diagnoses appearing only in risk assessments or linked chart reviews, with no other visit, test, procedure or supply carrying those diagnoses, generated $7.5 billion of payments in 2023 for 1.7 million enrollees. In-home assessments and the reviews linked to them represented 63% of that amount. Twenty companies generated 80% of the payments. The inspector general does not conclude that all $7.5 billion was improper. It presents a more troubling fork: either the diagnoses were inaccurate and payments were improper, or serious conditions had been identified without patients subsequently receiving necessary care. Either way, the code travelled better than the care. The same report says CMS identified $12.7 billion in net overpayments in fiscal 2023 from plan-submitted diagnoses unsupported by medical records. That measure should not be added to the $7.5 billion: the scopes overlap and the methods differ. It confirms only that the distance between a payable diagnosis and a demonstrable diagnosis is a budget category, not an anecdote. Three cases expose the chain Cases resolved in 2026 trace the incentive from insurer to coding contractor. They do not all have the same legal status. In January, Kaiser Permanente affiliates agreed to pay $556 million to settle False Claims Act allegations. The government said physicians were pressured to add diagnoses, sometimes more than a year after the consultation, that had nothing to do with the visit. Financial targets were allegedly assigned to physicians and facilities. The Department of Justice expressly states that the settlement is not a determination of liability. In March, Aetna agreed to pay $117.7 million to resolve separate allegations. The case describes a striking asymmetry: chart reviews were used to add codes producing extra payment, while previously submitted codes that the same reviews did not support were allegedly left in place. $106.2 million of the settlement addresses that mechanism; $11.5 million concerns morbid-obesity codes inconsistent with recorded body-mass index. Again, the settlement is not a liability judgment. In June, Matrix Medical Network, an in-home assessment contractor, entered a $36.5 million settlement and a five-year corporate integrity agreement. The distinction matters: according to the federal prosecutor in Manhattan, Matrix made factual admissions. It generally charged $350 to $450 per assessment and reported chronic conditions without sufficient clinical information. Some appeared in no record from any other provider in the two years before or after the home visit. These settlement values measure neither the system's return nor the full amount of improper payments. They document three places where the same data could be monetised: the medical-record addendum, the asymmetric chart review and the outsourced home visit. Intensive coding is not synonymous with fraud The analysis would be wrong if it treated every coding difference as a false diagnosis. MedPAC lists several causes. Plans may document real conditions more completely than traditional Medicare physicians, who do not always have an incentive to submit every possible code. Risk assessments can discover a neglected condition and trigger care. The model also provides essential protection for people whose treatment is genuinely costly. Medicare Advantage also offers benefits absent from traditional Medicare, reduces some cost sharing and includes an annual out-of-pocket limit. Enrollees generally report satisfaction with their coverage. The extra $76 billion helps finance these benefits, even though all Medicare beneficiaries, including those in traditional Medicare, subsidise them through taxes and premiums. MedPAC estimates that higher plan payments will raise aggregate Part B premiums by about $11 billion in 2026, or $175 per beneficiary over the year. The V28 model reform also shows that regulation can work. MedPAC estimated the payment gap at 20% in 2025 and lowers it to 14% for 2026, mainly because V28 finished phasing in and risk-score growth slowed. The mechanism is neither immutable nor entirely captured by insurers. But the correction remains incomplete. After the minimum 5.9% regulatory adjustment, Medicare Advantage scores are still projected to be 4% higher than scores for comparable traditional Medicare beneficiaries, producing the $22 billion coding component. CMS has the authority to impose a larger adjustment than the statutory minimum. According to MedPAC, it has never done so. One door closes in 2027 CMS has finally targeted one of the most contested tools. Starting in 2027, diagnoses found in a chart review that is not linked to a specific patient encounter will no longer count towards the score, except when a beneficiary switches from one Medicare Advantage organisation to another. Diagnoses originating only in an audio call will also be excluded. The agency estimates that removing these unlinked reviews will subtract 1.53% from payments relative to retaining them. Yet, driven in part by higher underlying costs, payments to plans are still projected to rise by 2.48%, or more than $13 billion. CMS also expects average risk scores to grow 2.5% because of population changes and coding practices. The reform therefore closes a door, not the building. It does not eliminate diagnoses associated with a real encounter, in-home risk assessments or the general incentive to maximise HCCs. Two HHS-OIG recommendations remain open: further restrict diagnoses produced only by home assessments and audit their validity. CMS had not concurred with either recommendation when the report was issued. The risk changes form In our investigation into private credit's regulatory data gap, risk came from missing data: authorities could not connect exposures. Here the incentive works in reverse. The data exists because it is payable. The system rewards production of an administrative representation of illness, sometimes faster than it verifies the real patient. The journey from consumer credit to an annuity showed how a claim changes holders. Medicare Advantage shows how information changes nature. A diagnosis leaves the consultation, becomes an HCC, then a score and finally a federal cash flow. At every stage, its clinical meaning moves a little further from its financial value. The test is now observable. If the 2027 exclusion of unlinked reviews durably reduces coding differences, if audits recover more improper payments and if conditions detected at home lead to verifiable care, the synthetic patient will become a weaker description. If scores continue to grow faster than the cost of comparable patients, the problem will not be a rogue contractor. It will lie in the price attached to the diagnosis itself. Limitations MedPAC's estimates rely on counterfactuals: they compare observed payments with what the same beneficiaries might have cost in traditional Medicare. They are sensitive to data, the risk model and selection assumptions. The $57 billion favourable-selection effect, $22 billion coding effect and $7.5 billion tied to risk assessments do not describe the same scope and should not be added. The Kaiser and Aetna settlements resolve allegations without a determination of liability. Matrix made factual admissions in its agreement, but its conduct does not establish that every contractor or plan behaves in the same way. Finally, a diagnosis missing from other care data may be false or may reveal failed follow-up; available data cannot always distinguish between the two. Sources 1. Medicare Payment Advisory Commission, The Medicare Advantage Program: Status Report, March 2026, especially pages 343 to 348, 351 to 352 and 386 to 387: payments, favourable selection, coding intensity, counterpoints and methodology. 2. HHS Office of Inspector General, Medicare Advantage: Questionable Use of Health Risk Assessments Continues To Drive Up Payments to Plans by Billions, October 2024: HCC model, $7.5bn, 1.7 million enrollees, concentration and limitations. 3. Centers for Medicare & Medicaid Services, 2027 Medicare Advantage and Part D Rate Announcement, 6 April 2026: exclusion of unlinked diagnoses, payment effects and expected score growth. 4. U.S. Department of Justice, Kaiser Permanente settlement, 14 January 2026. 5. U.S. Department of Justice, Aetna settlement, 10 March 2026. 6. U.S. Department of Justice, Matrix Medical Network settlement, 3 June 2026. 7. U.S. Department of Justice, complaint against three insurers and three Medicare Advantage brokers, 1 May 2025: alleged commissions, enrollee steering and absence of a liability determination. ============================================================================ ANALYSIS: Private credit: risk begins where the data ends URL: https://l0g.fr/en/analysis/private-credit-risk-begins-where-data-ends/ Canonical French source: https://l0g.fr/posts/credit-prive-risque-trou-donnees/ Date: 2026-07-29 (reviewed 2026-07-29) Topics: private credit, banks, insurance, regulation, systemic risk, data ---------------------------------------------------------------------------- import InfographicBars from '../../components/infographic/InfographicBars.astro'; The most important signal regulators have sent about private credit is not the announcement of a giant default. It is an admission about measurement. On 6 May 2026, the Financial Stability Board, or FSB, published fifty pages on a market estimated at $1.5 trillion to $2 trillion at the end of 2024. Its most unsettling conclusion lies in one gap: data supplied by its members capture about $220 billion of drawn and undrawn bank credit lines to private credit funds, while commercial databases suggest an amount more than twice as large. The FSB is not saying that $440 billion of losses are hiding on balance sheets. It is saying something more basic: for one direct exposure that authorities are trying to monitor, the range runs from one to more than two. After a decade of rapid growth outside traditional banking statistics, the regulatory turning point is therefore not yet an accident. It is the discovery that the system's map does not match its territory. The number that breaks the narrative The FSB report first makes the strongest case for the sector. Private credit finances companies that banks serve less well, offers tailored solutions and disperses risk beyond bank balance sheets. Closed-end funds, with no permanent redemption promise, are also less exposed to an immediate run than deposit-funded banks. But the strength of that defence depends on what is actually measured. The FSB's $220 billion is neither the global stock of private credit nor the loans made directly to its corporate borrowers. It covers bank credit lines, whether used or still available, extended to private credit funds in the data that members could aggregate. Commercial estimates exceed $440 billion, the lower bound obtained by doubling $220 billion because the report says "more than twice". Both estimates remain small relative to bank assets and CET1 capital, the FSB stresses. That is reassuring, but incomplete. It excludes banks also lending to the same companies as the funds, portfolio financing, bank-manager partnerships, NAV loans and synthetic risk transfers. An exposure that looks modest line by line may become less modest when several business units finance the same risk under different names. The problem is also concentrated. In research cited by the FSB on bank lending to US BDCs, the top five banks carry 63% of committed amounts and the top ten about 84%. These ratios come from a sample of large bank holding companies subject to Fed stress tests. They do not describe the entire market, but show why a system-wide average can obscure the nodes that matter. The American regulatory paradox The Federal Reserve has documented the hole itself. In a technical note published on 26 February 2026, it explains that private credit loans are not separately identified in the US Z.1 Financial Accounts. Domestic loans sit among businesses' "unidentified miscellaneous liabilities". The corresponding claims held by private debt funds are not assigned to a lender sector and appear as accounting discrepancies. Supervision then began compensating for statistics. On 10 April, Bloomberg reported via Fortune that Fed examiners were asking major banks for details of their lending to private credit funds. The report relies on people familiar with the matter and the Fed did not comment. It should therefore be treated as well-sourced reporting, not as a public collection whose questionnaire can be inspected. The Treasury made its move official. On 1 April, it announced a series of meetings with US and international insurance regulators on recent events, emerging risks and risk-management practices in private credit. The same Bloomberg report says an internal team was assembled for the work. The meeting with US state insurance supervisors is a public fact; the team's composition remains a press report. This sequence captures the paradox. Authorities want to determine whether private credit can transmit a shock through the system. To do so, they use ad hoc requests, commercial providers and proxies precisely because regular accounts do not yet isolate the object they are monitoring. Insurers and the denominator problem Life insurance makes the data gap visible. The FSB explains that assets can be classified as corporate bonds, private placements or structured products depending on the filer and jurisdiction. Its proxy, based on private placements and private ratings, suggests that about 10% of North American life insurer portfolios may be private credit, against roughly 3% for non-life insurers. In the International Association of Insurance Supervisors' collection, most jurisdictions remain below 5% of total insurance-sector assets, but definitions differ. A Chicago Fed research paper, revised on 27 April 2026, finds $849 billion, or 14% of US life insurer balance sheets in 2024, under a broader definition including private credit to financial borrowers and privately placed ABS. Barclays, according to the trade press, estimates that holdings grew by more than 20% in 2025 to around 10% of total assets, and exceeded 15% at some private-equity-affiliated insurers, including Athene and Global Atlantic. Moody's, finally, measures $807 billion of private credit and illiquid assets at year-end 2025, or 20% of the industry's fixed-income portfolio. These numbers do not form a time series. They measure different universes, dates and denominators. The dispersion is not an editorial footnote. It is the risk. If a supervisor cannot link a loan, its vehicle, its private rating, its ultimate insurer and any offshore reinsurance, it cannot aggregate leverage or detect that the same potential loss crosses several affiliated entities. The Apollo, Athene and reinsurance triangle showed how a manager originates credit, how its insurer provides long-term funding and how liabilities move between jurisdictions. The systemic question begins when that triangle is no longer one case but an architecture reproduced across groups. Our investigation into life insurers, retirement savings and Bermuda details that channel. From the triangle and the circle to the network The circle of synthetic risk transfers asked another question: what is a transfer worth if the bank also finances the protection buyer? The FSB report now connects these architectures. It cites credit lines to funds, revolvers to companies also borrowing from those funds, insurers holding the assets, private equity controlling some insurers and SRTs held by private credit funds. The network is also beginning to be measured. In May 2026, Chicago Fed researchers reconstructed life insurers' bond portfolios from regulatory filings. Between 2016 and 2024, private placements rose from 14% to 22% of their corporate bond holdings. Athene, Apollo's subsidiary, sits near the centre of the private-placement network with more than 40 systematic connections, compared with only five for MassMutual, despite the latter being the largest holder by volume. The same study supplies an important counterpoint. Overall portfolio similarity declined and diversification into private credit reduced some average interconnectedness. Systematic overlaps nevertheless increased in private placements and cluster among a subset of insurers sharing strategies or asset managers. The risk therefore does not look like a crowd holding exactly the same listed asset. It looks like a few clusters linked by the same origination channels. The Dimon-Bessent line still has force Jamie Dimon does not deny the sector's weaknesses. In his 2025 JPMorgan shareholder letter, he cites weaker standards, PIK, aggressive private ratings, poor transparency and insufficiently rigorous marks. Yet he concludes that, "in the great scheme of things", a $1.8 trillion market probably does not present systemic risk. On 15 April, Scott Bessent defended a similar line on CNBC: none of the Treasury's work had revealed a systemic problem. That reading is not merely industry messaging. In its May 2026 analysis, the European Central Bank finds that direct euro-area exposures are small in aggregate. In its severe scenario, bank losses tied to private credit do not exceed 1.3% of equity, while direct losses at insurers and pension funds remain absorbable. Closed-end funds have little redemption risk; insurers have long-dated liabilities; bank loans to funds are often senior. The FSB is equally careful: its members have reported no system-wide stress. Opacity is not evidence of hidden losses. A data gap does not mechanically turn a market into a crisis. The weakness of the "too small to matter" thesis lies elsewhere. The numerator is uncertain, scope changes by source, and indirect exposures are absent from the figure meant to demonstrate smallness. The sound conclusion is not that Dimon or Bessent are wrong. It is that the confidence placed in their conclusion depends on data that authorities themselves describe as incomplete. The real turning point The FSB already lists the elements of the future map: a harmonised definition, assets by strategy, bank lending by facility type, committed and drawn capital, insurer allocations, private-equity ownership of insurers, fund and borrower leverage, redemptions and SRT holdings. It also calls for identifiers that can connect one borrower to several lenders and for transparency beyond the first layer of investment vehicles. That is the turning point. A spectacular default would provide a name, a date and a loss amount. It would arrive too late to reveal the structure. The data hole acts earlier: it prevents authorities from knowing whether a bank line, a private bond held by an insurer, a reinsurance arrangement and an SRT tranche are four independent risks or four entries around the same borrower. The simultaneous awakening of the FSB, Fed, Treasury, ECB and insurance supervisors does not prove that a crisis is imminent. It proves that the market has outgrown the statistical machinery built to follow it. Private credit's first systemic risk may not be its volume. It may be the width of the uncertainty interval around that volume, its holders and its connections. Sources 1. Financial Stability Board, Report on Vulnerabilities in Private Credit, 6 May 2026: market size at end-2024, roughly $220bn of bank lines, commercial estimates above twice that amount, interconnections, data limitations and proposed metrics. 2. Federal Reserve, Financial Accounts Z.1, Technical Q&A, 26 February 2026: private credit loans not separately identified, unidentified miscellaneous liabilities and discrepancies between borrower and lender sectors. 3. U.S. Treasury, meetings with insurance regulators, 1 April 2026. 4. Bloomberg via Fortune, Fed seeks details on U.S. banks' exposure to private credit firms, 10 April 2026: Fed supervisory queries and the Treasury team, attributed to people familiar with the matter. 5. Federal Reserve Bank of Chicago, Life Insurers' Private Credit Investments and Annuity Market Share Capture, revised 27 April 2026: $849bn, 14% of life insurer balance sheets in 2024 and the role of PE-owned insurers. 6. Federal Reserve Bank of Chicago, Assessing Life Insurers' Interconnectedness Through Corporate Credit Investments, May 2026: private placements, portfolio overlaps, Athene and MassMutual. 7. European Central Bank, Stress in global private credit markets and its implications for euro area financial stability, May 2026: exposures, loss scenario and definitional limits. 8. JPMorganChase, Jamie Dimon's 2025 shareholder letter: relative size, transparency, valuations, credit standards and systemic-risk assessment. 9. American Investment Council, transcript of Scott Bessent's CNBC forum remarks, 15 April 2026. An interested trade-group source, used only for the remarks attributed to Bessent. 10. Insurance Business, Barclays estimate of life insurer exposures, 26 May 2026. A secondary estimate because no public Barclays note was located. 11. Moody's, Private Credit: $807bn of private credit and illiquid assets at year-end 2025, equal to 20% of US life insurers' fixed-income portfolio. ============================================================================ ANALYSIS: Held to maturity URL: https://l0g.fr/en/analysis/held-to-maturity/ Canonical French source: https://l0g.fr/posts/detenu-jusqu-a-l-echeance-pertes-latentes/ Date: 2026-07-29 (reviewed 2026-07-29) Topics: international, banks, accounting, risk, united-states ---------------------------------------------------------------------------- A bond does not lie about its value: the market sets it every day. A bank, though, has a choice the market does not, the choice of whether to look. US accounting offers it, for that, a reassuringly named category, held-to-maturity, in which a bond stays recorded at its original cost, indifferent to its real value, as long as the bank promises not to sell it. At the end of March 2026, US banks lodged $214 billion of unrealized losses there, invisible on the balance sheet by the sole grace of a promise. The promise holds as long as nothing forces the sale. In March 2023, Silicon Valley Bank discovered, in a few hours, the fragility of that conditional. Three drawers for one bond The same bond changes nature depending on the accounting drawer the bank files it in, and the choice of drawer governs the value the world will grant it. The first drawer, trading, marks the bond to market and runs the swings through profit and loss: the loss shows at once. The second, available-for-sale, also marks the bond to market, but lodges the loss in equity, without touching earnings: the loss shows, discreetly, on the balance sheet. The third, held-to-maturity, marks nothing at all. The bond stays at amortized cost, as if its value had not moved, and the loss appears nowhere, neither in earnings nor in equity. It exists economically, it is absent in the accounts. The scale of the silence can be measured. The FDIC's quarterly profile puts banks' total unrealized losses at $325.1 billion in the first quarter of 2026, of which $214.5 billion in held-to-maturity and $110.6 billion in available-for-sale. Two thirds of the loss, in other words, sleep in the one drawer where the rule permits not counting it. The rate rise of 2022 gouged the price of the long bonds bought when money was free; held-to-maturity made it possible not to keep the ledger of it. The move of 2022 The rule offers more than shelter, it offers an exit. A bank whose available-for-sale securities were bleeding into equity as rates rose could reclassify them into held-to-maturity, freezing the loss at the transfer date and, from then on, ceasing to remark it. In January 2024, Fifth Third thus transferred $12.6 billion of securities from available-for-sale to held-to-maturity, carrying $994 million of unrealized losses out of the reach of remeasurement. The operation erases not a cent of loss; it moves it to the blind spot. Academics watched that move closely, and their conclusion is uncomfortable. A study by the Becker Friedman Institute in Chicago on bank fragility and reclassification into held-to-maturity shows that the banks reclassifying the most were also, on average, the most fragile: the accounting move was not neutral, it signalled a strain it served precisely to mask. The gesture that renders the loss invisible is also the one that betrays its weight. The filter that neutralizes even the visible losses The visible part of the loss, that of available-for-sale securities, nonetheless does not dent the banks' stated strength, and the reason is a second device, less known than the first. Regulatory capital, the famous ratio that measures solvency, excludes these losses thanks to an option, the AOCI filter, which about 99% of US banks have used. In plain terms, a bank can display a comfortable capital ratio while carrying, in its equity, losses that genuinely impoverish it. Held-to-maturity removes the loss from the balance sheet; the AOCI filter removes from regulatory capital the part of the loss the balance sheet does acknowledge. Between the two sieves, the number that governs confidence, the solvency ratio, ignores most of the gap between the bonds' book value and their real value. March 2023, the promise that breaks The defensive reasoning is solid, and it must be laid out before it is tested. A government bond held to maturity repays its principal in full; its market loss is only a bump along the road, bound to fade as the bond nears its term. Held to the end, it costs nothing. In that light, held-to-maturity does not mask a loss, it avoids recording a loss that will never be one. The argument is right, on one condition, a single one: that the bank can actually hold to maturity, hence that it is never forced to sell. Silicon Valley Bank lived the fall of that condition in real time. Sitting on a base of volatile deposits, it had piled up long bonds classified held-to-maturity, whose unrealized losses did not appear in its ratios. When its depositors moved to withdraw their funds, it had to sell those securities to find cash, and the sale turned, in a few hours, an accounting non-loss into a real and fatal one. The "held-to-maturity" category is a bet on calm; a run is exactly the stress that voids the bet, at the precise moment the hidden loss turns lethal. The accounting had given no warning, not by error, but by design. To postpone the fatal instant, SVB had, moreover, funded itself at the window of the Federal Home Loan Banks, that lender of next-to-last resort which lets a bank push back the sale, and thus the recognition, a little longer. The remedy in reverse The lesson of 2023 had been drawn, on paper. The major prudential reform known as the Basel III endgame planned to remove the AOCI filter for banks above one hundred billion dollars in assets, forcing them to make their available-for-sale losses finally weigh on their capital, with a phase-in beginning 1 July 2025. The correction aimed straight at SVB's flaw. It had two limits, though: it touched only available-for-sale, never held-to-maturity where the largest share of the loss sleeps, and it covered only the biggest banks. Those limits are now the least of it, because the reform itself is ebbing. Signals from the Federal Reserve point to a capital-neutral Basel III endgame in 2026, a polite formula for saying the tightening will be gutted. The regulatory unwinding we documented elsewhere, that rollback on bank capital, carries away with it the one measure that answered directly the cause of the most spectacular failure since 2008. Three years after SVB, the flaw is wider than it was then in held-to-maturity terms, and the warning light someone had set out to switch back on is being unscrewed. One must guard against catastrophism, and the qualification is the same as the defence: most of these $214 billion will fade if the banks are never forced to sell, and many will not be. The danger is not a $214 billion hole opening tomorrow. It is narrower, and more insidious: the accounting certifies that a bank is fine right up to the instant it is not, and the condition that turns the invisible loss into a fatal one, a deposit flight, is precisely the one no balance sheet displays. Held to maturity is not a lie, it is a promise, and a promise is worth only the stability that lets it stand. The market, for its part, does not file its prices in drawers. It knows the value of a bank's bonds before the bank does, and it does not wait for the run to remind it. --- Sources - FDIC, "Quarterly Banking Profile, First Quarter 2026" (total unrealized losses $325.1bn, of which $214.5bn held-to-maturity and $110.6bn available-for-sale) - Becker Friedman Institute, University of Chicago, "Bank Fragility and Reclassification of Securities into HTM" (the most fragile banks reclassified the most; 99% of banks use the AOCI filter) - Securities and Exchange Commission, Fifth Third Bancorp, Form 10-K 2023 (transfer of $12.6bn from AFS to HTM in January 2024, $994m of losses at transfer) - Congressional Research Service, "Banks' Unrealized Losses, Part 1: New Treatment in the Basel III Endgame Proposal", IN12231 (removal of the AOCI filter for banks above $100bn, phase-in from 1 July 2025) - Bloomberg, "Fed remarks point to capital-neutral Basel III Endgame in 2026" - Office of Financial Research, "The State of Banks' Unrealized Securities Losses", May 2025 (scale and dynamics of unrealized losses) ============================================================================ ANALYSIS: Apollo, the triangular domino URL: https://l0g.fr/en/analysis/apollo-the-triangular-domino/ Canonical French source: https://l0g.fr/posts/apollo-domino-triangulaire/ Date: 2026-07-28 (reviewed 2026-07-28) Topics: international, private credit, insurance, risk, governance ---------------------------------------------------------------------------- There are two ways to tell the story of Apollo Global Management. The first is the one in the quarterly filings: a house founded in 1990, become in the first quarter of 2026 the first alternative asset manager to cross a trillion dollars under management, a private credit champion, a retirement engine for hundreds of thousands of Americans. The second follows the money and the risk through the side doors, and it tells something else: a closed triangle where the same small group of players holds all three corners, an insurer lodged in Bermuda to slim down regulatory capital, and at the top a founder whose payments to Jeffrey Epstein brought him down, without the wave stopping at him. It is this second story we tell here, piece by piece, source by source. Not because the first is false, but because it is incomplete. And because a domino, when arranged in a triangle, never falls alone. The birth certificate: Drexel, a dead insurer and a jackpot To understand what Apollo has become, you have to go back to what the house was born from. From 1977 to 1990, Leon Black ran the mergers and acquisitions department at Drexel Burnham Lambert, where he was regarded as the right hand of Michael Milken, the "junk bond king". When Drexel collapsed in 1990 under the weight of prosecutions, Black founded Apollo in its wake, with Josh Harris and Marc Rowan, on a simple idea: buy the discounted debt of troubled companies on the cheap, a strategy known as distressed-to-control. The first big score is telling, because it already put an insurance company at the centre of the game. In 1991, the Californian insurer Executive Life collapsed: its high-yield bond portfolio, bought precisely from Milken and Drexel, was falling sharply in value. California's insurance commissioner auctioned the portfolio, and in November 1991 Black won the auction with a bid of $3.5 billion, of which $3.2 billion was for the bond portfolio alone. Black did not have the money; to finance the deal he turned to the French bank Crédit Lyonnais, then state-controlled. The affair would become a resounding scandal, US law barring a foreign bank from owning an insurer, and Leon Black would himself be named in lawsuits alleging a conspiracy to seize the insurer's assets illegally. The portfolio bought at $3.2 billion would be worth billions more once the junk market recovered. Apollo's fortune was launched, and its DNA was written: buy the wreckage, and if possible the wreckage of an insurer. Thirty years later, Apollo would no longer buy a dead insurer's portfolio. It would own the insurer. The triangle Here is the machine as it works today, and you have to see it whole to understand why it is at once so profitable and so hard to take apart. Apollo is no longer just a fund: it is a three-sided structure whose sides feed one another. At the first corner, the asset manager, Apollo, which originates private credit, that is, direct loans to companies, without listing or an active secondary market. At the second corner, an insurer, Athene, founded in 2009 in Bermuda, an annuity specialist, which Apollo absorbed through an all-stock merger valuing the insurer at about $11 billion, closed in January 2022. Athene sells retirement annuities and takes over corporate pension plans, which gives it a giant reservoir of very long-duration capital. At the third corner, that capital is deployed, by Apollo, into the private credit Apollo originates. Chief executive Marc Rowan has summed up the bet bluntly: an asset-heavy balance sheet, fuelled by Athene's hundreds of billions of long-duration liabilities, should deliver superior, repeatable returns and carry assets toward $1.5 trillion. In the first quarter of 2026, the house crossed a trillion dollars under management for the first time, at $1.026 trillion exactly, with record inflows of $115 billion in the quarter alone. Bermuda, or the art of slimming down risk The third corner deserves a pause, because that is where part of the profitability, and part of the risk, is decided. The mechanism is called asset-intensive reinsurance: an insurer cedes its annuity or life reserves to a reinsurer, often affiliated with the same group and located offshore, in Bermuda in particular, where the prudential regime is lighter than in the United States. The reinsurer takes over the liabilities, reinvests the assets, frequently in private credit and structured products, and the whole ends up holding less regulatory capital for the same liability. It is exactly the logic we described in our investigation into life insurers, retirement savings and Bermuda. The scale is no longer confidential. According to the Financial Stability Oversight Council's (FSOC) 2024 annual report, published in March 2025, more than 40% of annuity reserves ceded by US life insurers now go offshore, and close to 40% of those reserves reach Bermuda, a share that climbs to about 60% if you look only at 2023 transactions. A forensic accountant cited by American Banker estimates that life insurers have shifted about $2 trillion of liabilities to offshore or captive reinsurers, of which some $1.3 trillion abroad, while the sector has placed nearly a third of its $5.6 trillion of assets in private credit. FSOC made these structures an explicit point of concern in its 2025 report, and state regulators responded by adopting actuarial guideline 55, which requires testing the adequacy of the assets behind reinsurance ceded offshore. Athene is not an abstract textbook case: it is the vehicle through which the savings of real people tip into this circuit. According to Bloomberg, the insurer has struck at least 49 deals with companies such as Alcoa, AT&T and Lockheed Martin to convert $53 billion of pensions into annuities, covering about 535,000 people as of mid-2025. For those employees and retirees, the manager watching over their pension is no longer a traditional insurance company but a subsidiary of a private equity firm whose trade is yield. Fairness is due here, because this is the heart of the controversy and it has two readings. The first, argued by Apollo, is that this model makes the system sturdier: an insurer owned by a sophisticated manager invests better, spreads risk toward long-duration holders, and Athene shows precisely a lower level of related-party investment than some of its peers. The second, held by part of the regulatory and research community, is that these private-equity-backed insurers hold fewer liquid assets than average, which makes them more vulnerable to a wave of defaults or downgrades in a slowdown, according to a 2023 IMF study. Both readings can be true at once: a model that performs better in calm and is more fragile at the worst moment. That is the nature of tail risk, the one we tracked from bank to fund in our investigation into synthetic risk transfer. The method: Caesars, or the art of sorting assets Before reaching the founder, a word is due on how Apollo treats the other end of the chain, its creditors, because it forged the firm's reputation for toughness. The emblematic case is the casino group Caesars Entertainment, bought by Apollo and TPG at the top of the market in 2008 in one of the largest leveraged buyouts in history, then smothered by its debt. What followed is a textbook of aggressive restructuring. In January 2015, junior creditors, led by attorney Bruce Bennett of the law firm Jones Day, filed an involuntary bankruptcy petition against the group's operating unit and demanded an examiner to investigate more than fifty transactions carried out by Apollo, TPG and management since 2008. The charge is blunt: creditors accuse the owners of having stripped the operating unit of its best assets, valuable casinos and properties transferred to a healthier sister structure, before letting the indebted shell sink into bankruptcy. The independent examiner gave weight to those grievances: he estimated that the potential damages tied to those fraudulent transfers could reach between $3.6 billion and $5.1 billion. Apollo and Caesars disputed those conclusions, and the matter was ultimately resolved within the reorganisation plan, the parent agreeing to contribute substantial value to creditors rather than face a judgment. No liability was therefore adjudicated by a court, but the episode durably installed the image of a player willing to play the hardest line at its creditors' expense. It is the same temperament, applied this time not to lenders but to a man's private life, that we find in the next file. The first domino: the man Now to the top of the triangle, to the man who built the machine, and to what brought him down. For Apollo's story is not only that of a financial construction: it is also that of a governance scandal rare in its scale, and it begins with a simple question. Why did the co-founder of one of the world's largest asset managers pay a fortune to an already convicted sex offender? The central facts are not in dispute, because they come from an investigation commissioned by Apollo's own board. In late 2020, prompted by press revelations, the board tasked the law firm Dechert with an independent review of the ties between Black and Epstein. The report, made public on 25 January 2021, established that Black paid Epstein $158 million between 2012 and 2017 for tax and estate planning advice, that he further lent him more than $30 million and gave $10 million to his foundation. According to the report, Epstein had helped Black solve an estate-structuring problem that could have created a tax liability of a billion dollars or more, and Epstein estimated he had saved Black $600 million. The Dechert report nonetheless concluded that Black had committed no wrongdoing and took no part in Epstein's crimes. Black, for his part, said he "deeply regretted" any involvement with Epstein. That report, commissioned to close the affair, did not close it: it opened it. As soon as it was published, Black announced he would step down as chief executive, then brought his departure forward and gave up on 21 March 2021 his roles as CEO, director and chairman of the board; Marc Rowan took the helm. Then public authorities seized the file. In early 2023, Black agreed to pay $62.5 million to the government of the US Virgin Islands to settle, without admission of guilt, potential claims linked to Epstein, and obtained in exchange criminal immunity for Epstein-related acts in the territory. His spokesperson maintains that Black paid Epstein for "legitimate financial advisory services, which he very much regrets," and that there is "no suggestion" that he knew of or took part in any misconduct. Then came the Senate investigation, and it is far more corrosive. Senator Ron Wyden, then head of the Finance Committee, ran a four-year "follow-the-money" inquiry. On 23 March 2026, in a letter to Black, he set out findings that must be presented for what they are, the assertions of a senator in a congressional investigation, not a verdict: " You were among Jeffrey Epstein's primary sources of income, flooding him with cash at a time when he was already a registered sex offender ". According to the same letter, the rates Black paid Epstein were thirty times higher than those of the elite tax advisers he already employed; $10 million was "papered over" through a sham 501(c)(3) charity, an Epstein lawyer writing in an email that routing the money through that structure would "avoid public disclosure" and "maximize deductions"; Black was overpaid $141 million by a family trust, whose reclassification could pull billions back into his taxable estate; emails indicate he paid millions to women using Epstein as a middleman, sums described as "gifts"; Epstein allegedly provided the Russian government with the location of women on Black's payroll; and Epstein, together with the head of the law firm Paul Weiss, allegedly surveilled women on Black's behalf. The Senate inquiry puts the total of the payments to Epstein at $170 million over several years, and Wyden referred his findings to the House Oversight Committee in June 2026. The matter then moved to Congress. On 26 June 2026, after Black refused to answer certain questions about non-disclosure agreements during a closed-door hearing he walked out of, House Oversight chairman James Comer subpoenaed him to appear again, under oath and on camera, on 16 July 2026, and to produce the non-disclosure agreements in question. Black has never been criminally charged, and he denies any wrongdoing. The court records The scandal's wave spilled from the tax terrain into the civil courts, where Black was both defendant and plaintiff. These files deserve close reading, and without indulgence in either direction: several ended in Black's favour, one remains open, and the exact statuses matter as much as the allegations. The first suit is that of Guzel Ganieva, a former model who accused Black of harassment and sexual assault. After a March 2021 interview in which Black acknowledged a consensual affair and claimed she had subjected him to extortion, the trial judge dismissed Ganieva's claims in May 2023, and a New York state appeals court ruled for Black on 16 January 2025, by four votes to one, holding that a 2015 non-disclosure agreement covered all her grievances and that she had "ratified" it by accepting $9 million, including a $100,000 monthly stipend. The second file is that of Cheri Pierson, who, under a New York law reopening the statute of limitations, had accused Black of raping her in Epstein's Manhattan townhouse. She ended her case: according to a New York State Supreme Court filing, the suit was "discontinued with prejudice and without costs to any party" in February 2024, meaning she cannot revive it, and that Black paid nothing to extinguish it. The third file is the most serious and the only one still open. In July 2023, a woman identified as "Jane Doe," autistic and born with mosaic Down syndrome, sued in Manhattan federal court, alleging that Black raped her in 2002, when she was a minor, at Epstein's townhouse; the complaint, detailed in the press, was reported by NBC News. Black denied it outright, his lawyers calling the action "frivolous and sanctionable." The file went through an extraordinary procedural battle: federal judge Jessica G. L. Clarke denied Black's motion to dismiss in September 2024, letting the case proceed, but the firm representing the plaintiff, Wigdor, asked to withdraw, and one of its lawyers was sanctioned by the judge for having "lied repeatedly". At this stage the matter remains an unadjudicated allegation, vigorously contested by Black, in a procedural framework damaged on the plaintiff's side. Fairness requires saying so as plainly as one states the allegation. Finally, Black did not only face lawsuits: he launched them. He sued his co-founder Josh Harris, Guzel Ganieva, the firm Wigdor and a public relations consultant, alleging an "unholy alliance" to destroy him under the anti-racketeering RICO statute. Federal judge Paul Engelmayer dismissed those claims "with prejudice" in 2022, finding them "glaringly deficient in fundamental respects," and the Second Circuit affirmed that dismissal on 2 March 2023. Black's legal offensive, too, therefore shattered. The scandal that climbs A founder's scandal could, in principle, stay confined to the man who caused it. Apollo did everything to keep it there: the Dechert report had taken care to conclude that neither Marc Rowan nor Josh Harris had hired Epstein or consulted him on their personal matters, and that no Apollo employee other than Black had ever seriously considered employing him. But the wave did not stop there, and that is the nature of a triangular domino: it climbed two of the three sides. First toward the current leadership. In February 2026, the exploitation of the Epstein documents released by the Justice Department revealed, according to CNN, that Marc Rowan, today Apollo's chief executive, had had several meetings and email exchanges with Epstein years after his 2008 conviction: in February 2016 they reportedly discussed a tax inversion strategy, and an executive at an Apollo affiliate reportedly asked, in September 2016, that Epstein continue to be copied on tax matters for his "substantive expertise". Apollo rejected the criticism: its president James Zelter said that "from an Apollo perspective, there's nothing new in these documents," and that Epstein's attempts to secure work beyond Black had been "declined at every turn." That denial must be restored with the same care as the allegation: nothing, at this stage, establishes any wrongdoing by Rowan, and the board-commissioned report expressly cleared him. But the plain fact, that the current leader kept professional contact with Epstein after 2008, now belongs to the public record. Then toward the next generation, and toward political power. On 31 January 2025, President Trump appointed Ben Black, Leon Black's son and a former Apollo associate, to lead the US International Development Finance Corporation (DFC), the federal development-finance agency; the Senate confirmed him, and he took office on 7 October 2025. The appointment fed questions about possible conflicts of interest, one investigation noting that Apollo had shown interest in debt tied to X, Elon Musk's network, just as the Black son was reaching a strategic agency of the administration. Nothing there is illegal, and one may legitimately judge a man on his merits and not on his father's name. But for a house whose brand has been dented by the Epstein file, seeing the Black name reinstalled atop a federal financial instrument carries an irony that has not escaped commentators. What Apollo would answer, and what remains true anyway An investigation without indulgence must also lay out the strongest defence, without which it charges only one side. Apollo and its defenders have arguments that are not mere formality. On the man, first: Leon Black has never been criminally charged, the board-commissioned report concluded he took no part in Epstein's crimes, two of the three civil proceedings ended in his favour or without any payment from him, and both his offence and his defence are his right. Senator Wyden's findings, however damning, are congressional allegations, not adjudicated facts. On the machine, next: the Apollo-Athene model is legal, supervised, and its supporters contend that a well-run insurer backed by a sophisticated investor serves its policyholders better than a traditional company; Athene highlights a lower level of related-party investment than several of its peers, and the firm raises record capital precisely because knowledgeable institutions trust it. On offshore reinsurance, finally: it disperses a risk otherwise concentrated, and regulators have framed it, not banned it, which suggests they judge it manageable. Marc Rowan himself, challenged on the liquidity of private credit, waved the worry away with a provocative line, calling an "idiot" any lender unable to meet 5% of redemptions on a fund. The confidence on display is real, and so far the facts have proved it right. Yet the soundness of a model in calm says nothing of its resistance to shock, and that is where the triangle worries. Why the triangle is fragile Let us gather the pieces. Apollo's risk is not that one of its three corners is rotten; it is that they are correlated, and held by the same hands. A manager originates the credit, an affiliated insurer holds it with retirees' savings, an affiliated reinsurer in Bermuda thins the capital the whole carries against it. In a normal regime, each corner reinforces the others. In a shock, the same property runs in reverse: there is no independent third party to cushion. If private credit deteriorates, it is Athene's assets that lose value; if Athene must rebuild capital, it is the offshore reinsurance mechanism that is tested; and if the house's reputation cracks, it is inflows, the model's raw material, that dry up. The three nets tear together, because they are woven from the same thread. It is the insurance version of the circle we described for banks in the risk that goes in circles and for the saver in from the credit card to the annuity. To that financial fragility is added one the models capture poorly: governance and reputational risk. For a house managing the retirement of hundreds of thousands of people, the integrity of those who run it is not a nice-to-have, it is an asset on the balance sheet. A founder subpoenaed by Congress, a chief executive whose emails with Epstein reach the public record, a family name reinstalled at the heart of the federal state: these are not celebrity anecdotes, they are risk factors that weigh on trust, therefore on inflows, therefore on the very substance of the triangle. The investor contemplating Apollo's record performance should ask the question that all our work on private credit, Bermuda insurers and the silent contagion invites: not "how much does this earn today," but "who carries the risk, where, and what happens when the three corners move at once." To judge the real soundness of such a group, you now have to read the insurer as much as the manager, an exercise our guides on a life insurer's health and private credit risk help carry out. A single domino makes no noise when it falls. Arranged in a triangle, with the same hands on each corner, it makes a great deal. Apollo has built the most elegant of machines for turning long savings into yield; it runs remarkably well as long as nothing pushes it. The only question worth asking, for the analyst as for the retiree, is what happens the day something pushes it. The founder, for his part, has already shown that one corner of the triangle can give way on its own. --- Sources - Apollo Global Management, "Apollo Reports First Quarter 2026 Results" ($1.026tn AUM, record inflows of $115bn) - Apollo Global Management, "Apollo Completes Merger with Athene…", 3 January 2022 (Apollo-Athene merger, permanent capital) - FinancialContent / Finterra, "Apollo Global Management (APO): The Trillion-Dollar Credit Engine" ($1.5tn target, asset-heavy model, Athene) - Los Angeles Business Journal, "Lawsuit: Executive Life's New Twist" (November 1991 auction, $3.5bn, Crédit Lyonnais, Leon Black named in the lawsuits) - Wikipedia, "Leon Black" (Drexel Burnham, Michael Milken's right hand, Apollo founded in 1990) - Fortune, "Caesars: A private equity gamble in Vegas gone wrong" (2008 buyout, January 2015 involuntary bankruptcy petition, more than 50 transactions probed) - Forbes, "Caesars Bankruptcy Examiner: Fraudulent Conveyance Damages Could Reach $5.1B" (potential damages estimated at $3.6bn to $5.1bn) - Insurance Business, "FSOC raises alarm on insurers' use of offshore reinsurance" (over 40% of ceded reserves offshore, ~40% to Bermuda, ~60% of 2023 transactions) - U.S. Department of the Treasury, FSOC 2025 Annual Report release - American Banker, "Is private credit a $2 trillion-dollar insurance timebomb?" (~$2tn of offshore/captive liabilities, ~1/3 of $5.6tn assets in private credit, AG 55) - Bloomberg, "Apollo's Athene Led Private Equity's Move Into Pensions, Shifting Risk Offshore", 17 November 2025 (49 deals, $53bn of pensions converted, ~535,000 people) - Retirement Income Journal, "Private Credit Anxiety and the Bermuda Triangle" (2023 IMF study: PE-backed insurers, fewer liquid assets) - CNBC, "Apollo's Leon Black to retire as CEO…", 25 January 2021 (Dechert report, $158m 2012-2017, loans, gift, departure of 21 March 2021) - CNN Business, "Leon Black made a $158 million payment to Jeffrey Epstein" (estimated $600m tax saving, $1bn liability avoided, Senate inquiry) - Artnet News, "Leon Black Agrees to Pay $62.5 Million… Virgin Islands" (early 2023 settlement, immunity, no admission of guilt) - Senate Finance Committee, 23 March 2026 release (Wyden letter: "primary sources of income," 30x, sha [...] ============================================================================ ANALYSIS: The lender of next-to-last resort URL: https://l0g.fr/en/analysis/the-lender-of-next-to-last-resort/ Canonical French source: https://l0g.fr/posts/fhlb-preteur-avant-dernier-ressort/ Date: 2026-07-28 (reviewed 2026-07-28) Topics: international, banks, private credit, risk, united-states ---------------------------------------------------------------------------- Eleven public banks, daughters of a 1932 law against the Great Depression, today manage more than seven hundred billion dollars of advances without almost anyone, outside Washington, knowing their name. They do not lend to the public, they barely finance homes anymore, and yet the state implicitly guarantees their debt, a privilege the Congressional Budget Office prices at close to seven billion dollars a year. That windfall benefits homebuyers less and less, and an insurer owned by Apollo more and more, one that borrows at the subsidised rate to fund private credit. Behind that drift hides an even heavier legal privilege: when a member bank collapses, these institutions come before everyone, the FDIC included, that is, before the fund that guarantees your deposits. The Federal Home Loan Bank system is the blind spot of American finance. It deserves a look. Eleven banks, one privilege, no scrutiny The system was born in 1932, at the trough of the Depression, to irrigate with liquidity the thrifts that financed home ownership. A century later, it survives as eleven regional banks, cooperative and member-owned, overseen by a discreet regulator, the Federal Housing Finance Agency. Their trade fits in one word: the advance, a collateralised loan made to a member, a bank, a credit union or an insurer. At the end of March 2026, these advances reached about $724 billion, on an aggregate balance sheet exceeding a trillion. These banks take no deposits. They fund themselves on the markets, at an abnormally low cost, because investors take for granted that Washington would never let them fall. That guarantee has never been legislated; it is simply anticipated, and the anticipation alone makes them government-sponsored enterprises, on a par with Fannie Mae or Freddie Mac. The Congressional Budget Office has put a price on that advantage: a net federal subsidy of $6.9 billion for fiscal 2024 alone, in a range of $5.3 to $8.5 billion, owing mostly to the implicit guarantee that lowers their rates, plus tax exemptions. Seven billion dollars of public money, every year, for a mechanism whose very existence escapes debate. The super-privilege of 1987 The singularity of the Federal Home Loan Banks lies not in their subsidy, shared with other GSEs, but in a legal prerogative unique of its kind. The Competitive Equality Banking Act of 1987 granted them a super-lien: when a member fails, their claim on the collateral ranks ahead of every other creditor, the FDIC and the deposit insurance fund included. Combined with an over-collateralisation requirement, that priority makes their losses practically nil. A Federal Home Loan Bank, so to speak, never loses. The safety has a price, and the reasoning is worth following to its uncomfortable conclusion. If these banks never lose, the loss does not vanish for all that: it is displaced. When an institution collapses, the best collateral goes first to repay the advances, and the FDIC inherits the residue, an asset stripped of its firmest pledges. The safety of the Federal Home Loan Banks is therefore bought by subordinating everyone else, starting with deposit insurance, itself funded by the surviving banks and, ultimately, backed by the taxpayer. Some lawyers object that this lien is nothing exorbitant, no more "super" than the security of an ordinary secured creditor; the objection is right on principle, but it misses the practical effect. The point is not that the bank seizes unlawfully, it is that at the end of a failure, the quality pledges are already spoken for, and the hole that remains falls to someone else. March 2023, the window that delays the fall The mechanism stayed theoretical until the spring of 2023, when it showed itself at full scale. Eaten away by unrealised losses on their bond portfolios, Silicon Valley Bank, Signature Bank and First Republic turned to their Federal Home Loan Bank to hold on. The advances let them delay the asset sales that would have forced the accounting recognition of losses, and dodge, for a while, the supervisor's hard stare. Then they failed. After they were placed in resolution, the Federal Home Loan Banks were repaid in full for SVB's and Signature's advances, thanks to the super-lien, while the FDIC absorbed the bill. New York Fed economists had named that role well before the crisis, in a study whose title has become proverbial, the lender of next-to-last resort. The phrase is surgical. The central bank is the lender of last resort, the one that steps in when everything else has given way; the Federal Home Loan Banks slip in just before, lending to already-shaky institutions they keep afloat without bearing their risk. The regulator itself half-admitted it in its major 2023 review: the system, the FHFA writes, was not designed to be a lender of last, or next-to-last, resort for struggling institutions. The admission is notable, coming from the supervising authority. It does not say the role was refused, only that it was never intended. A serious defence exists, and it deserves its place. In 2023, Federal Home Loan Bank liquidity cushioned a panic, bought time, perhaps averted a more brutal contagion. Time, though, is not always a public good: granted to an already-insolvent bank, it widens the hole rather than filling it, and it postpones a resolution that would have cost less earlier. Liquidity that saves a sound institution is a blessing; the same liquidity, lent to a doomed one, only displaces and enlarges the bill. From the thrift to the carry trade There remains the question of the recipient, and this is where the investigation turns sharp. The historic client of the Federal Home Loan Banks was the thrift financing home loans. The rising client finances no houses: it is the insurer owned by a private equity firm. Insurance companies' borrowings from the system hit a record $177.8 billion last year, up 10%, nearly a quarter of all advances. And the first among them bears a name familiar to our readers: Athene, Apollo's insurance arm, whose advances rose from $15.6 billion in 2024, seventh place, to $28.2 billion at the end of March 2026, third in the entire system. Apollo makes no secret of it, and its frankness is almost disarming: the group describes these advances as an investment spread strategy. Borrow at the rate lowered by the public guarantee, reinvest in better-paying assets, private credit most often, pocket the difference. The full machinery we have described elsewhere from another angle: the insurer backed by an asset manager issues annuities, lightens its capital requirements through Bermuda reinsurance, and pushes its investment towards private credit, all within entities of one group. The super-lien adds a final touch: the public housing guarantee now funds, at low cost, a yield arbitrage for the benefit of a private equity shareholder. The loan the bank judged too heavy, the retiree's annuity, the subsidised advance of a housing GSE: the same pipes, borrowed by the same players, that we followed from the credit card to the annuity. The arithmetic of a forgotten mission The figures, set side by side, return a verdict no speech can undo. In 2024, the system paid $3.7 billion in dividends to its member banks and about $350 million to its affordable housing programmes, more than ten times less. Set against the $6.9 billion public subsidy the CBO ascribes to it, the housing contribution becomes a decimal. A GSE designed to house Americans returns to its members, as dividends, more than ten times its housing contribution, and captures seven billion of public money to do so. The FHFA did try to close one door: its 2016 rule excluded captive insurers, those shells created to reach the window. But full-fledged insurance subsidiaries, like Athene's, remain eligible, and they poured through the next one. The regulator diagnosed the drift in its 2023 review and promised a remedy; two years on, insurers' borrowing sets records. Between the diagnosis and the cure, the gap is measured in tens of billions. The system's defences, and their blind spot The picture calls for its qualifications, and they are real. The advances are over-collateralised, so that the Federal Home Loan Banks have, historically, almost never taken a credit loss; the system is sound, well run, and has cost the federal budget not a direct cent. Insurers argue, not without reason, that these advances are a legitimate asset-liability management tool, a stable source of liquidity for long-term commitments. And the implicit guarantee stays, precisely, implicit: no public outlay as long as the system holds. The taxpayer pays nothing, today. The blind spot of these defences is that they all answer the wrong question. The soundness of the advances is not in doubt; their destination is. That the system never loses is precisely the problem, because that invulnerability is the product of a priority that makes others lose. And that the taxpayer pays nothing today says nothing of the price they would pay one day of real stress, when the implicit guarantee would turn explicit. The risk of the Federal Home Loan Banks is not that of a sudden crash, a default, a rout. It is slower and quieter: a public subsidy captured for a private arbitrage, a legal priority that rewrites in advance the hierarchy of a bank failure's losers, and a housing mission reduced to ornament. None of it makes the headlines, because none of it breaks. The danger here does not wear the face of collapse, it wears that of drift, and drift trips no alarm. An investor mapping the risks of American finance is used to hunting for bombs. This one is not a bomb, and that is exactly why it deserves the look: the most subsidised, most senior and least watched organ of the system does not threaten to explode, it merely serves, in silence, a purpose quite other than the one for which the state guarantees it. The lender of next-to-last resort will not cause the next crisis. It will only have, quietly, named its losers in advance. --- Sources - Congressional Budget Office, "The Role of Federal Home Loan Banks in the Financial System", March 2024 (net federal subsidy of $6.9bn in 2024, implicit guarantee, exemptions) - Federal Housing Finance Agency, "FHLBank System at 100: Focusing on the Future", November 2023 (first major review in decades; the system was not designed as a lender of last or next-to-last resort) - Federal Reserve Bank of New York, Staff Report 357, "The Lender of Next-to-Last Resort" (Ashcraft, Bech, Frame) - U.S. Government Accountability Office, "Federal Home Loan Banks: Actions Related to the Spring 2023 Bank Failures", GAO-24-106957 (SVB, Signature and First Republic; advances repaid in full) - Congressional Research Service, "The Federal Home Loan Bank (FHLB) System and Selected Policy Issues", R46499 (structure, members, super-lien of the Competitive Equality Banking Act of 1987) - American Banker, "Is private credit a new risk for the Federal Home Loan banks?" (insurers, Athene, "investment spread strategy") - American Banker, "Private credit making itself at home at Federal Home Loan Banks" (insurers' borrowing at $177.8bn, up 10%) - Bloomberg, "Apollo's Athene Rises to Third-Largest Borrower in FHLB System With $28 Billion", 14 May 2026 - American Banker (opinion), "Federal Home Loan bank liens are no more 'super' than other creditors'" (counterpoint on the super-lien's reach) - Consumer Federation of America, "Private Equity is Gobbling Up Subsidized Housing Loans" ($3.7bn in dividends versus ~$350m to affordable housing) - Federal Reserve, Financial Accounts (FHLB Advances series, FRED): advances outstanding of about $724bn in Q1 2026 ============================================================================ ANALYSIS: The risk that goes in circles URL: https://l0g.fr/en/analysis/the-risk-that-goes-in-circles/ Canonical French source: https://l0g.fr/posts/transfert-synthetique-risque-srt-cercle/ Date: 2026-07-28 (reviewed 2026-07-28) Topics: international, banks, private credit, securitisation, risk ---------------------------------------------------------------------------- There is in US finance a sleight of hand so elegant it becomes unsettling. A bank holds a portfolio of loans it would rather not carry, because they weigh on its regulatory capital. The intuitive solution would be to sell the loans. It does something subtler: it keeps the loans on its balance sheet but sells their risk to an outside investor, a hedge fund or a private credit fund, which agrees to absorb the first losses in exchange for a double-digit return. The loan does not move an inch. The risk, though, seems to evaporate. And the capital tied up behind that risk is suddenly freed, ready to fund new loans. This operation is called synthetic risk transfer, and it has quietly hedged more than a trillion dollars of bank loans. The problem is not that it exists. The problem is what happens when you follow the risk to the end: it often comes back, through a side door, into the very bank that thought it had shed it. The sleight of hand Let us start with the mechanics, because everything else follows. In an SRT, the bank keeps legal ownership of its loans but buys protection against their default. Technically, it proceeds like a securitisation, cutting the portfolio into risk tranches, but without selling the assets: it sells only insurance on the first-loss tranche, the one that absorbs defaults first. The most common instrument is the credit-linked note: the investor pays capital upfront, collects a high coupon, and gets its capital back at maturity, less any losses on the reference portfolio. If the loans perform, it pockets a double-digit return; if they sour, it loses its stake, and the bank is compensated. The intended effect is not economic, it is regulatory. By transferring the first-loss tranche, the bank can show its supervisor that it has shed most of the portfolio's credit risk, and therefore reduce the capital it must hold against it. Recent deals let banks cut their capital requirements by an average of 43 basis points, a considerable relief on balance sheets of hundreds of billions. The loan stays on the books, the client sees nothing, but the capital behind it is freed. It is the logical extension of the fight over capital we described in our analysis of the Basel III rollback: what regulation demands on one side, engineering takes back on the other. A market steps out of the shadows Long confidential, reserved for a few European banks and a handful of specialist funds, the SRT has become a mass market. The trigger, on the American side, dates to 2023, when the Federal Reserve recognised credit-linked notes as eligible for capital relief. US banks rushed in, to the point of now accounting for nearly 30% of global flow. The scale is dizzying: by the end of last year, banks had transferred the credit risk of more than €905 billion, roughly a trillion dollars of loans, up 26% year on year. The reference pool of European deals hit a record €260 billion in 2024, and US issuance rose from $29 billion to $41 billion in a year. On the other side of the table, a constellation of buyers has specialised. The big names of private credit and hedge funds, Magnetar, Ares, Apollo, Blue Owl, KKR, Blackstone, compete for the first-loss tranches, whose target returns reach the mid-teens. Some have built dedicated lines of tens of billions. In December 2025, Blackstone took the first-loss protection on a €2 billion portfolio of large corporate loans from ABN AMRO. These investors take real risk and book real losses when a portfolio sours; on that point, the market works as advertised. The awkward question: how much risk, really? This is where the elegance starts to crack. An SRT transfers only one tranche, usually thin, the first loss. The bank keeps the senior tranche, that is, the catastrophe risk: the one that materialises only if losses exceed the cushion sold. In normal times, that tail risk is negligible, and the transfer looks complete. In a correlated shock, where many loans default at once, losses can pierce the first-loss tranche and climb back to the bank, precisely when it thought itself protected. Researchers put the question bluntly in a note with a telling title, "synthetic, but how much risk transfer?": the capital relief is immediate and certain, the disappearance of the risk is partial and conditional. Add counterparty risk. The protection is only worth something if the investor can pay. In structures backed by a credit-linked note, the capital is paid upfront and locked, which limits that risk; but in unfunded variants, where the protection rests on a mere contractual promise, the bank stays exposed to the failure of its insurer. And these insurers are leveraged funds, less regulated than banks, and that is where the structure reveals its hidden flaw. The circle Here is the heart of the investigation, and the reason an attentive investor should worry. To buy these first-loss tranches at attractive returns, private credit funds and hedge funds use leverage, that is, borrowed money. And from whom do they borrow? Often from the banks themselves. A bank sells the risk of its loans to a fund, and another bank, sometimes the same one, lends that fund the money to buy the protection. The risk goes out the door and comes back through the window. The Financial Stability Board has put a name on this: "circles of risk", where bank credit lent to the funds that buy back bank risk reintroduces that risk into the system. The International Monetary Fund devoted a working paper to the mechanism with a limpid title, "Recycling Risk". The consequence is twofold. First, the capital relief can be partly illusory: the bank shows less risk, but the banking system as a whole still carries as much, or more, since it now has a leveraged intermediary in the middle. Second, the risk has changed regulator: leaving a supervised, marked and capitalised bank balance sheet, it landed at a less regulated fund, whose leverage amplifies losses, and which we file under non-bank financial intermediation. It is the same translation we documented for consumer credit, from the credit card to the annuity: the risk does not disappear, it migrates towards the least visible compartment. The new fuel: AI data-centre debt If this already-strained market suddenly worries more, it is because of what is pouring into it. US banks have lent colossal sums to finance the construction of data centres for artificial intelligence, a debt whose architecture we described in our investigation of the debt behind AI. That exposure has swelled to the point of becoming, in a Bank of America survey, the top systemic credit risk named by 48% of managers for 2026. What do banks do with this parcel that has grown too heavy? They transfer it. Morgan Stanley, Citi, JPMorgan and Goldman Sachs have begun offloading the risk of their AI infrastructure loans to private credit, hedge funds and pension funds via SRTs. The head of credit risk sharing at Man Group sums up the worry in a phrase: the sums involved are "out of scale to anything we've thought about, ever." The structure then becomes doubly circular. A bank lends to a data-centre developer; it transfers the risk of that loan to a private credit fund; that fund is sometimes the same one financing, elsewhere, the construction of the data centre or the AI company that will fill it. The risk turns inside a small circle of players who carry, through different vehicles, both ends of the same chain. If the AI bet disappoints, it will not be independent counterparties that absorb the shock, but a handful of funds exposed everywhere at once. The last circle: when risk turns liquid The chain does not stop at the private credit fund. It now has an extra link, perhaps the most vertiginous, because it brings the risk all the way to the ordinary saver in a form that erases every trace of it: the ETF, that listed index fund which trades on an exchange like a stock. To grasp the danger, one must first grasp how an ETF manufactures its liquidity, because that is exactly where the trap closes. An ETF does not keep its assets in a frozen vault. Its liquidity rests on a discreet mechanism, the creation and redemption of shares. Authorised intermediaries, the authorised participants, can at any moment create new shares by delivering the underlying securities to the fund, or destroy shares by taking those securities back. This back-and-forth anchors the share price to the portfolio's real value: if the share trades too dear, more are created to bring the price down; too cheap, some are destroyed to support it. The system is ingenious, but it rests entirely on one condition: that the underlying securities themselves buy and sell without friction. As long as the underlying is liquid, so is the share. Yet that is precisely what private credit lacks. A private loan does not sell in a day, often not in a month; it has no continuous market price, only an estimate, a subject we dug into in our analysis of one asset, two prices. Wrapping such assets in an ETF amounts to promising daily liquidity on assets that have none. The creation-redemption mechanism jams the moment too many holders want out at once: the authorised participants cannot liquidate the underlying fast enough, the share detaches from its theoretical value, and the exit, wide in appearance, turns out to be narrow. It is a risk transfer of a new kind, no longer credit but liquidity, and it is more insidious because it looks painless as long as flows come in. Bond ETFs did, it is true, come through the March 2020 shock without breaking, their discount later closing; but they held listed bonds, not private loans stripped of a price. This is no textbook hypothesis. The first widely distributed private credit ETF was launched in late February 2025 by State Street with Apollo. To keep its liquidity promise, it was allowed to hold between 10% and 35% of private assets, well beyond the usual 15% illiquid limit in an ETF, thanks to an agreement under which Apollo commits to buying those assets back, which immediately worried the SEC. The regulator asked the only question that matters: if a single player, Apollo, provides the liquidity by buying back assets it originated itself, at what price will it do so, and what happens the day it stops buying? The promised liquidity then no longer rests on a deep market, but on the goodwill of a single counterparty, in a position of conflict of interest. Here is the point to remember, and it reaches well beyond this one fund. In these wrappers, liquidity is not a property of the assets, it is a promise made by a counterparty. And a promise is worth only as long as the one making it has an interest in keeping it. Yet that interest evaporates exactly when it would be needed, when everyone wants to sell. The journey of risk, which we followed to a retiree's annuity from the credit card to the annuity, now sometimes ends in a retail brokerage account, in an instrument with the look of a stock and the substance of an illiquid loan. The saver thinks they hold liquidity; they hold, in reality, the last link of a chain that began with a loan a bank judged too heavy to keep. The other reading: a legitimate tool, not a bomb It would be dishonest to paint the SRT as pure artifice, because the instrument has real virtues, and the regulators themselves have not banned it. The first argument in its favour is that it achieves genuine risk-sharing. A bank heavily concentrated on one sector, commercial real estate, AI, leveraged credit, can, through the SRT, redistribute that concentration to long-term investors, insurers and pension funds, who seek precisely that return and hold their positions to maturity. Seen that way, the SRT makes the system more resilient, not less, by dispersing a risk otherwise lodged in a few balance sheets. The second argument is that these deals are bilateral, documented and known to the supervisor, unlike the opaque derivatives of before 2008. The Basel Committee published in February 2026 a detailed report on these markets, a sign that authorities are following them closely rather than discovering them after the fact. Its conclusion is not a ban, but tighter monitoring, possible limits on capital relief, and better coordination between bank and non-bank supervisors. The third argument, made by the buying funds, is that banning leverage on these deals would dry up financing useful to the economy without removing the underlying risk. The first-loss tranche finds sophisticated buyers who know what they are buying, and this market has so far absorbed its losses without systemic incident. What the circle will not forgive These counterpoints hold in calm times. They all say the same thing: as long as losses stay within the thickness of the sold tranche, as long as the buying funds can pay, as long as leverage stays contained, the SRT is a prudent management tool. The problem is that these three conditions degrade together, and precisely at the wrong moment. A correlated shock, on AI debt for instance, would do three things at once: it would pierce the first-loss tranches and send losses back to the banks; it would test the leveraged funds' ability to honour their protection; and it would push the lending banks to cut those same funds' leverage lines, drying up the market just when it needs to work. The three nets would tear at the same time. That is why this mechanism should be read not as a fraud, but as an optimisation at the seam between the regulated and the unregulated, where risk is not removed but relabelled. The investor contemplating a US bank's freed capital should not ask "where did the risk go", but "who holds it now, and is it the bank that lent them the money to hold it". The answer, more and more, draws a circle. And a circle, in finance, has an unpleasant property: it has no end by which to hold it when everything starts turning the wrong way. To judge a bank's real soundness, one must now read what it has transferred as much as what it holds, an exercise our guide on a bank's soundness no longer exhausts on its own. The risk that goes in circles always ends up back where it started. --- Sources - Bloomberg, "Banks Offload $1 Trillion Loan Risk to SRT Investors, IACPM Says", 4 June 2026 (hedged exposure €905bn / ~$1tn, +26% year on year) - Bloomberg, "Banks Love Significant Risk Transfers, and That Has Regulators Worried", 8 December 2025 (FSB warning on "circles of risk") - Risk.net, "SRT issuance hits €260bn as capital relief grows" (record reference pool in 2024, average 43 basis points of capital relief) - Philadelphia Fed, "Banking Trends: Synthetic Risk Transfers" (2023 Fed guidance on credit-linked notes, mechanics and rise) - Basel Committee, report on synthetic risk transfer markets, February 2026 (Jones Day summary: monitoring, possible limits on capital relief) - IMF, "Recycling Risk: Synthetic Risk Transfers", working paper 2025/200 (circularity and risk recycling) - SUERF, "Synthetic, but how much risk transfer?" (share of risk actually transferred, retained tail risk) - Fortune, "Morgan Stanley explores significant risk transfer for data center and AI infrastructure exposure", 4 December 2025 - Startup Fortune, "AI data center debt has climbed to the top of Wall Street's credit risk watchlist" (top systemic risk named by 48% of BofA managers) - ABN AMRO, "ABN AMRO announces significant risk transfer transaction with Blackstone", 11 December 2025 (first-loss protection on €2bn of corporate loans) - CNBC, "State Street, Apollo team up to launch first of its kind private credit ETF", 27 February 2025 (launch of the PRIV ETF) - WealthManagement, "State Street, Apollo's Private Credit ETF Raises SEC Concern" (10 to 35% private assets, Apollo as sole liquidity provider, SEC concerns on valuation and liquidity) ============================================================================ ANALYSIS: The deferred bill URL: https://l0g.fr/en/analysis/the-deferred-bill/ Canonical French source: https://l0g.fr/posts/privilege-exorbitant-addition-differee/ Date: 2026-07-28 (reviewed 2026-07-28) Topics: international, dollar, debt, united-states, risk ---------------------------------------------------------------------------- There is a magic trick in economics that only the United States has managed to sustain. A country that owes the rest of the world far more than it owns abroad should, in good logic, pay a net rent to its creditors every year. America did the opposite: a net debtor of more than $21 trillion, it nonetheless collected positive net income on its external holdings. Economists gave that anomaly a name, the exorbitant privilege, and for half a century they debated its magic without seeing it weaken. It is weakening now, and the mechanism behind it is jamming. The meal the world served America for free is starting to cost something. The trick rests on a balance-sheet asymmetry simple to state. The United States holds risky, rewarding assets abroad, equities and direct investment, while the world holds in return safe, low-yielding American assets, Treasuries first. The US net international investment position was minus $21.27 trillion at the end of March 2026, $43.37 trillion of assets against $64.64 trillion of liabilities, a chasm. But the higher return on American assets abroad offset the gap in holdings, so the income balance stayed positive despite the debt. That is what we described in our guide on the balance of payments: the debtor that earns money, a paradox that financed the American deficit for free across two generations. The mechanism that jams Two forces are closing the trap. The first is the rise in global rates. As long as the world lent to the United States at near-zero rates, the cost of its liabilities stayed negligible and the return differential ran full. Since rates normalised, every Treasury held by foreigners costs more, and the service of the external debt swells. The second is the accumulation of liabilities itself: by dint of financing current account deficits, the stock of debt held by foreigners has grown to the point where its cost, even at a modest unit yield, ends up weighing heavily. The result is mechanical: the investment income balance, long in surplus, is coming back down towards zero and about to tip into deficit. That tipping is no small thing. The day the income balance turns negative, the American current account deficit stops being financed for free: the United States will have to pay the world a net rent, for the first time in decades, on top of borrowing to fill its trade deficit. The planet's most indebted country will become an ordinary debtor again, one that pays to be so. The dark matter The most troubling feature is what still keeps the balance afloat. A growing share of the positive net income comes not from a true return differential but from an accounting artefact: the profits US multinationals offshore into tax havens and repatriate as investment income. Economists call that component dark matter, income that inflates the accounts without matching any genuine yield superiority. Yet stripped of these offshored profits, the American income balance is already negative. The exorbitant privilege, in its recent form, thus rests partly on the tax optimisation of the tech giants, not on the magic of the dollar. It is a trompe-l'oeil privilege, propped up by a structure that the slightest reform of international taxation could deflate. The other reading: the privilege is not dead Before burying sixty years of American exception, one must grant the thesis its counterpoints, because they are real. The first is that the external position has recently improved: from a trough of nearly minus $26.5 trillion at end-2024, it recovered by about $5 trillion to minus $21.27 trillion. That upturn owes mostly to valuation, the rise in American equities held by foreigners cutting the other way, but it recalls that the figure is not on a linear path to the abyss. The second is that the income balance, even in decline, stays close to balance: a tip into slight deficit is not a crisis, it is a slow, absorbable deterioration. The third is that the dollar remains the reserve currency, and as long as the world wants dollars, the United States can finance its deficit on terms no other debtor would obtain. The structural demand for American debt, which we track in our analysis of the marginal buyer, has not vanished. But the direction is structural The counter-reading reassures on the pace, not the direction, and it is the direction that matters. For the two forces gnawing at the privilege, high rates and growing liabilities, are not cyclical but structural, and they reinforce each other. The more the federal debt swells, the more the stock of Treasuries held by foreigners grows; the higher rates stay, the more it costs. The return differential that made the magic is compressing from both sides at once. And when the income balance turns durably negative, the effect becomes cumulative: paying a rent to the world widens the current account deficit, which swells the external debt, which heavies the rent. Yesterday's virtuous circle, where debt cost nothing, slowly turns into a vicious one. That is why this erosion, invisible and slow, speaks to the signals we track elsewhere. It is the hidden face of the return of the term premium: if investors demand more to hold American debt, the cost of the liabilities rises and the privilege erodes accordingly. It feeds the same underlying distrust as the debasement trade, that of a sovereign issuing without counting. The exorbitant privilege was the silent subsidy that let America borrow the world's savings without ever presenting the bill. That bill has not disappeared, it was deferred. It is starting, line by line, to arrive. --- Sources - Bureau of Economic Analysis, "U.S. International Transactions and Investment Position, 1st Quarter 2026" (net international investment position -$21.27tn, assets $43.37tn, liabilities $64.64tn, recovery from the end-2024 trough) - Econbrowser, "NIIP and Primary Income: Dark Matter vs. Exorbitant Privilege" (income balance coming down to zero, already negative excluding offshored profits) - Council on Foreign Relations, "The U.S. Income Balance Puzzle" (erosion of the income balance, role of dark matter and profit-shifting) ============================================================================ ANALYSIS: From the credit card to the annuity URL: https://l0g.fr/en/analysis/from-the-credit-card-to-the-annuity/ Canonical French source: https://l0g.fr/posts/de-la-carte-a-la-rente-qui-detient-risque-consommateur/ Date: 2026-07-27 (reviewed 2026-07-27) Topics: international, private credit, securitisation, united-states, risk ---------------------------------------------------------------------------- A household stops repaying its auto loan somewhere in the Midwest. The scene is ordinary, and we showed elsewhere how much it repeats, subprime auto delinquency being at its highest since the 1990s. The question left open is not whether the American consumer cracks, it is who takes the loss when they do. The answer is counterintuitive: almost never the bank that lent. The risk has been sliced, packaged, resold, and it keeps travelling, tranche by tranche, all the way to an insurer's balance sheet and a retiree's annuity. That journey is the most important and least told story in consumer credit. The starting point is a widespread misunderstanding. One imagines the subprime lender bears the risk of its loans, like a classic deposit bank. It does not. Specialised lenders do not keep these loans on their books: they securitise them into ABS and sell the tranches to institutional investors around the world, bond funds and pension funds first. The originator collects a fee and offloads the risk almost at once. Understanding where that risk goes means following the chain, link by link. The securitisation waterfall The first link is mechanical. The receivables, auto loans, card balances or instalment payments, are pooled in a trust bankruptcy-remote from the originator, which issues securities backed by those assets, the ABS. These securities are cut into ranked tranches, exactly as in the CLOs our guide describes. The senior tranche, rated AAA, is paid first and absorbs losses last; the equity tranche, at the very bottom, takes the first defaults in exchange for the highest yield. Between them, mezzanine tranches. The rule is simple: when a borrower defaults, the loss climbs from the bottom up, and the equity tranche is designed to be wiped out first to protect senior investors. That dispersion has one reassuring consequence and one misleading one, and both must be held at once. The reassuring one: the AAA rests on a cushion of subordination, and it would take massive losses to reach it. The misleading one: the risk did not vanish, it concentrated in the lower tranches, and those tranches found an eager buyer. The new buyer: private credit That buyer is private credit, and its appetite has changed the nature of the market. The big alternative asset managers, Apollo in the lead with more than $1 trillion in assets at the first quarter of 2026 and the leading private credit firms together weighing more than $3.4 trillion, have rushed into asset-based finance, the compartment that securitises the receivables of daily life. They no longer just buy the tranches, they settle across the whole chain. KKR signed a €6 billion forward-flow deal with PayPal to fund its instalment lending directly, and launched the first BNPL securitisation in Europe; the Pagaya platform issued about $300 million of securities backed by Klarna loans, arranged with Apollo's help; Affirm has closed more than a dozen ABS deals on its point-of-sale loans. The shift is decisive. Through these forward-flow deals, private credit no longer just buys the risk once created, it funds the loan's origination. In other words, the funds' reach for yield directly feeds the subprime credit expansion we described from the borrower's side, those raised card limits and that proliferating instalment lending. The one who will bear the loss is also the one who supplied the ammunition. This circularity, private credit lending in order to securitise what it will hold, is the newest and least discussed feature of the cycle. The last link: the annuity What remains is where the journey ends, and the answer closes the loop in an unsettling way. Private credit does not hold these assets on its own account: it lodges much of them on the balance sheets of the insurers it controls. Apollo recycles the premiums of its Athene affiliate into private credit and asset-based finance; the whole sector follows, with US life insurers' private credit holdings reaching $849 billion in 2024, more than double their 2014 level. These insurers seek long-dated yield to back their annuity commitments, and a growing share runs through offshore reinsurance structures, often Bermudian, whose opacity we described in our investigation of life insurers and private credit in Bermuda. The journey is therefore complete. Starting from the dashboard of a used car financed at a high rate, the risk has crossed a securitisation trust, a mezzanine tranche, a private credit fund, a reinsurance captive, to settle at last beneath the annuity of a retiree who never went near a subprime loan. From the credit card to the annuity, the risk changed hands five times without ever leaving the system, and at each step it became a little harder to see. The other reading: dispersion is a strength Before crying the next 2008, one must grant this construction what is solid in it, because the subprime analogy is misleading. Several counterpoints hold. First, dispersion is precisely what was missing in 2008. The mortgage risk of the time was concentrated, correlated and lodged in highly leveraged banks that had to sell in a panic. Here, consumer credit is spread in small tranches across hundreds of investors, and consumer ABS is an old, tested asset class that weathered the crisis better than mortgage CDOs. Second, the matching makes sense: an insurer that must pay annuities over thirty years has good reason to hold long, illiquid assets it intends to keep to maturity, never forced to sell. A holder who does not sell does not spread panic. Third, subordination works: as long as losses stay within the thickness of the lower tranches, the pension funds' AAA does not move, and that is exactly what it is for. But dispersion hides a re-concentration The counter-reading has its limits, though, and they are serious. The first blind spot is that apparent dispersion masks a real re-concentration. The risk leaves thousands of banks to gather in a handful of mega-managers that now originate, structure, hold and insure the same asset: diversification across investors comes with concentration across firms. The second is valuation: lodged in private credit funds, these assets are marked to model, often near par, not to market, which delays loss recognition, a problem we dug into in our analysis of one asset, two prices. The third is funding: the forward-flow deals and warehouse lines that feed origination can be cut in stress, abruptly drying up credit where it is most fragile. The fourth, the most disturbing, is one of identity: the ultimate risk-bearer is an annuitant, through an offshore structure they do not understand and that no state supervisor fully oversees. The conclusion is therefore neither alarm nor relief, but a shift of gaze. The American consumer's risk did not grow by changing address, but it became more opaque, more concentrated among its managers and slower to reveal itself. The real question, once the borrower's fragility is established, is not whether the banking system will shake, it barely holds this risk anymore, but what happens the day an annuitant discovers that their retirement rested, through five intermediaries, on the punctuality of a stranger repaying their car. The risk left the light of bank balance sheets for the shadow of private credit. It did not vanish. It is just waiting somewhere else, where no one is looking. --- Sources - Wolf Street, "Auto Loan Balances, Debt-to-Income Ratio, and Delinquencies of Subprime & Prime Auto Loans in Q1 2026" (specialised lenders securitise subprime auto into ABS sold to investors; prime delinquency 1.9%, subprime 60-day at 6.90%) - GlobalCapital, "KKR's debut lays foundation for BNPL ABS asset class in Europe" (€6bn forward flow with PayPal, first BNPL securitisation in Europe) - HedgeCo, "Apollo Tops $1 Trillion in AUM and Moves Toward Daily Private Credit Pricing", May 2026 (Apollo above $1tn, Athene, asset-based finance) - American Banker, "Is private credit a $2 trillion-dollar insurance timebomb?" (life insurers' private credit at $849bn in 2024, more than double 2014; offshore reinsurance) ============================================================================ ANALYSIS: The average consumer does not exist URL: https://l0g.fr/en/analysis/the-average-consumer-does-not-exist/ Canonical French source: https://l0g.fr/posts/fissure-consommateur-americain-economie-en-k/ Date: 2026-07-27 (reviewed 2026-07-27) Topics: international, consumer, credit, united-states, risk ---------------------------------------------------------------------------- The aggregate number tells a calm economy. US household debt reached $18.19 trillion in the first quarter of 2026, a record, but spending is not buckling, employment holds, and equity markets sit on their highs. At that level of reading, the American consumer is solid, and that is the conclusion the hurried commentators keep. The trouble is that this single consumer does not exist. Beneath the average hide two Americas whose paths have never diverged so far, and the second is cracking without the first noticing. This is the signature of a K-shaped economy: the top of the distribution prospers, spends and holds up the aggregate on its own, while the bottom falls away under the double weight of cumulative inflation and the cost of credit. Well-off households keep clean balance sheets and keep buying; modest households now borrow less to invest than to get by. The average adds these two worlds together and draws a reassuring figure that describes no one. To understand where the American consumer is really heading, do not look at the average, look at the tail. The car, the canary in the mine The clearest signal comes from auto credit, that mass market touching nearly every home. In January 2026, 6.9% of subprime borrowers were more than sixty days past due, the highest level since the 1990s and nearly double the historical average. The New York Fed recorded the highest auto delinquency rates in its entire series. And the deterioration is ongoing, not behind us: the median credit score on new auto loans slipped from 724 to 716 in the last quarter of 2025, the steepest quarterly drop in years, a sign that lenders are moving down the risk ladder at the very moment it grows fragile. The student-loan wake-up To that strain is added a timing shock: the resumption of student-loan repayments, long suspended, hits an already fragile generation head-on. The share of balances ninety or more days past due runs around 17%, and serious defaults worsened to 10.3% at period end, from 9.6% at the close of 2025. It is not the same borrowers everywhere: among 18-to-29-year-olds, the serious-default rate is near 5%, roughly double a year earlier and the highest of any age group. Young households, stacking high-rate revolving debt, a thin savings cushion and maximum exposure to the student reset, absorb the shock for everyone. The most worrying trait is the simultaneity. Borrowers in default are rarely so on a single loan: they pile up arrears on the card, the car and the student loan at once, the mark of a perfect storm where rising prices and drained savings combine. This is not an isolated incident by product, it is a household sinking on every front at the same time. The subprime that hides the subprime Then comes the indicator that seems to deny the alarm: the credit-card delinquency rate, calm on the surface, at 2.9% across commercial banks. But that calm is deceptive, because it is bought on credit. Over twelve months, subprime card openings jumped 18.6%, and the limits granted to that segment rose 37.6% versus the prior year. In other words, more credit is being extended to the most fragile households, which mechanically pushes back the moment of delinquency: as long as the limit rises, the account holds. The aggregate rate stays tame because fresh debt covers old debt, until the day it can no longer. To this is added the least visible debt of all, that of instalment payment. Buy now, pay later is poorly captured by the credit bureaus and barely shows up in household debt statistics, hence its nickname of phantom debt. It nonetheless weighs on the cash flow of the same households already stacking auto and student arrears. The overall picture is of a bottom of the distribution piling on layers of credit, some of which we do not even measure, to finance not projects but daily life made dearer by the comeback of US inflation. The other reading: a social fracture, not yet a systemic shock Here the analysis must guard against catastrophism, because from these figures a 2008 remake is too quickly drawn, and that would be an error of scale. Several counterpoints hold firmly. First, the aggregate really is contained. A 2.9% card delinquency is nothing like a crisis; it remains close to its long-run norm. The top of the distribution, which concentrates most of the spending, shows clean balance sheets and a debt-service-to-income ratio well below that of 2007. Subprime auto and the cards of fragile households make up only a modest fraction of total credit outstanding, and above all their risk is dispersed, securitised in small tranches spread across many investors, a world away from the correlated mortgage concentration that blew up the system nearly twenty years ago. The consumer fracture is first a social and distributional problem, not the fuse of a financial crisis. Second, the K is not immutable. Some measures show a recent narrowing of the spending gap between income groups, a slightly less clear-cut configuration than the one-way worsening the word "fracture" suggests. The bottom is falling away, but it is not collapsing, and a still-firm labour market remains the best firewall: as long as employment holds, delinquency stays manageable. But that is exactly where the shoe pinches, and the antithesis has its limits. The stress at the bottom of the distribution is a leading indicator, not a mere social blind spot: historically, subprime delinquency turns before unemployment rises, because it is the households without a cushion that crack first. And the risk, dispersed as it is, has not vanished: it has changed address, migrating towards non-bank intermediation, specialised auto lenders, BNPL platforms and, increasingly, the private credit that buys up these receivables. The question is not whether the average consumer is fine, it has no meaning. It is how long the top of the distribution can carry the GDP while the bottom carries the risk, and who will hold the bill when the labour market itself finally bends. The average consumer does not exist. The two who compose it have never been so far apart. --- Sources - Equifax, "U.S. Consumer Debt Hits $18.19 Trillion in Q1 2026" (record household debt, surge in subprime cards) - Bridgeforce, "Auto Loan Statistics Show Market Stress in 2026" (subprime auto 60-day-plus delinquency at 6.9% in January 2026, highest since the 1990s; median score 724 to 716) - Protect Borrowers, "American Families Hit Record Levels of Financial Distress" (student loans: 90-day-plus at ~17%, serious default 10.3%, 18-29 at ~5%, multiple arrears) - WalletHub, "Credit Card Delinquency Rates and Charge-Offs for 2026" (card delinquency at 2.9% across commercial banks) - American Default, "Credit Card Default Statistics 2026" (subprime card openings +18.6%, limits +37.6% year on year) ============================================================================ ANALYSIS: The debasement hangover URL: https://l0g.fr/en/analysis/the-debasement-hangover/ Canonical French source: https://l0g.fr/posts/debasement-trade-gueule-de-bois-or/ Date: 2026-07-27 (reviewed 2026-07-27) Topics: international, gold, bitcoin, reserves, dollar ---------------------------------------------------------------------------- It was the bet that seemed unable to lose. In late January 2026, the ounce of gold closed at an all-time high, just short of $5,600, silver had passed $120, and bitcoin was sitting on its autumn records. One logic drove all three: take shelter in supply-constrained assets to flee a currency that abyssal deficits and runaway debt would eventually debase. Wall Street had christened the move the debasement trade. It looked as solid as a truism. Then it broke, and the break teaches more than the rally did. Six months on, the scene is unrecognisable. Gold has fallen back below $4,000, down about 28% from its January peak; silver has dropped more than half, under $59; bitcoin has slid below $62,000, half its record, brushing its two-hundred-week moving average. The three supposed havens were swept away together, while a single asset held the lead: US equities, pulled by semiconductors and memory. The bet on the end of money lost to the bet on artificial intelligence. What made the trade fold The turn is no mystery, and it comes down to a single variable: real rates. The debasement trade thrives when holding cash is costly, that is, when inflation eats into low rates. It collapses the moment the central bank hardens its tone. That is exactly what the new-look Federal Reserve did. Under Kevin Warsh's chairmanship, markets began pricing two rate hikes by March 2027, lifting the Fed funds towards 4.00% to 4.25%, a shift we saw beginning at his first FOMC. A yield-free asset like gold, or a speculative one like bitcoin, sits poorly with the prospect of better-paid cash: its scarcity premium no longer offsets the opportunity cost. To that tightening was added a rotation. Capital did not leave risk, it changed horse: it fled gold and bitcoin for AI, the only story able to promise a real return rather than mere protection. The debasement trade was a defensive bet; against a tech bubble hoovering up all the performance, defence had no more buyers. The lesson is old and stubborn: a conviction trade is still a positioning trade, and positioning reverses. The debasement the prices do not see Here begins the interesting part, because concluding from this rout that debasement was an illusion would be too quick. The tactical bet blew up; the underlying move did not flinch. And that underlying move sits in a statistic that should have made the front page: for the first time since 1996, gold represents a larger share of central bank reserves than US Treasuries, about 27% against 22%. The world's dominant reserve asset changed, quietly, while commentators watched the spot price tumble. This shift is not a market accident, it is a repeated policy decision. Central banks are buying gold at a record pace, an extension of the move we track in our piece on the tonnes accumulated and de-dollarisation: Poland added 102 tonnes to lift its reserves to 550, Kazakhstan set an annual record, Brazil returned after four years away. And the intent is explicit: in the World Gold Council's 2026 survey, 89% of the central banks polled expect their gold reserves to rise over the next twelve months. The American backdrop feeds the reflex: a federal deficit above 6% of GDP and debt service exceeding $1 trillion a year. There are, then, two debasements. One is a trade, it lives and dies to the rhythm of real rates. The other is a regime, and it moves at the slow pace of official reserves. The other reading: neither bubble nor prophecy The symmetric trap must also be avoided, that of taking the structural signal for a prophecy of the dollar's collapse. Two qualifications are in order, and they cut both ways. First, against the sceptics: gold's 28% fall does not prove debasement was a mirage. A large part of the rout is a deflation of speculative positioning, not a revision of the fundamentals. That central banks keep buying even as the price falls is precisely the proof that their horizon is not the trader's: they vote with their reserves, indifferent to the monthly noise. Not confusing the price with the regime is the whole point, and JPMorgan and Morgan Stanley in fact keep high targets, around $6,000 an ounce ahead, driven by a weaker dollar and persistent official buying. Second, against the believers: gold's crossover above Treasuries in reserves is partly a valuation effect, not only a flow. When the gold price jumps, its share of reserves swells mechanically, without a single extra ounce being bought; the 1996 milestone owes as much to the metal's rise as to a deliberate rebalancing. And the very label of debasement is debatable: what central banks have done since 2022 looks less like a flight from inflation than a hedge against sanctions risk, a sovereignty bet after the freezing of Russian assets, which we separate out in our reading of the de-dollarisation narrative against the numbers. The real driver may not be fear of the printing press, but fear of depending on an asset another state can freeze. The synthesis fits in one sentence: the debasement trade got the timing wrong, not necessarily the direction. Real rates command the price in the short run, and they had the last word in 2026; deficits and geopolitics command the composition of reserves in the long run, and they keep pushing the other way. The investor who bought gold at $5,600 made a bad trade. The central bank accumulating it below $4,000 may be making a good bet. They are not the same people, not the same horizon, and that is why they can be wrong and right at once. The term premium waking up on US debt and the gold that states hoard tell, at bottom, the same distrust: distrust of a sovereign that issues without counting. The trade had its hangover. The distrust has not sobered up. --- Sources - CoinDesk, "Gold, silver and bitcoin tumble as debasement trade unwinds", 24 June 2026 (gold below $4,000, -28%; silver -50% under $59; bitcoin below $62,000; pricing of Warsh Fed hikes) - Mining.com, "Gold overtakes US Treasuries in global reserve shift", 2026 (gold ~27% of central bank reserves versus ~22% for Treasuries, first since 1996; ECB, IMF, World Gold Council data) - Crux Investor, "Gold Overtakes US Treasuries & 89% of Central Banks Expect Higher Gold Reserves" (World Gold Council 2026 survey) - CoinDesk, "Investors are throwing in the towel on the debasement trade, JPMorgan says", 28 May 2026 ============================================================================ ANALYSIS: Italy's borrowed calm URL: https://l0g.fr/en/analysis/italys-borrowed-calm/ Canonical French source: https://l0g.fr/posts/rachat-italien-btp-bund-trompe-l-oeil/ Date: 2026-07-26 (reviewed 2026-07-26) Topics: europe, bonds, italy, spreads, sovereign risk ---------------------------------------------------------------------------- The number has the force of a symbol. In January 2026, the yield gap between Italy's ten-year bond and its German counterpart closed at 64.6 basis points, its lowest level since 2008. Three years earlier, in September 2022, that same gap brushed 251 points, and markets were speculating on Rome's ability to fund itself without the shadow of the ECB. In between, Italy collected seven rating upgrades in a single year, moved back below France, and saw its deficit fall under the 3% line. The story of Italian fiscal redemption writes itself. It is not false. It is only half true, and that missing half changes everything. Let us start by giving Rome its due, because the improvement is real and it would be dishonest to deny it. Moody's raised Italy's sovereign rating from Baa3 to Baa2 on 21 November 2025, its first upgrade in twenty-three years, praising political stability and execution of the recovery plan. S&P had opened the ball in April, and seven upgrades in all punctuated the year. The public deficit fell to 2.98% of GDP in 2025, below the European threshold, paving the way for an early exit from the excessive deficit procedure. Italy even ran a primary surplus in 2024, the only G7 country to do so. And foreign capital came back: it held €1,038 billion of Italian debt in August 2025, some 33.7% of the total, up €121.6 billion over the year. Nothing cosmetic in that. The Meloni government, now one of the longest-lasting of the Republic, has delivered the rarest commodity in Italian politics: fiscal steadiness. But look at what moved The problem is not in these facts, it is in how they are read. A spread is a gap, therefore a subtraction, and a subtraction can tighten because the first term falls or because the second rises. In 2025, almost all the movement came from the second term. The yield on the ten-year Italian BTP went from 3.52% at end-2024 to 3.46% in November 2025, a mere six basis points lower. Over the same stretch, the German Bund's yield climbed 34 points, to 2.7%. The 40-point tightening of Italy's spread was therefore not a BTP rally: it was, more than 80% of it, the Bund rising. This is not a coincidence, it is a regime. We described elsewhere the end of Bund scarcity, that once-rationed asset, throttled by the debt brake and ECB purchases, now issued in floods to finance defence and infrastructure. When the German anchor stops being unfindable, it yields more, and everything measured against it tightens mechanically, without the periphery having done a thing. A good part of the European spread "normalisation" is in reality a normalisation of Germany. Italy has the merit of not having widened while the Bund rose, which in past cycles was far from guaranteed; it does not have the merit of having caused the convergence on its own. Slipping below France, an equivocal compliment The most spectacular proof of Italian redemption was the moment, in late 2025, when the Italian spread slipped below the French one. Rome was borrowing more cheaply than Paris relative to Berlin: an unprecedented inversion, instantly turned into a trophy. Here too the reading deserves to be turned around. If Italy slipped below France, it was at least as much because France disappointed as because Italy shone. Over 2025, the Italian spread tightened by 40 points while the French spread tightened by only 6, weighed down by a fiscal instability we analysed in our piece on French rates. The ranking flipped because the rival fell back, not only because the runner sped up. And the fragility surfaces the moment the wind turns. In the first quarter of 2026, the shock of the Iran war hit the BTP harder than the OAT, pushing the Italian spread some thirty points above its pre-war level and back above France. In the calm, Italy holds; at the first stress, it remains the first to be sold. The trompe-l'oeil dissolves precisely when it would be most needed. What the calm forgets Beneath the tranquil surface, the underlying numbers have not vanished. Italian public debt is still expected around 136% of GDP by 2028, and growth, 0.5% in 2025 and 0.7% in 2026, is too weak to erode that burden through the denominator. The deleveraging path is only due to resume after 2027, once the superbonus drag on revenues fades. Until then, Italy sails by sight between a recovery plan that is ending, an election in 2027 and an ECB backstop that is little discussed but works in silence. That backstop is the Transmission Protection Instrument, the anti-fragmentation tool created in 2022, which durably lowered the correlation between rate expectations and peripheral spreads. As long as it hovers, the redenomination risk, the spectre of a euro-area break-up that blew spreads apart in 2011, stays capped. But the instrument is conditional and discretionary: were the gap to settle durably at a high level and Rome-Brussels relations to sour over the fiscal rule, its conditionality would again become the heart of an already divisive debate within the Governing Council. Italy's calm rests in part on an implicit promise no one has yet been forced to test. Some investors are indeed starting to grow wary of it, as the government's political troubles mount. The other reading: resilience is earned Should Italy's spread be reduced to a mirage manufactured by Germany? That would go too far the other way, and honesty demands acknowledging what is solid in the current regime. That the convergence came mechanically from the Bund takes nothing away from a remarkable fact: when core yields rise, the historical rule is that peripheral spreads widen, because investors flee risk. This time, Italy held firm while the Bund climbed. Not widening in a bond bear market is itself a performance, and it signals that the political risk premium attached to the BTP has structurally fallen. The foundation of that resilience is anything but illusory. The primary surplus, the stability of a government among the longest-lasting of the Republic, the execution of the recovery plan and the return of €121 billion of foreign capital sketch a genuine improvement, not a stroke of luck. And the Transmission Protection Instrument, whatever one thinks of its untested character, has indeed removed from the market the most destructive tail risk, redenomination. One may fairly argue that Italy earned the right to enjoy the Bund's normalisation, where others would have squandered it. The truth, then, lies in between, and that is what makes the case interesting. The Italian spread at 65 points is at once a real compliment and a half misunderstanding. The real test has not yet taken place: it will come when the Bund's tailwind fades, or when the 2027 electoral deadline draws near. That day, we will learn whether Italy's calm was borrowed from German supply or earned in its own right. For now, Rome pockets a dividend it did not author alone, and pretends not to notice. --- Sources - Il Sole 24 Ore, "Seven rating promotions and spread fall: the BTp's year of grace" (seven upgrades, Moody's Baa2 on 21 November 2025, breakdown BTP 3.46% versus Bund 2.7%, Italian spread -40 bp and French -6 bp, deficit 2.98%, foreign holdings €1,038.4bn) - Amundi Research Center, "View on Italy and its government debt" (2024 primary surplus, only G7; debt ~136% by 2028; Iran shock on the BTP; role and conditionality of the TPI) - Reuters (via Investing.com), "Markets falling out of love with Italian debt as Meloni's problems mount" ============================================================================ ANALYSIS: Three thousand euros, no more URL: https://l0g.fr/en/analysis/three-thousand-euros-no-more/ Canonical French source: https://l0g.fr/posts/euro-numerique-plafond-detention-depots-bancaires/ Date: 2026-07-26 (reviewed 2026-07-26) Topics: europe, ecb, money, banks, payments ---------------------------------------------------------------------------- There is something paradoxical about designing a currency in the hope that no one holds too much of it. Yet that is the exercise the European Central Bank is running with the digital euro, its central bank money for the general public. The project cleared the final negotiation stage this summer, and the entire debate crystallises around one number: the holding limit, the maximum a private individual will be allowed to keep in their digital wallet. The ECB sees it between €500 and €3,000. This is not an engineering constraint. It is a dyke. And a dyke always betrays what it fears. The timing has its irony. At the very moment Brussels launches the savings and investments union to pull deposits out of banks and into the markets, Frankfurt is locking down its own creation so that these same deposits stay exactly where they are. Two European projects, one pool of savings, two rigorously opposite directions. The contradiction is only apparent, and what it reveals is precisely what makes it interesting: Europe wants to mobilise household savings, but it is terrified of moving them too fast. The cap is the message Everything in the design of the digital euro is built so that it circulates without ever accumulating. At the co-legislators' request, the ECB communicated a holding-limit range of between €500 and €3,000 per person. The top of the range is anything but arbitrary: €3,000 is the average net monthly income of a euro-area household. In other words, one is allowed to hold enough to live on for a month, not enough to save. The cap is only the first of the locks. The digital euro will pay no interest, so that it never competes with a savings account or a term deposit. It will be off-limits to companies as a holding, to cut off any large-scale hoarding. It will be distributed by the banks themselves, not by the ECB directly, so the intermediary stays in the loop. And it will come with a waterfall mechanism: above the cap, any excess is automatically swept back to the linked bank account, while a reverse waterfall reloads the wallet from that account when a payment requires it. Every brick of the architecture answers the same obsession: money should pass through, never sit still. This bundle of restrictions tells a story the press releases do not write. You do not throttle an instrument this hard if you expect it to change people's lives. You throttle it because you fear what it might trigger. The number that explains the dyke What Frankfurt fears has an order of magnitude, and it is massive. If every private individual in the euro area filled their cap to the maximum by moving the money from their current account, the shift would represent about 15% of retail deposits, roughly €1 trillion, the equivalent of the value of banknotes currently in circulation. A trillion-euro wall that could leave bank balance sheets to lodge, risk-free and without an intermediary, directly at the central bank. For a retail bank, that is so much less funding to turn into credit. For the system, it is the spectre of a slow, permanent run into the safest asset there is. A currency built not to please too much The international comparison finishes off exposing the intent. Where the Bank of England is considering a cap of £10,000 to £20,000 and Canada the equivalent of about €17,000, the euro area sets a limit three to five times lower. It is not that Europeans pay differently: actual cash held in wallets ranges from €46 in the Netherlands to €121 in Austria, so even €3,000 vastly exceeds everyday cash use. European moderation therefore reflects not payment caution but financial-stability caution, that is, the banks' fear for their deposit base. The flaw in the waterfall The sharpest paradox is that these locks could turn against the project's very purpose. The ECB presents the digital euro as a monetary anchor, the fixed point guaranteeing that a euro stays a euro whatever its form, the singleness of money that private stablecoins and foreign payment networks threaten to erode. But the waterfall lets a user keep a zero balance while retaining full payment functionality, the money being drawn on the fly from their bank account. Pushed to its logic, this convenience empties the anchor of its substance: what use is central bank money that no one ever holds? Frankfurt wants the digital euro to be everywhere and to weigh nowhere. The two wishes are hard to keep together. The debate, as a result, plays out at the exact point where financial stability meets monetary sovereignty, and it is anything but academic. The European Parliament settled its position in spring 2026, capping holdings and imposing a twenty-four-month rollout. The three-way negotiations between Parliament, Council and Commission opened on 13 July 2026, with holding limits and the compensation model for distributing banks as the very first items. The Irish presidency is aiming for a political agreement before year-end; the ECB, for its part, is preparing a pilot in the second half of 2027 and a possible first issuance in 2029. The other reading: the dyke is a door Reducing the digital euro to the banks' fear would be unfair, and several counterpoints deserve to be put. The first is that the dreaded disintermediation could be a regulatory fantasy. Who will want to keep €3,000 earning nothing in a wallet, when a current account offers the same payments and a savings account pays interest? Actual cash balances, a few dozen euros, suggest modest demand. The ECB itself argues that the digital euro is an opportunity for banks, which keep the customer relationship, collect distribution fees and gain a pan-European payment rail they do not control today. The second counterpoint is that the project's real driver may not be a deposit flight but strategic autonomy. Card payments in Europe run massively through two American networks, and dollar-backed stablecoins are laying claim to the role of default digital money. Against that double dependence, a European public currency is less a weapon against banks than an insurance policy on sovereignty, a natural extension of the tokenisation projects driven by the ECB. Seen that way, the cap is not a dyke against savings, it is a prudent dial, deliberately set low at the start to be raised once stability has been tested. Still, that dial says everything. A cap designed to be loosened later is an admission that the instrument is being launched without knowing how far to let it run. The real question is not whether the digital euro will see the light of day, it is now on the rails, but at what level this limit will durably settle, because that number will measure, very precisely, how much trust Europe places in its own two-tier monetary system. Three thousand euros today is the measure of the fear. What that number becomes will be the measure of confidence regained. --- Sources - European Central Bank, "Preparation phase of a digital euro, Closing report", October 2025 (holding-limit range, waterfall, timeline, possible issuance in 2029) - Bruegel, "On the digital euro holding limits" (€3,000 = average monthly income, 15% of deposits and €1 trillion, international comparison, the waterfall flaw, actual cash balances) - Freshfields, "Digital euro enters trilogues", July 2026 (trilogues opened on 13 July, holding limits and compensation model first) - PPC Land, "MEPs cap digital euro holdings and force 24-month rollout after 43-14 vote", June 2026 - European Central Bank, blog "Digital euro: an opportunity for banks", 27 March 2026 ============================================================================ ANALYSIS: Ten trillion asleep URL: https://l0g.fr/en/analysis/ten-trillion-asleep/ Canonical French source: https://l0g.fr/posts/dix-mille-milliards-en-dormance-siu/ Date: 2026-07-26 (reviewed 2026-07-26) Topics: europe, savings, markets, regulation, securitisation ---------------------------------------------------------------------------- The figure has become a mantra in Brussels: ten trillion euros. That is what European households leave sleeping in bank deposits, a reserve that exceeds the euro area's annual GDP and makes the Union the thriftiest continent in the developed world. The paradox sits in the next sentence: this same continent complains that it lacks the capital to finance its defence, its transition and its technology champions. The money is there, it does not circulate. The Savings and Investments Union, unveiled by the Commission in March 2025, is the latest attempt to resolve that contradiction. It is one year old, on a tight timetable, with a blind spot no one wants to name. One statistic captures the waste better than any speech. Every year, European households save about €1.4 trillion, against €800 billion for Americans, and yet close to €300 billion of that European saving leaves to finance markets outside the Union, mostly US assets. Europe exports its savings to the very economy it accuses of outrunning it, then re-imports the return as dividends paid to others. The project labelled SIU wants to close that leak. Whether an action plan can repair what three decades of unfinished integration left broken is the open question. Rich savings, poorly put to work The diagnosis is old and stubborn. European households are not short of money: they immobilise it in the least productive form there is. Union households keep 34% of their financial assets in deposits and cash, against 14% in the United States, and hold only about 17% of their wealth in listed securities, equities, bonds and funds, where Americans place 43%. The cost of that allocation gap runs into lost points of wealth: over a decade of rising markets, the cautious European saver watched their American counterpart grow richer while their own savings account eroded purchasing power. This conservatism is not just a matter of temperament. It reflects the absence of a simple, legible, tax-neutral channel from deposit to long-term investment. Facing twenty-seven tax regimes, as many national savings products and financial information rarely comparable from one country to the next, inertia is the rational choice. It is that inertia the Commission wants to turn into flow, at the precise moment the Union discovers the scale of its needs. Because demand for capital has exploded. The Draghi report of September 2024 put the additional annual investment effort at around €800 billion, close to 5% of the Union's GDP, to close the innovation gap, finance decarbonisation and reduce strategic dependencies. The former ECB president was clear on the source: most of it will have to come from the private sector, not from already stretched public budgets. In other words, there is a pool of dormant savings on one side, a wall of investment to finance on the other, and no plumbing that efficiently connects the two. The SIU is meant to be that plumbing. The rebranding of an old project The savings union is not a new idea: it is an old project that changed its name because the old name had not kept its promises. From 2015 to 2025, Brussels carried the Capital Markets Union, two successive action plans meant to build a single market for financing and wean European companies off their dependence on bank credit. The verdict, ten years on, fits in one word: fragmentation held. Markets stayed national, listings kept migrating to New York, and the share of savings invested in securities did not move. Hence the March 2025 relaunch, under a refocused label. The very name shifts the emphasis: it no longer speaks first of markets, it speaks of savings, that is, of the citizen. That semantic shift, presented as a new strategy on 19 March 2025, owes as much to the Letta and Draghi reports as to a political observation: selling financial integration on the promise of an abstract single market never mobilised anyone, whereas promising the saver a better return might create a social demand. The bet is clever. It changes nothing about the fact that the obstacles remain the same, as a European Parliament overview notes, stressing the persistence of the same barriers under a renewed wrapping. The strategy is organised around four workstreams: channelling citizens' savings, widening the supply of financing for companies, integrating and scaling market infrastructure, and converging supervision. The first three are technical and negotiable. The fourth is political and explosive, and we will come back to it. What Brussels put on the table The Commission deserves credit for one thing: it did not stop at the press release. In a little over a year, several concrete texts have been tabled, and it is that legislative material now under negotiation. The first floor, the one that governs the banking logic of the plan, is the revival of securitisation. The Commission presented its securitisation package on 17 June 2025, the first concrete initiative of the SIU, designed to ease the capital charges and administrative burden weighing on good-quality deals. The idea: let banks move loans off their balance sheets to sell them to investors, freeing up capacity to lend again. Securitisation is the instrument that built the depth of the American credit market, and the one Europe has distrusted since 2008, for reasons our guide on CLOs and leveraged loans sets out. The text has run its course: committee review in ECON in early 2026, around five hundred amendments tabled, Parliament's position finalised in spring ahead of the trilogue. It already divides: the centre-right groups see the missing tool, social democrats and greens smell a return of pre-crisis machinery, and part of the industry judges that the version under discussion does not go far enough to genuinely revive the market. The second floor aims at the saver directly. The Commission issued a recommendation on savings and investment accounts, simple and tax-advantaged wrappers inspired by the Swedish model, meant to give every household a legible entry point to the markets. To this is added a pensions component, with the review of the pan-European personal pension product and recommendations on auto-enrolment, a file on which the Council agreed its negotiating position in June 2026. The goal is to build, through the funded pension pillar, the base of long-term investors that Europe lacks. The third floor takes on the corporate framework and the plumbing. The Commission proposed on 18 March 2026 a 28th regime, an optional European legal status branded EU Inc for innovative companies785710), which could opt for a unified company law rather than juggling twenty-seven national codes. And above all, in December 2025, it tabled the package with the heaviest consequences, the one everything else depends on. The real lock: supervision, and the twenty-seven Here is the crux of the file, the one the press releases dress up. On 4 December 2025, the Commission published its integration and supervision package, which proposes to centralise part of supervision at European level and to extend the direct powers of ESMA, the European markets authority, over significant infrastructure and new participants. In plain terms: to make ESMA the sketch of a single markets supervisor, on the model of what the ECB has become for the large banks. That is where the mechanism jams, because a single market for capital requires single rules and a single referee, and Europe has neither. An investor lending to a company in another country faces a foreign insolvency law, a foreign tax code and a national supervisor that is not their own. These twenty-seven regimes are not technical details: they are the reason a German pension fund prefers US Treasuries, liquid and under a single law, to the bonds of a Portuguese SME. As long as defaulting in Lisbon, Milan or Warsaw does not obey the same rules, cross-border capital stays expensive and scarce. Yet these three locks fall under the competences member states defend most jealously. Taxation requires unanimity. Insolvency law touches the heart of national legal orders. And handing supervision to Brussels means dismantling national authorities that employ, that regulate their champions and that weigh in the domestic game. Unsurprisingly, the shift towards a European-level supervised model runs into resistance from national regulators and from the players who thrive in the current fragmented system. Every financial centre, from Frankfurt to Paris to Amsterdam, wants integration provided the centre of gravity sits at home. It is the calculation that sank the Capital Markets Union for ten years, and nothing suggests it has changed. The Commission knows it, which is why it has swapped persuasion for pressure. Its 30 April 2026 message, "from strategy to delivery: member states must now act", is a polite but firm reminder that Brussels has done its part and the blockage now sits in the capitals. Commissioner Albuquerque kept up the pressure all summer, continuing in July 2026 her dialogue with the academic community on implementation, and is aiming for agreement on the remaining texts by the end of 2026. The timetable is achievable. The political will of the Twenty-Seven remains the unknown. There is, finally, an objection from the left of the chamber that it would be dishonest to ignore. Pushing households to convert their deposits into market investments also transfers to them a risk they did not carry. A parliamentary question tabled in 2026 worries explicitly about the project's impact on public social security systems: by over-praising funded pensions, does one not weaken pay-as-you-go pensions? The cautious saver one wants to turn into an investor will also discover volatility. The line between financial empowerment and risk transfer is thin. The other reading: what if incrementalism is enough The dark scenario is easy to write: without single supervision or tax harmonisation, the SIU would be mere packaging, the third disappointment after two Capital Markets Union plans. That reading is solid, but it deserves its counterpoints, because it assumes there is no reform except through the institutional big bang. European history often says the opposite. First, the securitisation lever depends on no unanimity. It is a prudential rules adjustment, adoptable by majority, and its effect on banks' lending capacity is mechanical. If it revives even a fraction of Europe's long-atrophied securitised credit market, it will have circulated capital without any state ceding an inch of sovereignty. The same reasoning applies to the 28th regime: an optional status asks no one to abandon national law, it adds a parallel lane companies will take if it is better. Integration by opt-in bypasses the blockage instead of confronting it. Second, the €300 billion leak deserves perspective. That part of European savings finances US assets is not in itself a pathology: it is diversification, and a repatriated return remains income for the continent's saver. The problem is not that Europe invests abroad, it is that it does not offer at home vehicles deep and liquid enough to retain a share of that flow. Yet that depth is being built, and it is already being built: the expanding pool of common European debt, the end of Bund scarcity that widens the continent's bond base and the tokenisation projects driven by the ECB are manufacturing, without saying so, the deep reference asset that was missing. The SIU does not need to succeed at everything at once; it needs these bricks to converge. Third, incrementalism has a virtue the big bang lacks: it advances when political will is missing. The banking union was built through crisis, piece by piece, never completed, and works all the same. Europe will probably not build the savings union in one legislature nor by a treaty. It will build it, if it builds it, by sedimentation, a standardised savings account here, a unified company status there, a supervisor gaining one power at a time. It is not glorious. It may be the only method that works at twenty-seven. There remains the question that will decide everything, and to which neither Draghi nor the Commission can answer in the member states' stead: does Europe truly want a unified capital market, with what it implies of a common supervisor and champions that cease to be national? As long as the answer stays ambiguous, ten trillion euros will keep sleeping, and the best action plan in the world will not wake them. --- Sources - European Commission, "From strategy to delivery: Member States must now act on the SIU", 30 April 2026 - European Commission, Commissioner Albuquerque's dialogue with the academic community on the SIU, 15 July 2026 - European Commission, Draghi report on competitiveness (investment gap of about €800 billion a year) - Euronews, "EU Commission unveils plan to channel 10 trillion of citizens' savings", 19 March 2025 (€10tn in deposits, €1.4tn saved, €300bn exported) - Bruegel, "EU savers need a single-market place to invest" (34% in deposits versus 14%, 17% in securities versus 43%) - European Parliament, EPRS, "Savings and investments union: Overview and state of play", 19 November 2025 - DLA Piper, "EU Capital Markets Overhaul: European Commission Publishes Market Integration Package", December 2025 (centralised supervision, ESMA) - European Parliament, EPRS, "The 28th regime corporate legal framework" (proposal of 18 March 2026, EU Inc)785710) - Orrick, "European Parliament Finalises Securitisation Regulation and CRR Reform Proposals Ahead of Trilogue", May 2026 - Investment Company Institute, "EU Securitisation Framework Needs Stronger Reforms" - EU Perspectives, "EU capital markets in 2025: Savings and Investment Union takes shape", December 2025 (Council position on pensions, June 2026) - European Parliament, question O-000012/2026 on the SIU's impact on public social security systems ============================================================================ ANALYSIS: The cash that is not cash: the hidden liquidity buffer in bond funds URL: https://l0g.fr/en/analysis/cash-that-is-not-cash-bond-fund-liquidity-buffer/ Canonical French source: https://l0g.fr/posts/cash-pas-cash-coussin-liquidite-fonds-obligataires/ Date: 2026-07-25 (reviewed 2026-07-25) Topics: markets, funds, bonds, liquidity, systemic risk ---------------------------------------------------------------------------- When an investor asks an open-end bond fund for their money back, the fund does not necessarily sell a bond. It can first use cash, let a very short-term investment mature, reduce a repo position or redeem a money-market vehicle. Selling credit comes later. There is therefore a buffer between the redemption request and the bond sale. A Federal Reserve study published in May 2026 finally measures its composition. Its conclusion changes the question: the buffer is liquid, but very little of it is cash. This piece extends, without repeating, our analysis of high yield holding up while investment grade flees. Fund outflows show that an investor wants to be repaid. They do not show which asset the manager mobilised, or for how long the fund can avoid selling bonds. The Fed's answer is precise and limited. In its sample, the average buffer represents 4.7% of net assets and the median 3.4%. Yet in the aggregate series, cash and cash equivalents average only about 0.4% of assets. Most of the buffer comes from short-term investment vehicles and repo, two building blocks of the nonbank money market whose liquidity itself depends on market conditions. This finding does not prove that forced selling occurred in July 2026. The data end in the third quarter of 2025. It reveals something else: the daily liquidity of a bond fund is a funding chain, not a pile of banknotes. The exact perimeter: open-end mutual funds, not ETFs The FEDS Note published on 8 May 2026, by Erik Larsson, Ty Kawamura and Chaehee Shin, uses N-PORT and N-CEN filings submitted to the SEC. Its sample covers 369 US corporate bond mutual funds, observed from the fourth quarter of 2019 to the third quarter of 2025, with 5,458 fund-quarter observations. In the third quarter of 2025, these vehicles held $450.635bn in net assets. The authors select funds that are: - open-end vehicles registered on Form N-1A; - invested in portfolios with a weighted average maturity of at least three years; - invested in US-domiciled corporate bonds amounting to at least 55% of net assets. The perimeter explicitly excludes ETFs. It also excludes money market funds, which operate under a separate framework. The distinction matters: an ETF can manage flows through share creations and redemptions, sometimes in kind, using authorised participants. An open-end mutual fund redeems its investors directly under the procedures set out in its prospectus. The study therefore measures neither LQD nor HYG, nor the entire bond market. It describes one precise segment: long-term US mutual funds invested mainly in domestic corporate bonds. The ratio researchers had to reconstruct The SEC requires covered funds to report their portfolios every month on Form N-PORT. The form contains holdings, values, maturities, asset types, counterparties and several risk measures. Yet the most intuitive piece of information for this topic is missing from public filings: the liquidity category assigned to each position. Rule 22e-4 requires a fund to classify its positions at least monthly into four categories, from highly liquid to illiquid, taking into account conversion time, price impact and market depth. But N-PORT Item C.7, which carries that classification, remains confidential. The Fed researchers therefore built a measure that is neither a regulatory ratio nor an official threshold: the SLAR, or Short-Term Liquid Assets Ratio. Its numerator adds: - cash and cash equivalents; - Treasury bills maturing in 90 days or less; - US-domiciled repos maturing in 90 days or less; - US-domiciled STIVs. The denominator is the fund's net asset value. The formula is: SLAR = short-term liquid assets / fund net assets Form N-PORT defines a STIV category that includes a money market fund, liquidity pool or other cash-management vehicle. It is not a single legal wrapper. It is a reporting category that groups instruments designed to invest cash over short horizons. The reconstruction has one virtue: it examines what funds actually hold, not the liquidity implied by their name or stated strategy. It also has a limit: it measures a stock of assets deemed readily mobilisable, not the price at which each could be converted into cash during a crisis. A buffer near 5%, but only 0.4% in cash Across the full period, the average fund in the sample has a SLAR of 4.7%, with a median of 3.4%. The interquartile range is about 1.5% to 7.2%. The median fund therefore has a smaller buffer than the mean suggests, because the distribution is pulled upward by the most liquid vehicles. The aggregate composition is even more informative: - STIVs account for more than half of the buffer in most periods, equal to about 3.2% of net assets; - repo amounts to about 1.5% of assets; - cash and cash equivalents average only around 0.4%; - very short Treasury bills are only a minor component, with no exact figure published in the note's text. These orders of magnitude come from different statistics in the same study: a time average for cash, an asset-weighted aggregate series for the composition, and a value observed during most periods for STIVs. They must not be added as though they were the exact balance sheet of one fund on one date. Why a liquid asset is not cash A repo held by a fund is a cash loan secured by securities. At maturity, the counterparty repays the cash and receives its collateral back. A STIV is an interest in a vehicle that invests cash in short-term instruments. In both cases, the fund earns a return and retains strong liquidity under normal conditions. The qualification lies in the words "under normal conditions". Bank cash is already the settlement unit. A repo must mature, unwind or be transferred. A STIV interest must be redeemed by the vehicle that holds it. Their liquidity therefore depends on a second layer: collateral quality, money-market functioning, counterparty capacity, operational timing and market depth. The Fed does not say these instruments are about to break. It makes a more cautious implication: bond-fund liquidity may be shaped not only by investor redemptions, but also by market conditions in the nonbank money-market instruments the funds use. That dependence connects three compartments often analysed separately: 1. the bond fund, which promises daily redemption; 2. money market funds and liquidity pools held through STIVs; 3. the repo and collateral market, which turns a security into short-term funding. The buffer does not eliminate liquidity transformation. It temporarily moves it into assets whose conversion looks immediate for as long as their own market keeps functioning. What the manager actually sells The simplest mechanism would be a perfectly ordered queue: cash, then STIVs and repo, then liquid bonds, and finally hard-to-sell bonds. Academic research describes a more nuanced response. A study by Hao Jiang, Dan Li and Ashley Wang, published in the Journal of Financial and Quantitative Analysis in 2021, finds that corporate bond funds tend to reduce liquid assets to meet redemptions during calm conditions. When aggregate uncertainty rises, they sell liquid and illiquid assets in closer proportions to preserve the portfolio's liquidity profile. Sector-wide selling during high-uncertainty periods then creates price pressure followed by reversals, consistent with constrained-sale effects. Academic article and DOI. It would therefore be wrong to write that funds always sell their best bonds first. Managers choose between two risks: - consuming the buffer and leaving remaining investors with a less-liquid portfolio; - selling bonds as well, accepting transaction costs while trying to preserve a more stable portfolio structure. The choice depends on the scale of outflows, market liquidity, fund composition and the ability to rebuild the buffer quickly. SLAR measures the first line of defence, not the manager's entire strategy. A stock that falls after stress and is rebuilt later The Fed series shows a recurring pattern: stress episodes consume the buffer, then subsequent inflows allow it to be rebuilt. After the pandemic outbreak in the first quarter of 2020, the asset-weighted aggregate SLAR fell from 6.5% to 4.9%. The authors say the decline was likely an outcome of heavy redemptions, without presenting the attribution as a causal identification. The ratio then rebuilt to 5.8% in early 2021. After another trough in mid-2022, amid monetary tightening and bond-fund outflows, it returned to 5.5% by year-end. In the most recent part of the sample, it fell from 5.1% in the second quarter of 2025 to 4.3% in the third quarter, after redemptions associated with April volatility. The decline was 0.8 percentage point. The March 2020 precedent provides the market mechanism. Antonio Falato, Itay Goldstein and Ali Hortaçsu find that outflows were more severe in funds exposed to illiquid assets and fire-sale vulnerability. The Federal Reserve's corporate-bond backstop benefited the more fragile funds more strongly and helped reverse flows. Their study does not say such support will be repeated. It shows that in 2020, a backstop under the bond asset also stabilised fund liabilities. NBER Working Paper 27559, subsequently published in the Journal of Monetary Economics. What the regulator sees, and what the public does not The US framework contains several safeguards. Rule 22e-4 requires a liquidity-risk management programme, monthly asset classification, a general limit of 15% of net assets in illiquid investments and, for some funds, a fund-determined highly liquid investment minimum. The SEC states that breaching the illiquid-asset limit or remaining below that minimum triggers confidential notification. This framework does not impose one uniform minimum SLAR. The Fed notes that the highly liquid investment minimum does not apply in the same way to funds that primarily hold assets already classified as highly liquid, and the level is set by the fund. The information gap remains. The public can download N-PORT portfolios, but not the C.7 classifications used by the regulator. The authors had to infer them from asset type, domicile, maturity and reported value. In August 2024, the SEC adopted more frequent publication: one N-PORT report every month, filed within 30 days and made public within 60 days, instead of disclosing only the third month of each quarter. In April 2025, it delayed the effective date until 17 November 2027, with a compliance date of 18 May 2028 for fund groups below $1bn in net assets. Even after that reform, confidential fields will remain separate from public positions. More frequency is not full transparency, and a 60-day delay is not real time. Should the investor leaving pay the cost of exit? The buffer protects a fund from rushed sales, but consuming it can transfer costs to remaining investors. If an investor is redeemed at NAV before transaction costs and market impact are fully incorporated, the others may inherit a portfolio that is more expensive to liquidate. This first-mover advantage is already explained in our guide to reading money market funds. Its relevance here is regulatory: the Financial Stability Board says explicit and implicit redemption costs, including material market impact, should be borne by investors who redeem. Its revised 2023 recommendations call for anti-dilution tools and stress tests. IOSCO completed the framework in May 2025. Swing pricing, dual pricing and anti-dilution levies adjust the price paid by subscribing or redeeming investors. Quantity-based tools, redemption suspensions, gates, longer notice or settlement periods, side pockets and in-kind redemptions, instead limit the quantity or form of liquidity promised. Their availability depends on each jurisdiction's law. These tools do not make bonds easier to sell. They change who bears the cost and can slow the race to the exit. Some also carry a side effect: if investors anticipate a gate or suspension, they may try to redeem before it activates. IOSCO says so explicitly. What is known, inferred and unknown Observed fact: between 2019 and 2025, funds in the sample held a buffer near 5% on average, made mostly of STIVs and repo. The buffer fell after several outflow episodes and was rebuilt later. Academic result: funds adapt the mix of assets sold to the market regime. Under high uncertainty, bond sales can contribute to price pressure beyond the individual fund. Cautious inference: a simultaneous shock to fund redemptions and money-market liquidity would make the buffer less effective because two of its largest components depend on that same market plumbing. Unknown: the aggregate SLAR on 25 July 2026. The Fed note ends in the third quarter of 2025 and detailed liquidity classifications remain confidential. The public data available in this source set therefore do not show how much buffer has been consumed since then or which securities were sold. This separation prevents a structural vulnerability from being turned into a false immediate alarm. The useful dashboard SLAR should not be read in isolation. A high ratio can signal prudent management, but may also compensate for a less-liquid portfolio. A low ratio can be acceptable if assets are genuinely easy to sell and redemptions remain small. No 5% threshold mechanically separates safety from forced selling. Useful monitoring combines: - net flows and their speed relative to fund assets; - the share of cash, STIVs, repo and Treasury bills in N-PORT; - corporate-bond liquidity through TRACE volumes, bid-ask spreads and transaction costs; - credit spreads and dispersion across quality buckets; - any use of credit lines or interfund borrowing, reported in N-CEN; - changes in redemption policy and activation of anti-dilution tools. The right denominator is not only fund size. It is the speed at which investors can demand cash. A 4% buffer can be ample against daily outflows of a few basis points and insufficient against several days of heavy redemptions. Real coverage depends on a flow, not only a stock. The l0g view A bond fund's promise of liquidity does not rely directly on every bond being liquid. It first relies on a thin intermediary portfolio placed between the investor and the credit market. That portfolio does its job in normal times. STIVs pool cash management. Repo turns collateral into short-term cash. Treasury bills mature quickly. The fund can meet redemptions without immediately becoming a forced seller. But the architecture reveals a hidden dependency. The bond fund is also a user of the money market. When its investor asks for cash, another layer of the system must turn a short-term asset into the settlement unit. If a shock reaches redemptions, repo and money-market vehicles at the same time, the buffer stops being a passive reserve and becomes a transmission channel. Available data do not justify claiming that this shift occurred in July 2026. They allow the right question to be asked ahead of the next stress: how much liquidity remains before the fund sells what it intended to keep? --- Methodology - Main perimeter: 369 long-term US open-end mutual funds invested primarily in domestic corporate bonds, identified by the authors using N-PORT and N-CEN. - Period: 2019 Q4 to 2025 Q3. The data do not measure later conditions. - SLAR: cash and equivalents, Treasury bills maturing within 90 days, US repos maturing within 90 days and US STIVs, divided by net assets. - ETFs and money market funds are excluded from the main sample. - Composition figures retain the qualifications in the source. They are not the exact balance sheet of a particular fund. - Links between redemptions and SLAR declines are presented as observations and the authors' interpretations, not as certain causality. - Editorial cutoff: 25 July 2026. Primary sources - Federal Reserve Board, "Measuring Mutual Fund Liquidity with N-PORT", 8 May 2026. - SEC, Form N-PORT. - SEC, Investment Company Liquidity Risk Management Programs, Rule 22e-4. - SEC, delay to the N-PORT and N-CEN amendments, 16 April 2025. - Financial Stability Board, revised recommendations for open-ended funds, 20 December 2023. - IOSCO, "Guidance for Open-ended Funds", 26 May 2025. - Jiang, Li and Wang, "Dynamic Liquidity Management by Corporate Bond Mutual Funds", Journal of Financial and Quantitative Analysis, 2021. - Falato, Goldstein and Hortaçsu, "Financial Fragility in the COVID-19 Crisis", NBER Working Paper 27559, revised 2021. - Goldstein, Jiang and Ng, "Investor Flows and Fragility in Corporate Bond Funds", Journal of Financial Economics, 2017. ============================================================================ ANALYSIS: High yield holds up while investment grade flees: what the bond market is really measuring URL: https://l0g.fr/en/analysis/high-yield-holds-up-while-investment-grade-flees/ Canonical French source: https://l0g.fr/posts/le-high-yield-resiste-investment-grade-fuit/ Date: 2026-07-25 (reviewed 2026-07-25) Topics: markets, bonds, credit, rates, systemic risk ---------------------------------------------------------------------------- On 20 July 2026, US investment-grade bond funds suffered their largest daily outflow on record. Over the week, $7.1 billion left the category, while high-yield funds still received $534 million. Has the market suddenly decided it prefers fragile borrowers to strong companies? No. It is reminding investors that a bond carries at least two distinct risks: not being repaid, and being repaid too far in the future. A highly rated bond is not necessarily a defensive bond. When the risk-free rate rises sharply, long-dated, fixed-coupon, low-credit-risk debt can lose more than shorter, higher-paying speculative debt. July's paradox does not yet mean investors have stopped fearing defaults. It says that in the first stage of the shock, duration dominated credit quality. That distinction is central. It explains why investment grade can sell off before high yield, and helps identify the point at which a rates shock begins to contaminate credit. The observable fact According to LSEG Lipper data reported by Reuters on 24 July, US investment-grade bond funds recorded $7.1 billion of net outflows in the week ended 22 July 2026, a record. The 20 July session alone accounted for $8.2 billion of redemptions, also a record. This was not a uniform flight from all corporate debt. High-yield funds received $534 million, while leveraged-loan funds also attracted capital. In July, the iShares iBoxx Investment Grade Corporate Bond ETF, LQD, had lost 2.58%, compared with 0.93% for its high-yield counterpart, according to Reuters. The instinctive explanation would be risk appetite: investors selling strong credits to buy yield. It is incomplete. July's shock first came through the common foundation beneath all dollar bonds: Treasuries. l0g explains the mechanics in its guide to reading the US Treasury market. The 10-year Treasury yield rose from 4.55% on 17 July to 4.71% on 23 July, according to the Federal Reserve's H.15 series published by FRED. The 10-year inflation breakeven increased only from 2.25% to 2.28% between 20 and 23 July before falling back to 2.26% on 24 July, according to FRED. These observations do not allow an exact decomposition of the nominal-yield move. They nevertheless suggest that it was not simply a major unanchoring of long-term inflation expectations. The trap in the word "quality" LQD and HYG provide imperfect but useful representations of the two markets. As of 23 July 2026, BlackRock reported for LQD: - a 12.86-year weighted average maturity; - 7.78 years of effective duration; - a 5.63% average yield to maturity; - a 4.59% weighted average coupon; - an 84.55-basis-point option-adjusted spread, or OAS. For HYG, the same metrics were very different: - a 3.90-year weighted average maturity; - 3.05 years of effective duration; - a 7.39% average yield to maturity; - a 6.60% weighted average coupon; - a 271.59-basis-point OAS. Investment grade has better credit quality, but its main market vehicle locks investors into much more time risk. High yield carries more default risk, but its coupon is higher and its average maturity is much shorter. These are two credit portfolios, not two identical assets with different rating labels. What duration actually calculates For a small change in yield, the first-order relationship is: ΔP / P ≈ −D × Δy where D is effective duration and Δy is the change in yield in decimal form. FINRA explains that a one-percentage-point rise in rates implies, as a first approximation, a price decline equal to the duration. Under a uniform 50-basis-point shock, the calculation gives approximately: - LQD: −7.78 × 0.005 = −3.89%; - HYG: −3.05 × 0.005 = −1.53%. This calculation excludes coupon income, convexity, spread changes and ETF-specific flows. It should not be compared with observed performance to the nearest basis point. Its value lies elsewhere: the duration difference alone produces a gap of about 2.36 percentage points under the same yield shock. Credit quality protects against default. It does not protect against time. A long-dated investment-grade bond can be a highly aggressive rates position even when its issuer is unlikely to fail. Conversely, high yield contains several duration buffers: shorter maturities, higher coupons and more callable securities. Those features reduce the reaction to a pure rise in the risk-free rate. They clearly do not eliminate credit risk. Our guide to reading credit ratings separates assigned quality, rating migration and default risk. The signal is already not perfectly clean Saying that "it is all duration" would be as careless as declaring a credit crisis. On 23 July, the ICE BofA US Corporate Index OAS stood at 79 basis points, compared with 78 basis points from 20 to 22 July. The spread remained contained, so most of the increase in the yield demanded on investment grade still came from Treasuries. High yield showed more strain. Its aggregate OAS rose from 269 basis points on 20 July to 277 basis points on 23 July. More importantly, the CCC and lower segment moved from 977 to 991 basis points over the same period. The paradox therefore needs precise wording: Flows into high yield are still holding up better than flows into investment grade, but the market price of credit risk has already begun to rise beneath the surface. Flows and spreads do not measure the same thing. Flows describe the net allocations of a population of funds; spreads describe the premium demanded by the market on the bonds in an index. Inflows into high-yield funds can coexist with wider spreads if demand concentrates in shorter, better-rated or more liquid segments while CCC credits weaken. Starting yield also matters. At a 7.39% average yield to maturity for HYG, investors have a larger income cushion than in LQD. That cushion can absorb a moderate rate rise for a time. It disappears quickly if spreads widen by several hundred basis points or expected defaults increase. Three regimes, not one signal July's divergence becomes useful when read as a transmission sequence. 1. Rates shock The Treasury yield rises, long bonds fall and spreads remain relatively stable. LQD underperforms HYG because its duration is more than twice as high. Floating-rate leveraged loans can receive inflows because their coupons reset with rates. This was still the dominant regime as of 23 July. 2. Credit contagion High-yield spreads widen faster than investment-grade spreads. CCC underperforms BB, the distress ratio rises, issuance becomes scarcer and refinancing costs more. The market is no longer debating only the Fed rate path; it is beginning to revise expected losses. The observed HY and CCC OAS moves are consistent with an early phase of this second regime, but their magnitude remains insufficient to call a systemic break. The mechanics of CLOs and leveraged loans then become a second dashboard. 3. Liquidity stress In this scenario, outflows hit investment grade, high yield and leveraged loans at the same time. ETFs trade at more persistent discounts to net asset value, bid-ask spreads widen and TRACE volumes concentrate in the most liquid securities. Dealers protect their balance sheets; the displayed price gradually stops being a price at which a large position can actually trade. This mechanism is documented, but it is not being observed on that scale in July. During the March 2020 stress, the Federal Reserve found sharply higher corporate-bond transaction costs, large fund redemptions and less dealer capacity to absorb sales. That is the useful precedent for defining the regime, not evidence that it is already active today. In this regime, the difference between duration and credit becomes secondary. Investors sell what they can sell, not necessarily what they want to sell. The useful dashboard No single series can identify the transition. Monitoring has to combine several families of signals. Signal Duration shock Credit shock Liquidity stress 10-year Treasury rises quickly can remain high movement may become disorderly Investment-grade OAS contained widens widens with volatility High-yield OAS contained or moderate accelerates accelerates sharply CCC versus BB limited gap strong CCC underperformance prices can become discontinuous Fund flows mostly long IG outflows rising HY outflows simultaneous outflows Leveraged loans resilient begin to weaken outflows and lower liquidity ETF versus NAV small gap occasional discounts persistent discounts Primary market open but expensive issuance postponed window closed LQD/HYG in isolation is therefore not the right thermometer. The ratio mixes duration, quality, coupons, sector composition and index construction. It becomes useful when compared with spreads, rating buckets, flows and the primary market. l0g's guide to reading credit spreads explains the construction and limitations of OAS. What would invalidate this reading Four developments would weaken the "duration first" hypothesis: 1. a rapid and persistent widening of high-yield OAS without an equivalent fall in Treasury yields; 2. marked CCC underperformance and a rising distress ratio; 3. simultaneous net outflows from high-yield and leveraged-loan funds; 4. a closed primary market for issuers that could still refinance only weeks earlier. Conversely, a stabilisation of the 10-year yield, investment-grade spreads near current levels and normalising flows would confirm that July was primarily a repricing of long rates. There is one final construction caveat. Lipper flows cover fund categories; LQD and HYG are two specific ETFs; ICE BofA OAS series describe different index universes. Comparing them does not create a perfectly controlled experiment. The exercise is not designed to attribute every dollar of flows to one variable. It is designed to distinguish the dominant mechanisms using observable instruments. The l0g view The bond market has not decided that speculative debt is safer than investment grade. It has reminded investors that a strong credit can be a poor refuge when it is locked into long duration as the risk-free rate rises. July's paradox is therefore less a reversal of the credit hierarchy than a temporary change in the hierarchy of risks. First the rate, then the spread, finally liquidity. As of 23 July, the first stage still dominated: the US 10-year yield had risen 16 basis points since 17 July, while investment-grade OAS had barely moved. But aggregate high yield and especially CCC had begun to widen. The signal was not yet a crisis signal. It was already no longer perfectly clean. The useful question is not why high yield is "winning". It is how long its lower duration risk can mask its rising credit risk. --- Methodology - Fund-flow data: LSEG Lipper, week ended 22 July 2026, reported by Reuters. - LQD and HYG characteristics: iShares / BlackRock, data as of 23 July 2026. - Treasury yield: Federal Reserve Board H.15 via FRED. - Inflation breakeven: Federal Reserve Bank of St. Louis via FRED. - Spreads: ICE BofA indices published via FRED. Observations cited are daily, not seasonally adjusted. - Price sensitivity: first-order approximation ΔP/P ≈ −duration × Δyield. It is neither a performance forecast nor investment advice. - Data cut-off: 24 July 2026. Primary sources - Reuters, "US investment-grade bond funds see $7 billion record weekly outflows", 24 July 2026. - iShares / BlackRock, LQD: iBoxx $ Investment Grade Corporate Bond ETF, data as of 23 July 2026. - iShares / BlackRock, HYG: iBoxx $ High Yield Corporate Bond ETF, data as of 23 July 2026. - Federal Reserve Board, H.15: 10-year Treasury yield, via FRED. - FRED, 10-year breakeven inflation rate. - ICE BofA US Corporate Index OAS, via FRED. - ICE BofA US High Yield Index OAS, via FRED. - ICE BofA CCC & Lower US High Yield Index OAS, via FRED. - FINRA, "Brush Up on Bonds: Interest Rate Changes and Duration", 19 September 2024. - Federal Reserve Board, "The Corporate Bond Market Crises and the Government Response", 7 October 2020. ============================================================================ ANALYSIS: The ghost kilowatt: who pays for the grid if the data center never arrives? URL: https://l0g.fr/en/analysis/the-ghost-kilowatt/ Canonical French source: https://l0g.fr/posts/kilowatt-fantome-reseau-data-center/ Date: 2026-07-25 (reviewed 2026-07-25) Topics: ai, data centers, electricity, infrastructure, risk, regulation, us policy ---------------------------------------------------------------------------- A data center says it will need 1,000 megawatts. The utility adds the load to its forecast, reserves capacity, reinforces lines and prepares new generation. Then the project is delayed, downsized, moved or abandoned. The electricity is never consumed, but part of the grid has already been planned or built. That is the ghost kilowatt: announced demand that changes investment decisions before becoming actual consumption. The phrase is an analogy, not a regulatory category. The risk itself is documented. The Federal Energy Regulatory Commission, or FERC, warns that speculative requests and applications filed with several grids can be counted more than once, distort forecasts and send bad investment signals. The Department of Energy explicitly identifies stranded-asset risk when infrastructure built for a large load is underused. The question is therefore not only whether data centers will lift electricity demand. It is who guarantees the bill before they consume. A pledge is not a tariff On 23 July 2026, the White House expanded its Ratepayer Protection Pledge. Its principle is straightforward: data centers should pay for the generation, delivery and grid upgrades they cause, even if they ultimately do not use the power they reserved. The official page asks signatories to negotiate separate rate structures and pay for promised capacity whether they use it or not. That announcement establishes a political doctrine. It does not, by itself, settle a utility bill. The page describes the rate structures as voluntarily negotiated. Protection for other ratepayers depends on less glamorous documents: a tariff approved by a state commission, an interconnection contract, a cost-recovery agreement, a parent guarantee, collateral and the rules applied after cancellation. Reuters reported scepticism about the voluntary nature of the pledge the next day. That concern is justified on one precise point: a national promise becomes enforceable only when it is written into the relevant tariff and contract. How a line or power plant enters the bill A regulated utility does not charge only for electrons consumed. It also recovers operating expenses, depreciation, taxes and an allowed return on its rate base. A Department of Energy baseline report summarises the mechanism: revenue requirement = operating expenses + depreciation + taxes + rate of return × rate base If a substation or line is built for a load that disappears, there are three possible outcomes. 1. The customer still pays through a minimum-payment commitment, an exit fee or callable collateral. 2. The utility and its shareholders absorb some or all of the loss if the regulator refuses recovery in rates. 3. The cost enters the rate base or network charges and is spread across other customers. The third outcome is the transfer new large-load tariffs seek to prevent. But writing “pay if you cancel” is not enough. The contract must cover the right assets, for the right period, with the right legal entity. The clearest filter is in Ohio AEP Ohio's experience shows why utilities want to separate a serious project from an opportunistic reservation. According to the utility's 13 February 2026 update, it had received more than 30,000 MW of expressions of interest or requests before the new tariff took effect. 13,022.7 MW paid to enter the formal study process and 5,642 MW then signed legally binding contracts backed by collateral. These three numbers are not a cancellation rate: the stages, dates and perimeters are not identical. They nevertheless show the distance between stated demand, demand mature enough to fund a study and load backed by a legal commitment. Grid forecasts must stop treating those three levels of maturity as a single certainty. The result is informative, but the source is the utility itself. The figures should therefore be read as its reported pipeline, not as an independent assessment of the tariff's effectiveness. The Public Utilities Commission of Ohio approved the mechanism in July 2025 and described protection against underused investment as its purpose. Virginia and Wisconsin: charging for the reservation Virginia, the largest US data-center market, created a separate class for new very large loads. The State Corporation Commission fact sheet provides, from 1 January 2027, at least a fourteen-year commitment for affected new customers. Their minimum monthly payment must cover 85% of reserved transmission and distribution costs, even if consumption is lower. Where credit is insufficient, collateral can reach 60% of minimum payments over the contract term. In Wisconsin, the Public Service Commission approved a regime in April 2026 for loads of at least 100 MW. Its official release requires a minimum fifteen-year term, removes an option that would have reserved only 75% of capacity and requires very large customers to pay 100% of the costs allocated to them. These regimes are not identical and their percentages are not directly comparable. They concern different components, thresholds and rate structures. Their common logic is nevertheless clear: charge for reserved capacity, not only energy consumed. The contract can still miss the wrong bill The best counterexample comes from a FERC decision concerning an agreement between ComEd and Aligned Data Centers. The accepted agreement requires the customer to pay transmission charges even if its project is delayed or cancelled, or to pay a termination fee. It looks exactly like the intended protection. Yet Commissioner Judy Chang identifies two limits in her FERC concurrence. The agreement does not identify the specific upgrades it secures. And some network costs could be rolled into formula rates paid by all customers. If the assets are large, a bilateral contract can therefore coexist with higher charges for others. The case gives the right editorial and regulatory test: paying something after cancellation does not prove the customer pays everything it caused. In June 2026, FERC launched proceedings aimed at the large regional grids. Commissioner David Rosner said cost-recovery agreements should prevent a data center that never appears from leaving households with the bill. He also called for more transparency on speculative requests, physical site control and duplicate applications. In his official remarks, he describes these outcomes as the intended effect of the reforms. That is not yet evidence that every final tariff will achieve them. The risk changes address A minimum-payment commitment and collateral do not eliminate risk. They move it from the collective pool of ratepayers to the customer's credit quality. Protection is robust when: - the signing entity is solvent or backed by a strong parent guarantee; - collateral remains sufficient when project cost rises; - the payment term matches the recovery period of the assets; - exit fees cover equipment that cannot be reassigned; - the contract follows the project through a sale, restructuring or developer change; - the regulator separately identifies costs directly caused by the large load. It is fragile when a thinly capitalised vehicle signs instead of the group, the deposit is capped too low, some upgrades are diluted into general transmission charges or the grid relies on projects still duplicated across several queues. This shift into credit extends the risks already examined in the debt financing AI and the residual value guarantee. The difference matters: the potentially stranded asset is not only a GPU or a privately owned building. It is regulated infrastructure whose cost can enter a public utility bill. Evidence boundary The sources establish four facts. 1. Regulators regard speculative, duplicate or insufficiently mature requests as a forecasting and cost problem. 2. Several states have created real contractual protections: minimum terms, billing for reserved capacity, collateral and exit fees. 3. AEP Ohio's pipeline narrows sharply as demand has to pay for a study and then sign a contract. 4. FERC acknowledges that a pay-after-cancellation contract may not cover every network reinforcement rolled into general rates. They do not support a figure for a national bill already shifted to households. No harmonised public dataset yet links, project by project, announced load, committed assets, guarantees received, cancellation and final recovery. Claiming an aggregate amount would fabricate the missing data. Falsifiability The hypothesis of a material ghost-kilowatt risk would weaken if regulators consistently published: - a reconciliation of megawatts requested, studied, contracted and actually energised; - the incremental cost of each asset and the financial security backing it; - fees actually recovered after delay or cancellation; - the absence of residual costs rolled into other customers' rates; - verifiable reassignment of equipment initially built for an abandoned project. Conversely, cancellations with fees below non-reassignable cost, or upgrades explicitly rolled into collective rates, would confirm the mechanism. The White House pledge has therefore stated the right rule. State commissions and FERC now have to publish evidence of execution. The meaningful indicator will not be the number of signatures under a promise. It will be the dollars of infrastructure made unrecoverable by load that never materialised, and the identity of whoever ultimately paid them. Sources 1. White House, Ratepayer Protection Pledge, accessed 25 July 2026. 2. White House, release announcing the pledge expansion, 23 July 2026. 3. FERC, Commissioner David Rosner's remarks on large loads, 18 June 2026. 4. FERC, Commissioner Judy Chang's concurrence on the ComEd-Aligned agreement, 26 February 2026. 5. FERC, Commissioner See's remarks on cost recovery, 18 June 2026. 6. FERC, NYISO order, 195 FERC ¶ 61,216, 18 June 2026. 7. Department of Energy, Electricity Rate Designs for Large Loads, 15 October 2025. 8. Department of Energy, Electricity Distribution System Baseline Report, 2016. 9. AEP Ohio, data-center request update, 13 February 2026. 10. Public Utilities Commission of Ohio, approval of the data-center tariff, 9 July 2025. 11. Virginia State Corporation Commission, Data Center Initiatives, February 2026. 12. Public Service Commission of Wisconsin, approval of large-load tariffs, 24 April 2026. 13. Reuters, scepticism around the pledge, 24 July 2026. ============================================================================ ANALYSIS: When the barrel becomes a margin call: the hidden liquidity bill of the oil shock URL: https://l0g.fr/en/analysis/when-the-barrel-becomes-a-margin-call/ Canonical French source: https://l0g.fr/posts/baril-appel-de-marge-liquidite-choc-petrolier/ Date: 2026-07-24 (reviewed 2026-07-24) Topics: oil, liquidity, derivatives, systemic risk, banks, macro ---------------------------------------------------------------------------- Oil is back at $100, but the shock's first financial bill appears neither in the CPI nor in import accounts. It arrives in cash, sometimes by the next day, at the producers, commodity traders, refiners and airlines that hedge their prices in futures markets. A hedge can protect their future earnings while draining their cash today. The risk is not theoretical: in 2022 it forced European energy firms to reduce their hedges and mobilised bank balance sheets. Nothing, however, establishes that a comparable crisis is already under way in July 2026. The task is precisely to separate the documented mechanism, the observable signals and what public data cannot yet tell us. On 24 July, Reuters reported that Brent had moved through $100 the previous day for the first time since May, as the market again worried about Middle Eastern flows. Our analysis of the Fed's barrel trap covers the macroeconomic bill. Another one arrives faster and is less visible: the liquidity need created by derivatives. A profitable hedge can run short of cash Consider a producer due to sell one million barrels in a few weeks. To lock in the price, it sells Brent futures. That short position loses value if oil rises, but the physical crude the company will deliver becomes more valuable at the same time. At maturity, the two legs should largely offset one another. This is a hedge, not necessarily a bearish bet. The calendar breaks that symmetry. ICE specifies that one Brent contract covers 1,000 barrels and that every open position is marked to market daily. The gain on physical crude becomes cash only after sale and settlement. The futures loss produces variation margin at the pace of the market. The European Central Bank notes that variation margin must be paid in cash, while initial margin can also be posted in high-quality liquid securities. The following example is a teaching simulation, not the exposure of a real company. One million barrels correspond to 1,000 ICE contracts. If the price rises by $10 per barrel, the economic value of the physical inventory increases by $10 million, but the short futures position also loses $10 million. The hedged result can remain close to zero while the cash need reaches $10 million before the cargo is paid for. This timing difference explains why a solvent participant can come under pressure. It connects commodity markets to the plumbing described in our analysis of repo and collateral: in both cases, owning a valuable asset is not enough. The right form of liquidity must be available in the right place and at the time imposed by the market infrastructure. Margin protects the counterparty, not the treasury The central counterparty interposes itself between buyers and sellers. Initial margin covers a potential loss during the time needed to close a defaulting member's position. Variation margin resets the current exposure to zero as prices move. This system reduces the risk that an unpaid loss spreads from one counterparty to another. It does not eliminate risk. It turns risk into a liquidity requirement. The Financial Stability Board summarised the tension in its December 2024 final recommendations: margin and collateral protect against counterparty risk, but can amplify liquidity demand when they rise unexpectedly across a large part of the market. The FSB therefore calls for contingency funding plans, stress tests, and reserves of cash or immediately available liquid assets. ICE's matrix published on 24 July 2026 provides a current reference point. For the September 2026 Brent future, it indicated initial margin of $15,217 for a long position and $11,776 for a short position. ICE explicitly warns that these are indicative amounts for a single position: actual incremental margin depends on portfolio size, direction and composition and may be substantially reduced by offsets. A clearing member can also add its own surcharge to the clearing house requirement. The figures therefore cannot estimate a trader's net bill, but they show that the initial deposit comes on top of daily variation. 2022, the measured precedent The full-scale test came from European gas and power after Russia's invasion of Ukraine. Its mechanism is not identical to oil in July 2026, but it is documented with unusual precision. According to the Bank of England, TTF prices reached ten times their average over the previous decade. In the first half of 2022, average daily variation-margin calls rose to more than sixteen times their level in the calm 2019-2020 period. Higher initial margin reduced leverage from more than five times in September 2021 to less than two times in March 2022. Traders that had sold futures to hedge physical gas not yet sold had to meet calls within a day. Some cut their hedges to find cash, and open interest in the main TTF contracts fell by around 20%. The United Kingdom created a loan-guarantee scheme for energy firms unable to finance extraordinary calls. It was ultimately not drawn, but its existence identifies the risk the authorities sought to contain. The ECB reaches the same diagnosis using EMIR and AnaCredit data. By mid-2022, initial margins on commodity portfolios had approximately doubled from late 2021. Credit lines granted by euro-area banks to power producers rose from about €3 billion to more than €6 billion between March and April 2022. From the trader to the bank balance sheet When internal cash is insufficient, the bank becomes both lender and gateway to the clearing house. This dual function concentrates risk. At the end of August 2022, four banks were directing around 85% of exchange-traded energy-commodity positions to central counterparties, measured by gross notional value, according to the ECB. A quarter of the energy firms in its sample used the same set of banks for credit and derivatives clearing. The 85% figure requires caution. The ECB notes that gross notional inflates intermediation chains and is not a flawless measure of economic risk. It nonetheless reveals a narrow passage: if a client fails to pay margin, the clearing member still owes the clearing house. The bank may therefore fund a client whose clearing risk and, in some bilateral contracts, counterparty risk it already carries. Another route is to move the hedge over the counter. In 2022 the ECB observed a decline in futures and greater use of non-centrally cleared swaps among some European traders. The client saves immediate margin, but the system exchanges transparency and collateral for more bilateral risk. It is a precise example of credit risk migrating beyond the regulatory gaze: the constraint disappears from one screen, not from the balance sheet. July 2026, what is established Three elements are observable on 24 July. First, Reuters recorded Brent's return to $100 amid greater risk around two shipping passages. Second, ICE marks its Brent contracts daily and publishes indicative initial margins for the nearby contract. Third, the official precedents show that an energy shock can turn hedges into cash demand and bank credit very quickly. The conclusion stops there. The public data reviewed do not show a wave of oil margin calls in 2026, extraordinary drawings on bank facilities or a forced contraction in hedging comparable to TTF in 2022. The CFTC's COT describes positions and open interest with a lag of several days, but not margin calls, portfolio offsets, clearing-member surcharges or private credit facilities. The ICE matrix describes risk parameters, not the liquidity available to clients. That limit does not weaken the analysis. It prevents a plausible channel from being turned into an imaginary crisis. A repeat of 2022 is not a given The opposing case is strong. The 2022 European gas shock was more violent than the oil move observed in July 2026. The ECB itself noted that oil prices moved far less than TTF. Brent has a deep global market, integrated participants able to offset part of their exposures and portfolios in which diversification can reduce margins. Market infrastructures and treasurers have also learned. Since 2022, the FSB has formalised eight recommendations on margin-call preparedness: governance, liquidity-risk tolerance, funding plans, extreme but plausible scenarios, liquid assets and collateral organisation. Publication does not prove uniform implementation, but it makes the assumption of a completely unchanged system less defensible. Finally, an integrated producer benefits directly from the higher value of the oil it extracts, whereas a refiner, airline or distributor has neither the same physical exposure nor the same hedge. Treating “energy firms” as one balance sheet would erase precisely the differences that determine who pays margin and who receives it. The breaking points The case for liquidity stress would become more credible if several signals converged: 1. a further increase in ICE margins beyond the indicative level of 24 July; 2. a sharp fall in open interest alongside a reduction in commercial hedging; 3. unusual drawings or extensions of bank facilities by traders and producers; 4. a move from cleared futures into less-collateralised bilateral contracts; 5. clearing-member surcharges or the creation of public liquidity guarantees. Conversely, a sustained decline in Brent, stable margins, resilient open interest and no emergency facilities would refute the systemic margin-call scenario. The relevant dashboard is therefore not a magic oil-price threshold but the combination of price, volatility, margin, hedging and bank funding. A $100 barrel is a market signal. It becomes a financial risk only when a daily loss must be funded before the physical gain can be collected. The hidden liquidity bill sits in that interval, outside the Brent chart. Sources 1. Reuters, “Take Five: A $100 question”, 24 July 2026: Brent at $100 and risks around Middle Eastern shipping passages. 2. Intercontinental Exchange, Brent future specification, 1,000-barrel contract, daily mark-to-market and the role of ICE Clear Europe, accessed 24 July 2026. 3. Intercontinental Exchange, IRM 2 Margin Rates, Brent Crude Futures, matrix dated 24 July 2026. Amounts are indicative and depend on the actual portfolio. 4. European Central Bank, “Financial stability risks from energy derivatives markets”, Financial Stability Review, November 2022. 5. Bank of England, Nathanaël Benjamin, “Late call: preparing for liquidity stresses”, 18 July 2024. 6. Financial Stability Board, “The Financial Stability Aspects of Commodities Markets”, 20 February 2023. 7. Financial Stability Board, “Liquidity Preparedness for Margin and Collateral Calls: Final report”, 10 December 2024. Further reading: how to read the oil market, how to read the CFTC COT report, our analysis of supply chains after Hormuz, ghost tankers and the cost of waiting and the mechanics of repo and collateral. ============================================================================ ANALYSIS: Intelligence on the cheap: China's open-source AI strategy against the capex bubble URL: https://l0g.fr/en/analysis/china-open-source-ai-strategy-vs-capex-bubble/ Canonical French source: https://l0g.fr/posts/intelligence-bradee-open-source-chinois-capex-ia/ Date: 2026-07-23 (reviewed 2026-07-23) Topics: ai, china, debt, systemic risk, macro, markets ---------------------------------------------------------------------------- On 17 July 2026, China's Moonshot released Kimi K3, a 2.8-trillion-parameter model billed as the world's largest open-source model, with weights set to open by month's end. In the same run, Alibaba's Qwen family passed a billion cumulative downloads on Hugging Face, faster than any other model lineage, and DeepSeek keeps shipping its models under a permissive licence. Seen from Silicon Valley, this is a technology rivalry. Seen from l0g, it is something else: an economic weapon. By making frontier-class models free, China is collapsing the price of intelligence at the precise moment US giants commit hundreds of billions of dollars whose repayment assumes that intelligence stays expensive. The analysis that follows does not judge the models' quality, it reads the price shock they cause. The strategy: commoditise the model layer China's dominance in open-source models is no accident of timing, it is a choice. Alibaba has released more than a hundred models under the Apache 2.0 licence, DeepSeek published its own under an MIT licence together with a paper describing its training method, and Moonshot is now opening its largest model. Giving away the model layer means destroying its market value for everyone, oneself included, for the sole purpose of denying competitors their pricing power. This mechanism is familiar to us. It is the exported deflation we described in the world's factory selling off, transposed from the solar panel to the language model: flood the world with free capacity to suffocate the other side's margin. Where Chinese industrial overcapacity drives down the price of goods, open-source generosity drives down the price of inference. In both cases, China exports a price drop its rivals absorb. The target: pricing power The model layer was supposed to be the moat, the scarce asset that justified the margins. It is becoming a commodity. A Chinese open-source model such as DeepSeek charges for its output inference around $0.28 per million tokens, against some thirty dollars for a leading US frontier model, a ratio near a hundred to one. On reasoning models, the gap stays an order of magnitude. The market rule is merciless: when someone offers comparable quality at one-hundredth of the price, the margin evaporates. The point is not that closed US models are technically outdated, they are not necessarily. The point is that their edge no longer monetises at the token level. If raw intelligence trends toward free, the rent must lodge elsewhere, in the application, distribution, proprietary data. Yet it is precisely the token level that was meant to repay the infrastructure. The point of impact: $725 billion of capex Here it is, where the Chinese strategy meets the American balance sheet. The four largest hyperscalers, Amazon, Alphabet, Meta and Microsoft, plan to commit about $725 billion of capital expenditure in 2026, up 77% year on year; CreditSights puts the top five, Oracle included, between $700 and $900 billion. Nearly all of it goes into AI infrastructure: processor clusters, proprietary silicon, data centres. And a growing share of that bill is debt-financed, a mechanism we dissected in the debt behind AI, circular financing and the residual-value guarantee of infrastructure credit. The return on these $725 billion assumes one thing: that the model layer keeps enough pricing power to yield the margin that will repay the debt. China's open-source strategy attacks that assumption directly. If the price of inference keeps trending toward zero, the gap between committed capex and the revenue it generates, already scrutinised by markets that are repricing, becomes a solvency problem, not just a profitability one. It is the fragility we flagged in the AI boom and financial fragility and in the bubble within the bubble. The geopolitical lever To the financial dimension is added a power dimension, just as concrete. You cannot sanction a weights file once it has been downloaded. Where US export controls on chips aim to slow China, the diffusion of open-source models bypasses the symmetric lever: it installs the Chinese stack as the default among developers in the Global South, in universities and administrations that have neither the budget nor the access to closed US APIs. Setting the free standard means capturing the ecosystem and future dependence, a long-game logic that revenue tables alone do not capture. The opposite reading Fairness demands setting out the counter-argument, because it is serious and could overturn the conclusion. The first point is the Jevons paradox: cheaper intelligence widens the market, and a fall in the token price can blow up usage volumes, hence demand for compute, hence justify the capex rather than doom it. On this reading, value does not disappear, it migrates toward compute, that is toward chipmakers and inference operators, and toward the application layer that US labs monetise with enterprises. The second point is that open source carries hidden costs, compliance, security, no support, that keep many large accounts on closed providers. The third, finally, is that Chinese generosity is not pure strategic altruism: constrained by chip restrictions, unable to easily monetise closed models globally, China makes open source a rational second-best as much as a weapon. These objections shift the question without cancelling it. Even if the capex is justified by volumes, it changes beneficiary: the silicon maker and the inference operator collect, while the closed lab that financed its moat on credit sees its margin thesis crumble. The deflation of intelligence is real; it does not destroy value, it redistributes it, and that redistribution does not follow the map of the debt. Where the value goes The moat shifts from the model to the layer above. That observation, banal on the surface, carries a heavy financial consequence: whoever financed the model layer's margin with debt has tied its repayment to an asset whose price trends toward zero. The likely winners are compute and the application layer; the relative losers, the closed labs whose token rent evaporates and the creditors who backed them. China's open-source strategy is therefore not an episode of the model wars, it is a risk factor for the heaviest capex cycle in the sector's history. The signals to watch fit in a few lines. The gap between the hyperscalers' capex and revenue, which markets are starting to punish at every earnings call. The pace of Chinese releases, of which Kimi K3 is only the latest. The adoption of open-source models by large enterprises, the only arbiter of the migration. And the token price, the most direct thermometer of the margin left to defend. Reading AI through the lens of debt means seeing that the real question is not which model wins, but who repays when intelligence no longer sells. --- Data and sources: MIT Technology Review, the future of Chinese open-source AI (Kimi K3, Qwen, DeepSeek); large-model API price comparison, 2026 (token price collapse, DeepSeek vs frontier models); 2026 hyperscaler capital expenditure, about $725 billion and CreditSights estimates. Kimi K3's claimed performance figures are vendor announcements, without independent validation at this stage; inference prices and capex guidance move, the levels cited are those of mid-2026. To go further: our analysis of the world's factory and Chinese deflation; our AI-and-debt cluster, with the debt behind AI, circular financing, the residual-value guarantee, the AI boom and financial fragility and the bubble within the bubble; and our critical look at AI productivity gains. ============================================================================ ANALYSIS: The Fed trapped by the barrel: the data before the 29 July FOMC URL: https://l0g.fr/en/analysis/fed-trapped-by-the-barrel-data-before-july-fomc/ Canonical French source: https://l0g.fr/posts/fed-piege-du-baril-donnees-avant-fomc-juillet/ Date: 2026-07-23 (reviewed 2026-07-23) Topics: fed, inflation, oil, rates, macro, markets ---------------------------------------------------------------------------- A central bank always decides while looking in two directions at once. On 23 July 2026, Brent crossed back above $100, up 6.4% on the session, on the twelfth consecutive day of US strikes on Iran and after fresh tanker attacks. Six days before its 28-29 July meeting, the Federal Reserve nonetheless holds a snapshot pointing the other way: the latest price index, for June, shows inflation cooling. The Fed is looking at a rearview mirror that is calming while the windshield catches fire. What follows is not a forecast on its decision, but a reading of the data that boxes it in, and they say something counter-intuitive: the barrel does not push the Fed to raise rates, it removes its option to cut them. The rearview: inflation cooling Start with the hard data, the only thing that counts as fact. June's consumer price index, published by the Bureau of Labor Statistics, shows a rise of 3.5% over the year and, above all, a 0.4% fall on the month, the largest monthly drop since April 2020. The driver of that decline is energy, whose index fell 5.7% in June. Core, excluding food and energy, comes in at 2.6% year on year, the heart of inflation staying contained. Our guide to reading the CPI sets out why this split between headline and core is decisive. This figure tells of a disinflation under way, extending the sequence we tracked in our analysis of the 2026 inflation risk. On June's basis alone, a central bank with a dual mandate would have arguments to ease its stance, all the more as energy was pulling the whole down. The trouble is that this snapshot predates the oil surge, and a central bank does not drive by looking only in the rearview. The windshield: the barrel back above $100 The surge is recent and sharp. Brent crossed $100 on 23 July, WTI climbing toward $91, on the combination of a twelfth day of US strikes on Iran and tanker attacks off Saudi Arabia. This is a supply shock, exogenous to the US economy, and therein lies the whole difficulty: the rise comes not from overheating demand that higher rates would cool, but from a geopolitical risk premium on the barrel, which we have documented for months in our coverage of the Iran war and its economic fallout and in our guide to the oil market. The arithmetic effect is delayed but mechanical. Gasoline was already up 26.7% year on year in June; the late-July push will read in the July price index, published in mid-August, that is after the meeting. The Fed thus decides on a figure the barrel is in the process of making stale, without yet holding the measure of the ongoing shock. The windshield shows what the rearview ignores. Why the Fed is boxed in The trap lies in the nature of the shock. Monetary policy acts on demand, not on the supply of oil. Raising rates does not lower the barrel; it would only add a brake to an economy the energy shock is already slowing by eating into purchasing power. Conversely, cutting rates just as crude soars would mean easing as an inflation push builds, at the risk of un-anchoring expectations. Caught between these two dead ends, the Fed has only one workable option left, waiting. The market has grasped it. The CME's FedWatch tool put the probability of a hold at the 29 July meeting at 83.4% as of 21 July. The federal funds range has been unchanged at 3.50-3.75% since December 2025, per the Federal Reserve Bank of New York, which puts the effective rate at 3.63%. Against 3.5% inflation, the real rate is near zero: policy is neither clearly restrictive nor accommodative. This will be the second meeting chaired by Kevin Warsh, whose stance we analysed at his first appointment in June. His room for manoeuvre has narrowed a notch with every dollar added to the barrel. The barrel is already in long rates The bond market, for its part, is not waiting for the meeting. The ten-year Treasury yield stood at 4.67% on 23 July, per market data, with a 36-basis-point slope on the two-to-ten-year. The crude surge feeds inflation expectations and the term premium, that extra yield demanded to hold long debt when the price outlook clouds. On top of that comes the liquidity constraint we described in our analysis of the drained reverse repo cushion: the Treasury is issuing heavily, and the buyer book is tightening at the same moment. The long end therefore already prices part of the shock the policy rate cannot neutralise. Our guide to the Treasuries market gives the reading grid. The lesson of 2011 Caution requires setting out the counter-argument, because it argues precisely for waiting. Core at 2.6% remains contained, and if the Iranian escalation recedes, the oil premium can deflate as fast as it rose, making the shock transitory. Holding rates rather than reacting in haste is, on this reading, the wise decision and not the mark of paralysis. The precedent exists, and it is instructive: in 2011, the European Central Bank raised rates in the middle of an oil shock, before having to reverse course a few months later, an error we recalled in our analysis of the 2011 remake against the barrel. Tightening against supply-driven inflation is fighting a fire with the wrong extinguisher. The serious objections therefore bear less on the July decision, a widely expected hold, than on what follows. If the barrel stays high, the July index, then August's, will climb through energy, and the Fed will have to hold against a headline inflation that reheats without being able to address its cause. Its communication will then matter as much as its rates: telling a transitory oil bump from a durable un-anchoring of expectations will be the most delicate exercise of the coming months. The data to read The list of markers is short and keeps every forecast at arm's length. The July price index, in mid-August, will tell the scale of the barrel's pass-through into consumer prices. The tone of the 29 July statement, more than the decision itself, will reveal how the Fed ranks the rearview and the windshield. Market inflation expectations, readable in breakevens and in the ten-year term premium, will measure whether the shock is still judged transitory. And the path of Brent, hostage to Iran, will decide the size of the problem. The Federal Reserve is not facing an excess of demand that a turn of the screw would correct, but an oil price no rate brings down. Its best option is also the most uncomfortable, to do nothing and explain it. The real stake on 29 July is not the level of rates, known in advance, but the reading a central bank makes of a shock it endures without being able to cure. Reading the data means seeing that constraint before it imposes itself on the message. --- Data and primary sources: Bureau of Labor Statistics, June 2026 consumer price index (headline 3.5%, core 2.6%, energy and gasoline); Federal Reserve Bank of New York, effective federal funds rate as of 20 July 2026; ten-year Treasury yield as of 23 July 2026; CME FedWatch, probabilities for the 29 July meeting. Analysis and press: CNBC, Brent surges above $100 on the Iranian escalation (23 July 2026). To go further: our pieces on Warsh's first FOMC, the 2026 inflation risk, the drained liquidity cushion, the economic earthquake of the Iran war and the 2011 remake against the oil shock; our guides to reading the CPI, reading the PCE, reading the oil market and reading the Treasuries market. Oil prices, yields and probabilities move continuously; the levels cited are those of 20 to 23 July 2026, the inflation reading being June's, published on 14 July. ============================================================================ ANALYSIS: The CLARITY Act under the scalpel: the new architecture of US crypto URL: https://l0g.fr/en/analysis/clarity-act-bill-text-sectoral-impact-analysis/ Canonical French source: https://l0g.fr/posts/clarity-act-texte-analyse-impact-sectoriel/ Date: 2026-07-23 (reviewed 2026-08-06) Topics: clarity act, crypto, regulation, us politics, stablecoins ---------------------------------------------------------------------------- Update, 7 August 2026. The analysis below concerns a text, not a regime in force. H.R. 3633 remains officially Passed House, not Passed Senate. After the 10 August to 11 September break, the calendar provides fourteen scheduled days, from 14 September through 2 October, before the election period. Procedure and scenarios are set out in CLARITY Act: the window before the midterms. The text exists, and it is massive. The CLARITY Act as reported in the Senate, referenced H.R. 3633 RS, runs to 594 pages and carries two official titles: the "Digital Asset Market Clarity Act of 2025" and the "Anti-CBDC Surveillance State Act." This is not a light-touch deregulation, it is the construction of an entire regulatory regime, with registration, mandatory disclosures, custody and anti-money-laundering compliance. Its keystone is one mechanism: a maturity test that decides whether an asset falls to the securities regulator, the SEC, or the commodities regulator, the CFTC. We read the text to draw a sector-by-sector impact analysis, each claim tied to its section. One preliminary point: this reported version does not contain the ethics clause targeting public officials whose negotiation made July's headlines, a separate matter we covered in our piece on the August countdown. The maturity test, keystone The whole scheme rests on a shift of jurisdiction. A token sold to the public through an investment contract remains, at issuance, an ancillary asset under the SEC. But once the blockchain that carries it becomes a "mature blockchain system," the asset qualifies as a "digital commodity" and moves under the CFTC's spot-market jurisdiction. The definition, in Section 104, is deceptively plain: a mature blockchain system is one "not controlled by any person or group of persons under common control." Decentralization thus becomes the legal criterion that reassigns the regulator. The passage is not automatic. The text creates, in Section 205, a new Section 42 of the Securities Exchange Act organising a certification: the issuer of a digital asset files with the SEC an attestation that the blockchain is mature. This is the text's most sensitive point, a self-certification of decentralization, framed by agency oversight but initiated by the issuer. To stop each actor inventing its own definition, Section 105 requires the SEC and the CFTC to define jointly, by rulemaking, the key terms, including "mature blockchain system," "decentralized governance system" and the decisive notion of "unilateral authority." The law sets the principle; the real boundary will be drawn by two years of joint rulemaking. Issuers: a framed launch ramp For anyone creating a token, the text opens what years of regulation by enforcement had closed: a legal financing path. Section 202 adds to the Securities Act an issuance exemption letting an issuer raise, on its ancillary asset, up to $50 million of gross proceeds per calendar year for a period not exceeding four years, the amount adjusted annually to the Consumer Price Index, or 10% of the total value of outstanding ancillary assets. A retail safeguard accompanies the opening: after an exempt transaction, a purchaser may not hold more than 10% of outstanding units. Insiders, for their part, are reined in. Section 104 counts as "affiliated persons" any holder of at least 5% of the units, any founder and any officer, and Section 204 caps their sales: over any twelve-month period, they may cover only a band between 5% and 20% of the units acquired. Section 411 adds a notification duty for "control persons" of a system certified mature before any sale. The message to the sector is twofold: financing a token on US soil becomes possible again within a framework, but founders' exit is slowed and watched. Exchanges and intermediaries: the federal licence, and its price The text's operational core is the registration of intermediaries. Title IV creates at the CFTC a full regime for "digital commodity exchanges," "brokers" and "dealers," with custody by qualified custodians (Section 405), product certification for trading and registration of associated persons. Section 106 provides expedited registration and provisional status, so existing players are not frozen while the regime deploys. Title III mirrors this with the SEC's residual role over ancillary assets and the intermediaries that touch them. The trade-off is real. Section 413 requires the CFTC to issue, within 360 days, rules on identifying and resolving conflicts of interest "among and across registered entities," naming vertically integrated market structures. This is the lesson of the FTX collapse written into law: a platform that combines exchange, brokerage, custody and market-making will have to wall off those functions. For compliant US exchanges, the text offers a long-awaited federal licence; in exchange it imposes a compliance architecture they must fund and document. Title IV as a whole takes effect 270 days after enactment (Section 414). DeFi and developers: the broad exclusion This is the most industry-friendly part, and the most debated. Section 309 inserts a new Section 15H into the Securities Exchange Act, and Section 409 does the same on the CFTC side: a person is not subject to these laws merely for compiling, relaying, sequencing or validating transactions, running a node or an oracle, providing bandwidth, offering an interface to read blockchain data, or developing and publishing software. Validators, node operators, oracle providers, front-ends and developers therefore fall explicitly outside the scope. Section 109 protects non-controlling developers in the same spirit. The reach is immense for decentralized finance, and that is also where the criticism concentrates. The exclusion hinges on the control test: a truly decentralized protocol escapes regulation, but an actor keeping unilateral authority stays in the net. And it is exactly the definition of that "unilateral authority" and of maturity that the text hands to joint rulemaking. Until it is written, the line between the protected developer and the regulated operator remains a grey zone, and self-certified decentralization feeds the fear of circumvention. Stablecoins and banks: docking to GENIUS On stablecoins, the CLARITY Act reinvents nothing, it docks. Section 104 defines the "permitted payment stablecoin" by reference to the GENIUS Act, and Section 301 states that such a token may be brokered, traded and custodied by a broker, dealer, alternative trading system or exchange, the SEC having jurisdiction only over the transaction, not the token as a security. The payment stablecoin is thereby confirmed outside the securities scope, extending the logic we described in our analysis of stablecoins as the marginal buyer of US Treasuries and of the bill deluge. Two provisions complete the institutional picture. Section 310 expressly authorises the custody of digital assets by banking institutions, a green light prudential regulators had long withheld. And Section 110 extends the application of the Bank Secrecy Act, the anti-money-laundering law, to digital-asset intermediaries: regulatory clarity comes paired with compliance duties identical to those of traditional finance. US crypto gains a seat in the banking system, at the price of entering its rules. The hidden title: banning a CBDC The text carries a second title, often overlooked in commentary: the "Anti-CBDC Surveillance State Act." Its Title VI bars the Federal Reserve from issuing a central bank digital currency, directly (Section 602) and indirectly through a financial intermediary (Section 603). This is a major monetary-policy decision lodged in a market-structure law: the United States renounces, by statute, the tool China and others are building, and effectively leaves the digital-dollar field to private stablecoins regulated by the GENIUS Act. The choice is coherent with the whole, but no less consequential for the monetary architecture. The blind spots Rigour requires naming the weaknesses, because they will decide the text's real reach. The first is self-certification of maturity: handing the issuer the initiative to declare its blockchain decentralized, even under agency oversight, creates a hazard critics judge exploitable. The second is the power shift toward the CFTC, a smaller, less-resourced regulator than the SEC, to supervise a spot market it had never had to watch; the text provides dedicated resources, whose sufficiency remains to be shown. The third is the definition calendar: until the SEC and CFTC have jointly written what "unilateral authority" and maturity mean, the promised clarity remains, on paper, a promise. The fourth is political: the ethics clause targeting public officials is absent from this version, and its possible addition on the floor remains the knot we described elsewhere. Three dated scenarios What follows is scenario, not data. In the fast-adoption path, the text is passed before the 7 August recess, with or without a floor ethics amendment, reconciled with the House version, then enacted; Title IV then takes effect about 270 days later, and the first conflict-of-interest rulemaking falls at 360 days. In the blockage path, the text slips past August and runs into the midterm campaign, pushing the workable deadline back by years. In the adoption-without-clarity path, the law passes but bogs down in two years of contested joint rulemaking, so that "clarity" stays theoretical until the definitions settle. Our piece on the Senate endgame tracks that calendar day by day. The net effect At the end of the reading, the net effect comes down to a few lines. The text ends regulation by enforcement by giving issuers a bounded financing path and exchanges a federal licence, it carves a broad exclusion for DeFi and developers, it cements stablecoins as the rails of the private digital dollar by referring them to the GENIUS Act, it brings crypto custody into banks and anti-money-laundering compliance into crypto, and it bans a central bank digital currency by statute. The regulatory centre of gravity slides from the SEC to the CFTC, through a decentralization test on which everything will hinge. The real unknown is no longer whether a framework exists, but how it is scored: the promised clarity will play out over two years of rulemaking, and in how two agencies draw, together, the boundary of control. For the European framework alongside, our guide to MiCA offers the point of comparison. --- Primary source: the official text of the CLARITY Act, H.R. 3633, as reported in the Senate (BILLS-119hr3633rs), 594 pages, and its XML version; bill page and status on Congress.gov. Every claim in this analysis points to a precise section of the text: definitions and maturity test (Sec. 104, 205), joint rulemaking (Sec. 105), issuance exemption (Sec. 202), insider sales (Sec. 204, 411), intermediary registration and custody (Sec. 106, 405), conflicts of interest (Sec. 413), deferred effect (Sec. 414), DeFi exclusion (Sec. 309 and 409, Section 15H), stablecoins (Sec. 301), bank custody (Sec. 310), anti-money-laundering (Sec. 110), CBDC ban (Title VI, Sec. 602 to 604). To situate: our analyses of the Senate endgame and its scenarios, of Trump as the first obstacle to his own law and of the CLARITY Act, a primer; our guides mapping the GENIUS Act and decoding MiCA; our pieces on stablecoins as marginal buyers and the bill deluge. This analysis covers the version reported in the Senate as of 1 June 2026; a text amended on the floor could alter certain provisions, starting with the ethics clause absent from this version. ============================================================================ ANALYSIS: Who buys the bill deluge? The US Treasury's new marginal buyer URL: https://l0g.fr/en/analysis/who-buys-the-bill-deluge-us-treasury-marginal-buyer/ Canonical French source: https://l0g.fr/posts/qui-achete-le-deluge-de-bons-tresor-acheteur-marginal/ Date: 2026-07-23 (reviewed 2026-07-23) Topics: us treasury, bonds, stablecoins, money market funds, macro, markets ---------------------------------------------------------------------------- A debt issuance always reads at two ends: how much the state sells, and who buys. We described the first end in our analysis of the drained liquidity cushion: the US Treasury is borrowing $671 billion net this quarter, largely through short-term bills, while the reverse repo facility, the window where money funds parked their cash and out of which the money came to buy bills, has fallen to $1.2 billion. The shock absorber is gone. That leaves the second end, less commented on yet decisive: who now absorbs the deluge? The answer traces an order book that has changed composition, and it is that composition, more than the volume, that deserves attention. The first buyer, and its limit The largest taker of Treasury bills remains, by far, money market funds. Their assets reached a record of about $7,890 billion in the week of 15 July 2026, per the Investment Company Institute, with government-securities funds making up the bulk. Such a mass constantly seeks short paper to remunerate, and Treasury bills are its natural vehicle, as our guide to money market funds details. The limit lies in one word: appetite varies. Those same funds saw their assets fall by nearly $60 billion in the week of 15 July, and their preference swings continuously between bills and repo, depending on the relative yield of the two. When repo pays better, cash leaves bills for the pledged loan. The first buyer is therefore vast, but it is not a captive acquirer: it compares, and it can turn away from the Treasury's window without notice. The new entrant the law compels Alongside this fickle giant, a buyer of an unprecedented kind has entered: stablecoin issuers. The market for dollar-pegged tokens passed $300 billion in 2026, and the GENIUS Act, in force since 2025, requires domestic issuers to hold full backing in high-quality liquid assets, essentially Treasury bills and overnight repo, with no ability to pass the yield to holders. In practice, every additional dollar of stablecoin mechanically translates into a purchase of short-term government debt. The amounts are becoming significant. Tether reports more than $141 billion of Treasury exposure, ranking it among the twenty largest holders of US government debt worldwide; Circle parks about 80% of USDC reserves in a BlackRock-managed government money fund that buys only bills and repo. This is a buyer structurally insensitive to price, bound by law to buy whatever the yield, exactly the mechanic we described in our piece on stablecoins as the marginal buyer of US Treasuries. Still small against the $7,890 billion of money funds, but growing, and above all captive, where the rest of the book is not. Foreign demand is changing in nature The third reservoir, foreign holders, whose total stood above $9,500 billion in early 2026, up 6% year on year. But composition matters more than the total. The Treasury's TIC data put Japan first at end-2025 with $1,186 billion, followed by the United Kingdom at $863 billion and China at $684 billion. The ranking tells of a shift: the United Kingdom has moved ahead of China, not because London saves more, but because the hedge funds domiciled on its market book the basis trade there, that highly leveraged bet on the gap between the cash and futures price of Treasuries. The nuance is first-order. Demand of official origin, like China's central bank, is stable and price-insensitive; demand from leveraged funds is opportunistic and can unwind at once. As China trims its pool and British leverage grows, the quality of the foreign book deteriorates: more demand, but more fragile. Japan, the top holder, faces its own bond-market strains, described in our analysis of Japanese fiscal dominance, which erode its capacity to absorb others' debt. The banks and the great absentee Two players remain, one constrained, the other gone. Banks could absorb more government debt, but their capacity runs into the leverage ratio, the SLR, which weighs Treasuries and reserves without risk-weighting. The US reform of late 2025 eased that constraint to give room back to Treasury-market intermediation, a project we tracked in our piece on the Basel III rollback by US regulators. The easing helps, it does not turn banks into an unlimited buyer. The great absentee is the Federal Reserve. Long the primary buyer through its purchase programmes, it is now shrinking its balance sheet and letting part of its portfolio roll off. The taker that, in 2020 and 2021, mechanically absorbed new debt has left the table. The order book must therefore do without both its largest historical buyer and its liquidity cushion at once. Demand is not lacking Caution requires setting out the opposite reading, because it is robust. At no point has the deluge struggled to clear: auctions cover, money funds have nearly $7,900 billion to churn, foreign demand is at a record $9,500 billion, and stablecoins add a new and growing buyer. The Treasury bill remains the most sought-after cash equivalent on the planet, and it always finds a taker. Talk of a buyers' strike would be a misreading. The serious objections bear not on the market's capacity to absorb, but on the price and the soundness of the marginal taker. A book where captive demand shrinks and leveraged demand rises absorbs the debt, but at a higher yield and with a nearer breaking point. The basis trade in particular already showed in March 2020 that a forced unwind could freeze the most liquid market in the world, a risk we x-rayed in our study of the Fed's record exposure to the basis trade. The price, not the taker What follows is a signal to watch, not a forecast. The first is the yield demanded at bill auctions, whose level and tail will tell at what price the book clears. The second is the share of leverage in foreign demand, readable in TIC data and in funds' futures positions. The third is the trajectory of stablecoin assets, the only buying flow that grows by regulatory construction. The fourth is money funds' behaviour between bills and repo, which decides at the margin the appetite for Treasury paper. The bill deluge will find a taker; that was never the real question. It is at what price, and on whose shoulders. A book that loses its central buyer, the Fed, and its cushion, reverse repo, must lean on a mix of fickle giants, law-bound buyers and leveraged funds. The debt will clear, but the yield it demands and the fragility of those who carry it are, from now on, the real barometer. --- Data and primary sources: Investment Company Institute, money market fund assets; US Treasury, foreign holders of debt (TIC data) and TIC system; Congressional Research Service, ownership of Treasury debt (RS22331). The stablecoin backing framework stems from the GENIUS Act, whose application we detail in our guide mapping the GENIUS Act regulators. To go further: our analysis of the drained liquidity cushion for the supply side; our guides reading TIC data, reading money market funds and reading the Treasuries market; our pieces on stablecoins as marginal buyers, the leveraged basis trade and the Basel III rollback. Money fund and foreign debt holdings move weekly or monthly; the levels cited are those of mid-2026 releases, with TIC country data dating from December 2025. Stablecoin holding figures come from their attestations and the specialist press, cited as such. ============================================================================ ANALYSIS: The cushion is gone: RRP drained, the Treasury reloads, reserves in the front line URL: https://l0g.fr/en/analysis/fed-rrp-drained-tga-rebuild-reserves-liquidity-2026/ Canonical French source: https://l0g.fr/posts/tga-rrp-vide-reserves-siphon-liquidite-2026/ Date: 2026-07-22 (reviewed 2026-07-22) Topics: fed, liquidity, repo, us treasury, systemic risk, macro ---------------------------------------------------------------------------- There is a place in the financial system where monetary tightening is not voted, it is operated. This quarter, a plumbing figure deserves more attention than the bets on the Federal Reserve's next meeting: the overnight reverse repo facility, the window where money market funds parked their spare cash, fell to $1.2 billion in the week of 15 July 2026, against a peak near $2.5 trillion at end-2022. It is, to all intents, empty. At the same moment, the US Treasury is rebuilding its cash account by borrowing hundreds of billions. The cushion that absorbed every liquidity drain has melted; the next one will therefore come from elsewhere, from bank reserves. What follows is neither a crash nor a prophecy, but the reading of a mechanism. The cushion that absorbed everything Let us take the mechanics from the right end. The Federal Reserve's balance sheet has a liability side shared by three main items: bank reserves, the Treasury's cash account, and the reverse repo facility. The latter, the RRP, played the role of shock absorber from 2021 to 2024: when the Fed had injected too much liquidity, money market funds deposited their surplus there, remunerated overnight. The RRP thus peaked around $2.5 trillion at end-2022, according to the Federal Reserve Bank of New York's series. Since then it has drained, and that looks like a good thing: this money came back out to buy the mountains of Treasury bills issued since 2023, without draining bank reserves. The cushion did its shock-absorbing job. The Fed's weekly release, the H.4.1, records it unambiguously: as of 15 July 2026, the facility, down to $1.2 billion, absorbs nothing more. The problem begins where the shock absorber stops. The Treasury reloads its account For the demand for liquidity does not fade. After the debt-ceiling resolution, the Treasury must rebuild its cash, the general account it holds at the Fed. The 15 July release puts it around $796 billion, and the Treasury's financing estimate targets a cash balance of $950 billion at end-September, funded by $671 billion of net marketable borrowing in the third quarter alone, largely through short-term bills. The third quarter is traditionally the heaviest of the year, the Treasury having borrowed $1,058 billion at the same point in 2025. Every dollar that enters the Treasury's account leaves the system: it exits a bank account, hence reserves, to sit on the Fed's liability side. As long as the RRP served as the source, this transfer was painless, money funds paying for the new bills with cash withdrawn from the window. That channel is now closed. Rebuilding the Treasury's account will therefore run, at the margin, against reserves. The drain changes target Bank reserves stand at about $3,143 billion, a level still called abundant. But abundance is not infinite, and the Fed is simultaneously shrinking its balance sheet, whose securities portfolio remains above $6,400 billion. Two forces thus pull reserves down at once: quantitative tightening, which removes assets, and the Treasury's cash rebuild, which shifts cash to the Fed. As long as the RRP supplied the counterpart, reserves were protected. That protection is gone. The stakes lie in a notion the Fed handles carefully: the lowest level of reserves compatible with a calm money market. No one knows this threshold in advance; it is discovered by crossing it, as in September 2019, when a Treasury cash rebuild and a tax-date spike sent repo rates jumping to 10% intraday. The 2026 mechanics are of the same nature, slower: the cushion is drained, reserves fall, and the market will end up showing where the floor sits. The signal in short rates That signal already reads, quietly, in overnight rates. SOFR, the benchmark secured funding rate, stood at 3.57% on 20 July 2026, below the effective federal funds rate of 3.63% and within the 3.50 to 3.75% target range, per the Federal Reserve Bank of New York. Nothing alarming in the average. But SOFR's 99th percentile, the rate paid by the most pressed borrowers, came in at 3.66%, above the fed funds rate and near the top of the range. On daily volume above $3 trillion, this overshooting tail is the first symptom of collateral that is starting to cost more to finance. Why 2019 does not mechanically repeat Fairness requires saying why the 2019 comparison has its limits, because they are real. The Fed learned from the episode. It now has a standing repo facility, which did not exist in 2019: a bank short of cash can borrow there against Treasuries at a ceiling rate, mechanically capping repo spikes. Reserves, above $3 trillion, are moreover still judged abundant, far from the $1.5 trillion of 2019 relative to a smaller economy. And the Fed has signalled its intent to slow then stop shrinking its balance sheet as reserves approach the ample level, precisely to avoid replaying the scenario. The serious objections therefore bear not on an imminent accident, but on the margin. No one knows the reserve floor, and the Treasury's cash rebuild can reveal it sooner than expected, especially around tax dates and quarter-ends, when bank balance sheets contract for regulatory reasons we described for regional banks and their liquidity ratio. The standing facility caps the spikes; it does not say at what reserve level the market turns nervous. The sensors to watch The list is short and reads in real time. SOFR relative to the fed funds rate, whose spread and month-end spikes will tell whether collateral is tightening. Use of the standing repo facility: the day banks tap it in volume, scarcity is here. The level of reserves in the weekly H.4.1, set against the ample threshold the Fed is trying not to cross. The issuance calendar of the next quarterly refunding, in early August, for the bill share of the $671 billion. And the pace of balance-sheet shrinkage, whose halt would signal the Fed judges the margin too thin. The most effective tightening of this cycle may not have been decided in a meeting. It is operated, mechanically, in the plumbing: a drained cushion, a Treasury reloading, reserves falling. Reading this plumbing means seeing the liquidity constraint coming before it shows up in the rates everyone watches. Our guides to Treasury liquidity and the repo market give the full grid. --- Data and primary sources: Federal Reserve, H.4.1 release (week of 15 July 2026) for the reverse repo facility, the Treasury account, reserves and the securities portfolio; Federal Reserve Bank of New York, SOFR and effective federal funds rate as of 20 July 2026; US Treasury, marketable financing estimate for net borrowing and the third-quarter cash target. To go further: our guides reading Treasury liquidity (TGA, RRP), reading the repo market and SOFR, reading the Fed balance sheet (H.4.1) and reading money market funds; our analyses of the repo liquidity factory and collateral as the keystone. Balance-sheet and rate levels move weekly; the values cited are those of the 15 and 20 July 2026 releases. The Treasury borrowing projections are its own quarterly estimates. ============================================================================ ANALYSIS: The tightening that didn't happen: the ECB's credit survey, and the private-credit crack URL: https://l0g.fr/en/analysis/eurozone-credit-tightening-ecb-survey-private-credit-crack/ Canonical French source: https://l0g.fr/posts/resserrement-credit-bce-enquete-fissure-credit-prive/ Date: 2026-07-22 (reviewed 2026-07-22) Topics: credit, banks, ecb, systemic risk, private credit, macro ---------------------------------------------------------------------------- A central-bank survey reads on two levels: the headline and the figures. On 21 July 2026, the European Central Bank published its second-quarter bank lending survey, and the easy headline is that euro-area banks keep tightening. That is true in the strict sense, but misleading. For firms, the tightening is far milder than announced, and loan demand is picking up. The real strain shifts to two zones the survey lights up less directly: households, and the parallel channel that bank regulation itself helped inflate, private credit, where the Financial Stability Board documents fragilities piling up. Two credit channels, one business cycle, trajectories pulling apart. The survey's message The figures first, since they belie the general impression. On lending to firms, euro-area banks report a net tightening of their credit standards of 7% in the second quarter, against 10% the previous quarter and above all against the 19% they themselves expected in April. The feared hardening thus materialised at only a third of its expected scale. Better still, firms' loan demand came in up a net 3%, where banks had projected a 10% decline. For the segment that weighs most on investment, the survey traces a normalisation, not a stranglehold. The picture worsens, by contrast, for households. Standards tightened by 9% on housing loans and 12% on consumer credit, and demand remains negative, at minus 15% for housing. The European Central Bank attributes this hardening to lower risk tolerance and higher risk perceptions, against a backdrop of geopolitical and energy concerns. For firms, the driver is mainly the cost of credit, consistent with the June 2026 policy-rate hike. The survey was conducted from 15 to 30 June among 159 banks, with a 100% response rate. Where the pressure moves If the regulated bank channel holds up better than expected, the question becomes that of the credit that left this channel. For a decade, a growing share of medium-sized firms' financing has shifted from bank balance sheets to private debt funds. The Financial Stability Board, in its 6 May 2026 report, sizes this market at between $1.5 and $2 trillion at end-2024, the United States ahead, followed by the euro area and the United Kingdom. The FSB explicitly names one of the drivers of this growth: post-crisis changes to bank regulation, which made certain loans less attractive for banks and pushed borrowers toward faster, more flexible nonbank lenders. This is the shift we have tracked for a long time, from the migration of credit risk out of the regulatory gaze to private credit as the new shadow banking. The ECB survey lights up the visible channel; the FSB report lights up the blind spot. And it is in the blind spot that the signals are tightening. The cracks in private credit The FSB is cautious in tone, precise in its findings. Private credit default rates remain low, but they rise as soon as broader measures are used, including selective defaults and distressed exchanges. Borrowers rely more on payment-in-kind, or PIK, where interest is capitalised rather than paid in cash: a short-term liquidity tool that, the FSB notes, often signals deteriorating credit among the weakest borrowers. Valuations, finally, are conducted less frequently and with a high degree of discretion, which amplifies uncertainty, a theme we dug into with one asset, two prices and the zombie funds. The specialist press puts examples on these mechanics. Non-traded private debt funds aimed at retail investors saw, in the first quarter of 2026, their first quarter of net redemptions exceeding subscriptions; a large fund came within a whisker of the 5% quarterly redemption cap, and several managers imposed gates, which we detailed for the gating of a semi-liquid fund and the record June default. Moody's moved the BDC sector outlook from stable to negative, and a major investment bank floats a direct-lending default rate that could climb toward 8%. These figures come from market analysis, not a regulator; they illustrate, without proving on their own, the FSB's cautious diagnosis. The thread linking the two channels The real systemic question is neither the bank channel alone nor the private channel alone, but their seam. The FSB describes an ecosystem where banks and private debt funds are intertwined: banks finance the funds through credit lines, they extend NAV-type facilities and portfolio financing that let funds take on leverage, and they lend on a revolving basis to companies that borrow simultaneously from the same funds. Banks' direct exposure to private credit funds is judged "relatively modest," but poorly measured: FSB member data captures only about $220 billion of drawn and undrawn credit lines, with large uncertainty. At the other end, insurers and pension funds are major investors, drawn by the illiquidity premium, and retail participation is rising, the loop we described for life insurers and private credit in Bermuda and via NAV loans. The reassuring side Fairness demands taking the opposite reading seriously, because it is solid. The ECB survey does not describe a credit crisis: tightening for firms is deflating, their demand is picking up, and the emerging rate-cut cycle would ease the constraint further. The FSB, for its part, judges banks' direct exposure to private credit "relatively modest," recalls that this financing supports activity by offering tailored solutions to underserved borrowers, and stresses that the presence of insurers and pension funds, with long liabilities, is consistent with the illiquidity of these loans. A channel funded by investors with matching mandates is less fragile than one funded short-term. The serious objections therefore bear not on today, but on the test that has not happened. The FSB says it plainly: private credit has never been tested by a prolonged downturn. To that it adds two blind spots it documents, the insufficiency of data on liquidity mismatches in semi-liquid funds, and the opacity of valuations. The rise of retail investors in vehicles with redemption options introduces precisely the liquidity risk that closed-end structures avoided. The real test ahead Three deadlines will tell whether the divergence closes or settles in. The next ECB survey, for which banks already expect further tightening across all segments in the third quarter: the turning point will read there. The second-quarter results of listed BDCs, due in August and September, which will give the first post-spring measure of non-accrual rates and dividend sustainability. And the first European stress test devoted to nonbank players, scheduled for 2026, which will force supervisors for the first time to quantify what the FSB, for now, only sketches. The 21 July survey will thus have rendered an unexpected service: a reminder that the channel we watch best, regulated bank credit, is also the one that holds up best. The risk has migrated to where measurement is weakest. Reading both channels at once, and the thread linking them, is now the only honest reading of the credit cycle. --- Primary and official sources: ECB, euro area bank lending survey, July 2026 (press release) and full second-quarter 2026 report; Financial Stability Board, "Report on Vulnerabilities in Private Credit" (6 May 2026). Analysis and press: Bloomberg, euro-zone banks tighten corporate credit standards (21 July 2026); Private Debt Investor, BDC redemptions exceed inflows and defaults at a record high. The survey and FSB figures are verified against their original sources; the market data on BDCs (redemptions, caps, Moody's outlook, default projection) come from private analysis, cited as such, and illustrate without proving the FSB's diagnosis. To go further: our guides to reading private credit risk and reading bank health. ============================================================================ ANALYSIS: CLARITY Act: Trump concedes on ethics, the August countdown begins URL: https://l0g.fr/en/analysis/clarity-act-trump-concedes-ethics-august-countdown/ Canonical French source: https://l0g.fr/posts/clarity-act-trump-cede-ethique-compte-a-rebours-aout/ Date: 2026-07-21 (reviewed 2026-08-06) Topics: crypto, regulation, us politics, stablecoins ---------------------------------------------------------------------------- Update, 7 August 2026. The in-principle agreement described here has not yet produced a law. H.R. 3633 remains officially Passed House, not Passed Senate. The 10 August to 11 September break shifts the next pre-midterm window to the fourteen scheduled days between 14 September and 2 October. See the procedural status and explicitly conditional scenarios in CLARITY Act: the window before the midterms. The blockage came from the top, and so did its lifting. On the evening of 20 July 2026, according to The Block, Donald Trump agreed to an ethics provision in the CLARITY Act, the text meant to split crypto regulation between the SEC and the CFTC. The provision would restrict how senior public officials, president, vice president, members of Congress and other federal officials, profit from digital assets while in office. This clause targets his own interests first, the obstacle we identified as central in our analysis of Trump, the first obstacle to his own crypto law. The paradox thus unwinds where it was tied. But the agreement comes late, on a text no one has yet read, and against a calendar now counted in session days. The 20 July reversal The sequence is fast and documented. The White House sent an ethics package to certain Senate Republicans on the evening of Monday 20 July, with Senators Bernie Moreno and Cynthia Lummis named among the recipients of the text, The Block reports. A high-level meeting held on 16 July had produced nothing; the 20 July one broke the sticking point. The amended text is expected "in the coming days," with no firm date. The exact content is what matters, and it is the knot. The language circulating, a rule applying to "all" officials rather than to a named office, matches the compromise the White House had long deemed acceptable, and that the firmest Democrats called insufficient. A general rule paired with a long transition period might never force the sitting president to shed his positions. The agreement in principle says nothing yet about real scope: family perimeter, enforcement mechanism, timing. On that last point, CoinDesk reported in mid-July that Democrats were pushing to extend the restrictions to family members, with disclosure requirements and ownership bans. The financial stake, for its part, is quantified and public. Donald Trump's certified annual report, published by the Office of Government Ethics, lists $635,068,835 in royalties tied to the memecoin bearing his name, within CIC Digital LLC. The press aggregates to more than $1.4bn the crypto income declared for the first year of the second term. It is this reality, entered into an official document, that the ethics clause claims to fence in. The ten days before The reversal takes on its full meaning set against the fortnight that preceded it. The Senate returned from recess on 13 July. On the 17th, the bill merging the Banking and Agriculture Committee versions was released after a White House meeting bringing together the president and industry players. That merged text omitted the ethics clause that Democrats had set as the price of their votes. That same day, Senators Chris Murphy, Chris Van Hollen and Jeff Merkley held a press conference to formally oppose the text. In three days, the file went from a bill with no safeguard, publicly rejected by the Democratic wing, to a presidential agreement on the principle of a safeguard. The move is real, but it returns the debate to where it stood: to the nature of the clause, not its existence. The 60-vote wall, still standing The White House deal changes nothing in the arithmetic of the floor. Breaking a filibuster takes 60 votes. With 53 Republicans, that means at least seven Democrats, and those seven had not yet seen the text at the time of the agreement. The Democratic negotiators, per CoinDesk, notably include Chris Murphy, Kirsten Gillibrand, Chris Van Hollen and Jeff Merkley; on the Republican and executive side, Senator Lummis and White House crypto adviser Patrick Witt. Everything turns on one word: whether the clause is binding or cosmetic. A provision "for all" paired with a multi-year grace period and a weak enforcement mechanism would offer a political exit with no real constraint on the sitting president. A biting provision, with a family perimeter and effective avenues of recourse, would satisfy Democrats but clash with the administration's initial red line. The idea of letting state attorneys general sue for violations has circulated as an enforcement route. The seven votes will depend on that dial, not on the principle now settled. The calendar as arbiter Even a satisfactory text would run into the clock. The Senate begins its summer recess on 7 August 2026, leaving about ten useful session days since the 13 July return. The realistic voting window runs from 27 July to 7 August, but no cloture motion has been filed and Majority Leader John Thune has not yet allocated floor time, per the coverage of the file. He had floated a vote "this month" in mid-July, without a formal commitment. The floor vote is not even the last step. The Senate text merges the Banking and Agriculture Committee versions, the latter with jurisdiction over the CFTC; once passed, it must be reconciled with the House version, H.R. 3633, adopted on 17 July 2025 and placed on the Senate calendar as No. 423 since 1 June 2026, before the presidential signature. Senator Lummis summed up the stakes of the window: a failure before August could push the next workable opportunity to the end of the decade, with a new Congress then having to rebuild the bipartisan coalition from scratch. Three scenarios under calendar pressure What follows is scenario, not data. Each branch is anchored to a concrete deadline, with no bet on which one prevails. The first scenario is an express passage. The amended text comes out in the next few days, its ethics clause satisfies at least seven Democrats, Thune files a cloture motion and books floor time, the Senate votes before 7 August, the House accepts the changes or a lightning conference ratifies them, and the president signs. A prediction market priced, after Trump's agreement, the probability of enactment in 2026 at around 44%, sharply up on the day before. The second scenario is a deadlock on substance. The text appears, but Murphy, Van Hollen and Merkley read into it a "for all" clause hollowed out by a long grace period and weak enforcement. Short of seven votes, the text slips past 7 August, and the window closes onto the midterm campaign. The third scenario is a calendar failure, independent of substance. Even with a deal judged serious, the absence of booked floor time, a cloture motion filed too late, or friction over reconciliation with the House would be enough to exhaust the counter before the recess. The agreement would hold, the law would not. The triggers to watch The list is short and dated. Publication of the amended text and the presence, or not, of a genuine ethics clause. Public reaction from Murphy, Van Hollen and Merkley, and a signal from Gillibrand, long crypto-friendly but firm on ethics. Filing of a cloture motion and booking of floor time by Thune. Finally, the reconciliation calendar with the House. The obstacle the president embodied appears lifted in principle; what remains to be seen is whether ten session days are enough to turn a last-minute deal into law, and whether the clause that unblocks everything truly deserves the name. --- Primary and official sources: Congress.gov, H.R. 3633, actions and status (Calendar No. 423); Senate Banking Committee, 15-9 advance of 14 May 2026; Senate Agriculture Committee, timeline announced by Chairman Boozman; Office of Government Ethics, Donald Trump's 278e report ($635,068,835 in memecoin royalties). Analysis and press: The Block, Trump agrees to the ethics provision (20 July 2026); CoinDesk, Trump's crypto riches loom over the talks (13 July 2026); CryptoBriefing, merged draft and Democratic opposition; CoinGape, Trump's agreement on the ethics clause; Yahoo Finance, countdown to 7 August. Figures and dates checked against the sources cited; the state of the text changes day to day, and the situation described is that of 21 July 2026. Items not consolidated in a primary source (the clause's content, the voting calendar, the market probability) are attributed to the media cited and presented as such. ============================================================================ ANALYSIS: Fiscal dominance made real: Japan subordinates its central bank to its budget URL: https://l0g.fr/en/analysis/japan-fiscal-dominance-honebuto-shock-boj-cornered/ Canonical French source: https://l0g.fr/posts/japon-dominance-fiscale-choc-honebuto-boj-cernee/ Date: 2026-07-21 (reviewed 2026-07-21) Topics: japan, jgb, boj, debt, macro, markets ---------------------------------------------------------------------------- Fiscal dominance is usually defined in textbooks: a situation where the level of public debt forces the central bank to keep rates lower than it would like, price stability yielding to the financing of the state. Japan has just supplied an almost literal illustration. In the draft of its economic guidelines, the Honebuto, Sanae Takaichi's cabinet dropped its fiscal consolidation targets and invoked the legal clause requiring the Bank of Japan to support the government's growth agenda, without mentioning the neighbouring clause that guarantees its independence. The bond market read the text for what it was. As of 21 July 2026, the Japanese 30-year yields 3.90%, the 40-year 3.91%, the 10-year 2.74%, the last of these up 122 basis points over twelve months. On a debt stock at 204% of gross domestic product, every tenth of a point matters. The Honebuto shock The catalyst was not an inflation print or a failed auction, but a drafting choice. Japan's annual budget guidelines frame the government's priorities; the 2026 edition erased the return-to-balance targets that previous cabinets displayed, at least for form's sake. Above all, the draft describes monetary policy as needing to be "guided appropriately to achieve a stronger economy" and points to the legal article asking the BoJ to align its decisions with the executive's economic strategy, while staying silent on the provision that protects its independence. The omission was enough. Long-dated JGB yields climbed to multi-decade highs, a move the financial press dubbed the Honebuto shock. The political context gives this text a weight an isolated statement would not have carried. Sanae Takaichi, Japan's first female leader since October 2025, dissolved the lower house and won a snap election on 8 February 2026, with a clear mandate to spend: tax cuts, higher defence spending, stimulus. The fiscal-year 2026 budget, the largest in the country's history, was passed in short order. The Honebuto is therefore no slip of the pen: it is the budget doctrine of a government that has just won on that platform, set down in black and white. The central bank cornered The Bank of Japan had begun normalising well before the shock. It raised its policy rate to 1% on 16 June 2026, its highest since 1995, after a late-2025 hike had lifted the 10-year back above 2%. The problem is not that it refuses to tighten, but that it is caught between two forces each pulling the opposite way. On one side, inflation and the currency call for higher rates. The yen trades around 162 to the dollar as of 21 July 2026, a level that keeps it near multi-decade lows and fuels imported inflation. On the other, gross public debt reaches 204.4% of gross domestic product in 2026 according to the International Monetary Fund, after 206.5% in 2025: on such a stock, each additional point of yield eventually swells debt service heavily as bonds roll over. Raising rates to defend the yen and contain inflation means driving up the interest bill of the most indebted state in the developed world. There is the vice, and it is precisely what the concept of fiscal dominance names: the budget constraint ends up setting the ceiling of monetary policy. The government did try to put out the fire it had lit. Minister for Economic Policy Minoru Kiuchi insisted that the BoJ's autonomy must be respected and that the executive would not give it advance signals on the timing or scale of its moves, with the revised Honebuto text due before the cabinet. A denial has rarely confirmed the subject it claimed to close quite so well. The detail of recent moves confirms the nature of the signal. Over the past month, the sharpest rise concentrates on the 40-year, the maturity most exposed to a long-run fiscal slippage. When investors demand more to hold the most distant debt, it is not the business cycle they fear, it is the trajectory. The nudge on pension funds Facing rising long yields, a state has several levers. Japan has just pulled a very old one. Finance Minister Satsuki Katayama said she wanted to encourage pension funds, starting with the GPIF, to substantially raise their holdings of Japanese financial assets. The GPIF is no ordinary investor: it managed roughly 293.4 trillion yen, close to 1.81 trillion dollars, at the end of December, split in near-equal parts between equities and bonds, domestic and foreign. Steering such a mass toward local sovereign debt means calling on a captive buyer when market buyers grow reluctant. This mechanism has a name, financial repression: mobilising domestic institutional savings to fund the state at a yield the free market would not accept at the same price. Japan has for this an asset few countries possess, deep domestic savings and a historical home bias. But the tool has a cost: it shifts interest-rate risk onto future pensions and signals that the state is relying on something other than investor confidence alone to place its paper. The global transmission belt What plays out in Tokyo does not stay in Tokyo. Japan remains the world's great net creditor, and the rate gap between a cheap yen and better-paid foreign assets has for years fuelled the yen carry trade, that borrowing in a weak currency to invest elsewhere. We documented its mechanics in our analysis of the fuse lodged in Japanese bonds and in our piece on the unwind risk tied to FX intervention. The rise in Japanese yields acts on this construct like a magnet. As long-dated JGBs offer 3.9%, Japanese savings parked abroad find reasons to come home, and the cost of carry climbs. A faster BoJ tightening, should it become necessary to defend the yen, would revive the risk of a disorderly unwind of carry positions, which the August 2024 episode showed could shake US technology stocks. How Japan resolves its budget vice therefore weighs on liquidity for the rest of the markets. The sceptics' counter-argument Caution demands laying out the other reading with equal rigour, because it is solid. Several research houses judge the panic premature. Capital Economics urges ignoring comparisons with the UK's Truss episode: the rise in Japanese yields would reflect less a crisis than a return to normal after decades of near-zero rates. Morningstar, for its part, sees a risk of future fiscal constraint but a limited impact on equities. The substantive arguments come down to three points. Japanese debt is yen-denominated, overwhelmingly held at home and backed by abundant domestic savings: a state that owes in its own currency to its own residents does not default the way a foreign-currency borrower does. The BoJ, moreover, remains the top holder of JGBs, with a share just back below half the market: it retains a capacity to intervene few central banks match. Finally, for all their nominal surge, real yields remain modest once inflation is stripped out, and a 10-year at 2.74% is still low by international standards. Normalising a zero-rate regime is not, in itself, a crisis. The tipping points to watch The rest is scenario, not observed data. Three markers will tell whether the vice loosens or closes. The first is the final Honebuto text adopted by the cabinet: a rewrite explicitly restoring the BoJ's independence would count as a denial more credible than any soothing statement. The second is the behaviour of the 30-year and 40-year: as long as they stay contained, fiscal dominance remains a latent threat; an acceleration toward new records would bring Japan closer to a genuine funding test. The third is the yen: a durable break of current lows would force the BoJ to choose between the currency and the debt, the very heart of the trap. The Japanese singularity, captive savings and a net claim on the world, has let the country carry the heaviest debt in the developed world without a funding crisis. The Honebuto shock tests the limit of that singularity, not through a market event, but through a sentence the government wrote itself. What follows will tell whether words count as much as numbers. --- Data and primary sources: JGB curve and policy rate, worldgovernmentbonds.com (21 July 2026); Japan gross public debt, IMF DataMapper, WEO April 2026; USD/JPY exchange rate (21 July 2026). Analysis and press: CNBC, Japan's bond market back in play and the Honebuto shock and the BoJ hike to 1%; InvestingLive, pressure on the GPIF and BoJ independence fears; CSIS, Takaichi's lower-house election win; Euronews, Japan's largest-ever budget; Japan Times, the 40-year at 4% in January and the yen carry revival; Capital Economics, ignore the Truss comparison; Morningstar, future fiscal constraint but limited equity impact. Yields, exchange rates and levels move continuously; the values cited are those recorded on 21 July 2026. Items not verifiable against a primary source (the Honebuto shock, the political calendar, the budget) are attributed to the media cited. ============================================================================ ANALYSIS: Argentina, half the lending window: anatomy of the largest exposure in IMF history URL: https://l0g.fr/en/analysis/argentina-the-imf-largest-ever-exposure/ Canonical French source: https://l0g.fr/posts/argentine-la-plus-grande-exposition-de-l-histoire-du-fmi/ Date: 2026-07-20 (reviewed 2026-07-20) Topics: imf, argentina, debt, macro, markets ---------------------------------------------------------------------------- On 28 July, Kristalina Georgieva lands in Buenos Aires for two days of meetings with Javier Milei and his economic team, a trip billed as a clear signal of support for the government. Behind the photo opportunity sits a number the International Monetary Fund usually keeps in its technical annexes: as of 17 July 2026, Argentina owes SDR 42.55 billion to the IMF's general lending window, roughly $57.9 billion at the current SDR rate. A single country concentrates 46.1% of the outstanding non-concessional credit of the institution charged with safeguarding the monetary stability of 191 member states. The managing director's trip is not just a show of support: it is a creditor coming to inspect its principal asset. One borrower, nearly half the window The orders of magnitude deserve to be laid out calmly, because they have no precedent. Argentina's financial position in the Fund as of 30 June 2026 shows SDR 42.55 billion outstanding under extended arrangements, 1,335% of its quota. The normal cumulative access limit to Fund resources is set at 600% of quota: Argentina sits at more than double that, a level reserved for the so-called exceptional access procedure. Its SDR holdings tell the rest of the story: 35 million against a cumulative allocation of 5,075 million, or 0.69%. The country has consumed its international reserve allocation down to the stub. Measured against the lender's portfolio, the concentration is starker still. Total outstanding credit of the general window, the General Resources Account, stood at SDR 92.25 billion on 17 July 2026: Argentina's share works out at 46.1%. Adding the concessional windows, the Fund's total credit outstanding reaches SDR 122.7 billion, of which Argentina still accounts for 34.7%, far ahead of Ukraine (10.4 billion), Pakistan (8.1), Ecuador (7.1) and Egypt (6.8). The five largest borrowers of the general window add up to 81% of the book. To situate the historical anomaly: the financial risk assessment published by Fund staff at programme approval recalled that the largest borrower averaged 27% of outstanding GRA credit between 1985 and 2010, and projected an Argentine peak of SDR 43.1 billion in 2026, "the Fund's largest-ever exposure in absolute terms", nearly 9 billion above the 2022 peak. Three programmes, one debt This mountain did not form in a single cycle. The current arrangement is, by the Atlantic Council's count, Argentina's twenty-third IMF programme since the 1950s. The recent sequence fits in three lines of the commitments table. In 2018, the Macri-era stand-by was approved for SDR 40.7 billion, the largest arrangement ever signed, of which 31.9 billion was actually drawn. In 2022, the refinancing EFF was calibrated at 31.9 billion: exactly the amount drawn under the stand-by, whose maturities it served to meet. In April 2025, the Fund approved a new 48-month, $20 billion extended arrangement, with $12 billion disbursed upfront. Of those $20 billion, about 11 will go to repaying the IMF itself over the life of the programme. Each arrangement papers over the previous one; the principal never comes down. The current programme has already had its stumble. The December 2025 target on net international reserves was missed, against a backdrop of massive dollarization of savings ahead of the October midterms, and the Fund granted a waiver while lowering the accumulation target. The second review, completed by the Executive Board on 21 May 2026 with a $1 billion disbursement, brings total disbursements to $15.8 billion of the planned 20. The end-June targets, recalibrated on that occasion, will be the yardstick of the next review. The Milei bet, seen from July 2026 It needs saying plainly, because objectivity demands it: as of mid-2026, the indicators side with Buenos Aires. June inflation came in at 1.9% on the month, the slowest pace in ten months, and 33.5% year on year, against 211% at end-2023. The IMF confirms growth of 3.5% in 2026 and around 4% in 2027. Country risk fell in early July to around 415-421 basis points, its lowest since 2018, lifted by rating upgrades whose mechanics our guide reading a credit rating sets out. The central bank, now running an exchange rate band indexed to inflation, has bought around $7.5 billion of foreign currency since January, rebuilding net reserves by $4.8 billion while the peso appreciated 13% in real terms. Debt management itself has turned professional. In early July, Economy Minister Luis Caputo presented a financing plan covering maturities through end-2027 without returning to the international bond market: $19.2 billion of needs in 2026, $22.9 billion of identified resources across FX purchases, domestic debt, multilaterals and privatisations. Bank offers to place $5 billion of ten-year paper were turned down, their 12.5% rate judged prohibitive. Instead, Decree 478 authorises up to $5 billion of borrowing partially guaranteed by the World Bank and the IDB, at a hoped-for cost of around 6.5%, to cover the $4.5 billion of July maturities. And the government has tabled a reform of the central bank charter refocused on inflation, which the IMF welcomed as strengthening its independence. The repayment wall The calendar tempers the enthusiasm. The schedule published by the IMF on existing credit alone, before the programme's remaining drawings, traces a steep slope: $2.8 billion due to the Fund over the second half of 2026, then $7.7 billion in 2027, 9.5 in 2028, 10.6 in 2029 and 11.5 in 2030, at the current SDR rate. Over five years, about $42 billion, of which $12.4 billion in charges and fees, including the surcharges applied to credit far in excess of quota. At the end of the programme, in April 2029, projected outstanding credit will still be SDR 35.5 billion, 1,115% of quota, still above the normal access limit. The official financing plan covers 2026 and 2027. The following steps, 9.5 then 10.6 then 11.5 billion, assume either an external surplus durably above projections, or a return to international markets at rates far below the 12.5% refused this summer. The investors who reckoned in March that the country had let an issuance window pass were already asking the question that will decide what follows: at what price, and when, can Argentina refinance itself without its official creditor? A creditor under influence Now walk the balance sheet in the other direction, because the dependence does not run one way. The staff assessment published at programme approval says it in unusually direct language: Argentina's capacity to repay remains subject to "exceptionally high" credit risks, and the exposure far exceeds the precautionary balances, the IMF's capital cushion, projected at SDR 25.9 billion for end-April 2025. Argentine credit represented 155% of that cushion at approval; it represents roughly one and a half times it today. Another figure from the same document, less commented on: the charges and surcharges paid by Argentina in fiscal year 2026 alone are equivalent to 249% of the Fund's residual burden-sharing capacity, the mechanism that would spread the cost of an arrears event across creditors and debtors. The IMF's largest risk is also its largest paying customer. On top of the financial concentration sits a political one. The October 2025 rescue was co-financed by Washington directly: a $20 billion currency swap backed by the US Treasury's Exchange Stabilization Fund, of which Argentina drew $2.5 billion, repaid in January 2026 according to Scott Bessent. The facility remains open, to the point that Senator Elizabeth Warren is demanding its termination. The IMF's dominant shareholder is thus, in parallel, the bilateral guarantor of the same debtor: the Fund's credit risk and American foreign policy towards the Milei government have become hard to tell apart. One pillar of our analysis of the emerging markets double squeeze applies here to the creditor itself: the global safety net heads into the next shock with a balance sheet already committed. The opposite reading The case for the defence is solid, and it deserves the same rigour. First, the IMF enjoys de facto preferred creditor status: in the history of the general window, arrears episodes, from Peru in the 1980s to Greece in 2015, were resolved without a definitive principal loss for the Fund. Second, an exposure is not a loss: it self-liquidates if the programme succeeds, and this programme is posting results none of the previous twenty-two achieved at this stage, a sustained primary surplus, rapid disinflation and a government elected precisely on that adjustment. The same staff assessment that flags the concentration also concludes that the Fund's liquidity would remain adequate even with this programme on the books. Finally, American backing, whatever one thinks of its political dimension, reduces near-term liquidity risk: a debtor leaning on two windows rarely defaults on either. The serious objections therefore bear not on 2026 but on repetition. Net reserves remain below the original trajectory despite the waiver, the exchange rate band has yet to weather a real storm, and the 2027 presidential election will put the programme's political continuity back in play, just as the 2025 midterms were enough to trigger defensive dollarization. The 2018 precedent hangs over the whole analysis: that programme too displayed, in its first eighteen months, completed reviews and falling spreads, before the credibility trap snapped shut. Three trajectories What follows is scenario, not observed data. The high path is the one official projections trace: reserve targets met, a return to the international bond market in the course of 2027 at single-digit rates, and an Argentine share of GRA credit falling back below 40% by fiscal year 2028, as the approval scenario envisaged. The middle path is an external shock, strong dollar and wartime barrel, the vice described in our emerging markets analysis: reserve slippage, fresh waivers, then, at the programme's end, a successor arrangement rolling the credit over for a fourth time since 2018. The low path, low in probability but major in consequence, would run through a political rupture in 2027 and would confront the Fund with a question it has never faced at this scale: an arrears case representing nearly half its portfolio and a multiple of its burden-sharing capacity. The signals to watch through year-end fit on a short list: the end-June reserve targets, on the menu of the third review, country risk holding below 400 basis points, the resilience of the inflation-indexed band, the fate of the US swap line, and the tone, ceremonial or substantive, of Georgieva's visit on 28-29 July. The IMF has lent to Argentina to the point of making it nearly half of its lending business. The success of the Milei bet would decide much more than the fate of one programme: it would determine whether the world's lender of last resort regains, or not, the freedom to act elsewhere. --- Primary sources: IMF, Argentina's financial position in the Fund as of 30 June 2026; IMF, GRA Credit Outstanding as of 17 July 2026 and Total IMF Credit Outstanding; IMF, SDR Valuation; IMF, assessment of the Fund's financial exposure and liquidity position (April 2025); IMF, approval of the $20 billion extended arrangement (11 April 2025); IMF, second review and Article IV consultation (21 May 2026); Congressional Research Service, U.S. Financial Support to Argentina (R48780). Analysis and press: Atlantic Council, four questions on the $20 billion rescue; PIIE, Argentina's credibility trap; Buenos Aires Times on the second review, the central bank reform and Georgieva's visit and the growth forecasts; Buenos Aires Herald on the reserve waiver and the inflation-indexed bands; MercoPress on country risk at an eight-year low; UPI on the financing plan through 2027; Rio Times on Decree 478 and Argentina's economy in 2026; Bloomberg on the missed issuance window and Elizabeth Warren's call to close the swap line; Fortune on the repayment of the swap drawdown. Figures checked against the sources cited; credit outstanding, SDR rates and spreads move continuously, the levels quoted are those of 17-20 July 2026. ============================================================================ ANALYSIS: The Gulf is thirsty: desalinated water, the petromonarchies' financial Achilles heel URL: https://l0g.fr/en/analysis/gulf-thirst-desalination-financial-achilles-heel/ Canonical French source: https://l0g.fr/posts/golfe-a-soif-eau-dessalee-talon-achille-financier/ Date: 2026-07-20 (reviewed 2026-07-20) Topics: geopolitics, energy, water, risk, iran ---------------------------------------------------------------------------- On Friday 17 July, Iranian missiles damaged a power plant and a desalination facility in Kuwait, starting a fire and triggering emergency plans. Two days later, a second strike hit the same type of installation. Markets read these attacks the way everyone did: through the barrel, up 13% on the week. The heavier reading lies elsewhere. In a country where 90% of drinking water comes out of desalination plants, targeting these facilities is not energy sabotage: it is holding an entire population within reach of thirst. And what holds for Kuwait holds, in varying degrees, for all the petromonarchies that finance the world's debt and its tech. An X-ray of a vulnerability that asset prices display nowhere. The facts first, because they are recent and precise. On 17 July, Kuwait confirmed that an Iranian attack had damaged power generation units and a desalination station, starting a fire and triggering emergency plans. On the 19th, a second strike hit a combined power and water installation, the second in two days. Earlier in the war, the Doha West plant had already been damaged by debris from interceptions near the port. The line that sums up the stakes comes from the wires themselves: were the big plants knocked out, some cities would lose most of their drinking water within days. The architecture of thirst The dependence is structural, quantified, and concentrated. The region operates more than 400 desalination plants producing 7.2 billion cubic metres a year, 40% of the world's desalinated water. As a share of total water use, desalination ranges from 61% in Qatar and 59% in Bahrain down to 18% in Saudi Arabia; as a share of the water people drink, it becomes overwhelming: more than 99% in Qatar, more than 90% in Bahrain, about 90% in Kuwait, more than 80% in the Emirates. The water table, where it still exists, is brackish or set aside: Bahrain became fully dependent as early as 2016 and keeps its groundwater as a contingency reserve. This architecture stacks three fragilities that multiply each other. Concentration: a few coastal mega-sites produce most of the output, and the Gulf coastline is precisely the geography that Iranian missiles and drones cover. Cascading dependence: no water without electricity, no electricity without gas or oil, so that combined power-and-water plants, like those struck in Kuwait, take down two networks at once. Short-term irreplaceability: a desalination plant takes years to build, and nothing trucks in water at the scale of a capital city. In our usual reading grid, each mega-site is an immobile chokepoint: Hormuz concentrates flows that can partly be rerouted, a desalination plant concentrates a need that reroutes nowhere. Days, not months The contrast with energy gives the measure of the problem, and it is cruel for states built on hydrocarbons. On the oil side, the world has equipped itself with thick cushions: the IEA's 90-day strategic reserve rule, Chinese stocks estimated at 1.24 billion barrels, the whole arsenal we inventoried in our analysis of strategic reserves in the second round. On the water side, the order of magnitude changes in kind: Gulf storage counts in days of consumption, not months. Qatar built mega-reservoirs at great expense to extend its autonomy to about a week; Bahrain hoards its aquifer as a last resort; and Friday's wires recall that in Kuwait, the loss of the major plants would translate into urban shortage «within days». This asymmetry is the strategic heart of the case. An oil embargo negotiates over months, the time it takes for stocks to run down; a water interruption negotiates over a week. An attacker who targets water thus buys a lever of coercion incomparably faster than anything the energy market can produce, at trivial military cost. The 2026 war has just demonstrated it publicly, three times, on the same small state. The fragile creditor's paradox The financial reading starts here, because these hydraulically mortal states are the structural creditors of the global system. Gulf sovereign wealth funds manage between $4 and $6 trillion in assets, and deployed $119 billion in 2025 alone, most of it toward the United States, from funding AI hyperscalers to infrastructure funds. The Council on Foreign Relations posed the question as early as May, under a title that says it all, "disappearing Gulf capital": a war that settles in forces these funds to repatriate capital, for defence, reconstruction, budget support, and this reallocation risk, "much less obvious than rising gasoline prices", would weigh first on the American markets that have grown used to the manna. The strikes on water harden that scenario by a notch. A sovereign wealth fund is a bet on the permanence of the state that owns it; demonstrating that this state can be made thirsty within a week changes the calculus of everyone who manages one. Concretely: more domestic resilience spending (storage, air defence, plant redundancy, interconnections), less capital available for Silicon Valley funding rounds, and a new precautionary premium on everything the region finances. The Middle East Council on Global Affairs already speaks of the "cost of crisis resilience" for the sovereign funds: war turns return machines into insurers of the state. The price markets do not display There remains the quintessential l0g question: where does any of this read in prices? In touches, and surprisingly quietly. Gulf issuers have shifted part of their funding to less visible channels, nearly $10 billion raised in private debt since the war began, the signature of a public market grown dearer or more scrutinised. Bahrain's CDS, the region's weak link, embeds an implicit support premium from the rest of the Gulf, meaning the market prices the neighbours' solidarity more than standalone strength. The IMF has formally warned that the war is feeding regional financial stability risks. But no spread break remotely commensurate with the demonstration of vulnerability that just took place: the pattern matches, feature for feature, the 235-point EMBI paradox we documented this morning, pre-shock pricing against post-shock fundamentals. Insurance tells the same story from the other end: maritime war-risk premia have exploded, as we tracked on the Gulf's tankers, but insurance on fixed installations renegotiates privately, away from any screen. The other reading The dark scenario deserves its counterweights, and they are real. The Iranian strike stayed calibrated: damage, a fire, no grid collapse and no declared shortage, and Kuwait's emergency plans worked. Deterrence cuts both ways: Iran itself depends on desalination for part of its coast, as the Qeshm incident early in the war recalled, and making civilian populations thirsty is the kind of threshold that coalises the entire world against its author. Redundancy exists: 400 plants do not go dark in one salvo, the six monarchies interconnect, and Qatar spent billions on reservoirs precisely to buy time. Finally, the region demonstrated in 2019, after the Abqaiq attack on the Saudi oil heartland, a speed of repair that surprised everyone. The vulnerability is proven; its exploitation at scale remains a choice nobody has yet made, because it would change the nature of the war. The dials This case will be tracked on precise signals. Recurrence first: a third strike on water would make it a strategy rather than a signal, and would move all regional pricing. Resilience announcements next: every billion invested in storage, in air defence for the plants or in interconnection is a priced confession of the threat, to be read in Gulf budgets and debt issuance. Sovereign fund flows again: the pace of big American funding rounds, AI first, will say whether the CFR's "disappearing capital" is a hypothesis or a trend; 2026 data against 2025's $119 billion will serve as the yardstick. CDS and sukuk finally, Bahrain as the scout: the day the Gulf's implicit solidarity premium starts to quote hydraulic vulnerability, the subject will have left the think-tank notes for the screens. The underlying irony will remain, whatever the scenario. The states that built their power on the world's most storable molecule, the one that keeps for months in salt caverns, are discovering that their existence hangs on the least storable molecule there is. Oil gave them the means to buy everything, except months of lead time in water. In a world learning to target infrastructure, that asymmetry is now a market fact. --- Primary sources and wires: Al Arabiya, Kuwait's official confirmation of the 17 July attack; Bloomberg (17 July 2026); AP via KSAT, on the vulnerability and Doha West; AP via Las Vegas Sun, the second strike; Al Jazeera, regional dependence data (March 2026) and on the targeting of plants; Middle East Eye on Bahrain and water security. Analysis: CFR, Rebecca Patterson, "Disappearing Gulf Capital" (1 May 2026) ($4-6 trillion of sovereign assets, $119 billion deployed in 2025); Middle East Council on Global Affairs, "Gulf Sovereign Wealth Funds and the Cost of Crisis Resilience"; Atlantic Council on the post-war economic phase; CAIA, "War, the Gulf, and the Pricing of Systemic Risk" (March 2026); Ifri, "The Geopolitics of Seawater Desalination"; CNN on water, more vital than oil (March 2026). Market: Gulf private debt raises (~$10 billion); Tangency Point Capital on Bahrain's CDS and the Gulf support premium; the IMF's warning on regional financial stability. Figures and dates checked against the sources cited; the military situation evolves daily, the state described is that of 20 July 2026. ============================================================================ ANALYSIS: The world's factory marks everything down: China's deflation, shock absorber and poison URL: https://l0g.fr/en/analysis/china-deflation-shock-absorber-and-poison/ Canonical French source: https://l0g.fr/posts/usine-mondiale-brade-deflation-chinoise/ Date: 2026-07-20 (reviewed 2026-07-20) Topics: china, macro, inflation, geopolitics, markets ---------------------------------------------------------------------------- Mid-July delivered three Chinese figures that look contradictory. On the 10th, inflation at 1% and producer prices up 4.1%, their strongest rise since 2022. On the 14th, second-quarter GDP at 4.3%, the first missed target since Covid, published the same day as record exports of $412 billion and the second-largest monthly surplus in history. An economy that disappoints at home, reflates on the surface and exports like never before: the three facts are one. After three years manufacturing deflation for itself, China has found the way to ship it elsewhere, and the West, in the middle of an oil flare-up, has become its first customer, willingly or not. The statistical turn first, because it is real. China's GDP deflator, the broadest measure of prices, turned positive at 1.6% in the second quarter, ending twelve consecutive quarters of decline: the official exit from "quasi-deflation". Producer prices jumped 4.1% year on year in June, a near four-year high. From a distance, the page looks turned. Up close, the composition tells another story: this reflation comes from the wartime barrel inflating input costs and from administered price floors, not from demand. Consumer inflation remains stuck at 1%, core too, and growth slowed to 4.3%, below consensus and below target for the first time since Covid, with property still weighing as we documented in our anatomy of China's real estate risk. Rising prices without demand are not a recovery: they are a removal. The export machine at full throttle The removal has a departure address: the factory. June exports reached $412.4 billion, up 27% year on year, an absolute record, carried by global demand for artificial intelligence hardware and by autos. The monthly trade surplus came in at $125.6 billion, the second-largest in history. Vehicles illustrate the mechanics to the point of caricature: 5.1 million units exported in the first half, up 65%, including 2.35 million electric vehicles, a doubling, while domestic sales fell 13% and only three brands in the whole sector remain profitable. Solar pushes the logic further still: panels sold below variable cost, with capacity utilisation down under 40%. This is not trade in the textbook sense, where you sell what you produce at a profit. It is overcapacity in motion: industries sized for a domestic demand that never came, forced to sell at any price, anywhere, just to keep running. The Chinese have a word for this self-destructive competition, involution, and their government has made it the official enemy of the year. Until the campaign bites, every departing container is a piece of deflation emigrating. Europe, the unloading dock The main destination has changed since the trade war: it is Europe. The European deficit with China reached €360 billion in 2025, up 18%, as goods banned from the American market poured onto the old continent. The movement is accelerating: in June, China's surplus with the European Union hit a record $32.9 billion, up 27%, with the German surplus more than doubling year on year. This dumping ground has an effect the ECB itself quantified before living it: a redirection of Chinese exports toward the euro area could shave up to 0.3 percentage point off core inflation over two years, through the extra supply and compressed prices. The prediction now reads in the data: in June's European inflation, non-energy industrial goods trail at 0.9% while energy burns at 8.7%. Put differently: at the precise moment the wartime barrel pushes European inflation up, discounted Chinese goods pull it down, and the balance of those two forces is what still allows the ECB to debate a pause rather than a full hiking cycle, the trade-off we framed in our analysis of the 2011 remake. Europe's oil-shock absorber is made in Shenzhen. So is its price: European industry pays it, in market share. The American wall, lowered but standing The transatlantic contrast lights up the other half of the picture. The United States largely closed itself to the overflow: after the 2025 escalation that pushed some duties beyond 100%, the Supreme Court struck down the IEEPA-based tariffs in February, bringing the effective rate on Chinese goods to around 30%, still the highest of any partner. The result: America deprived itself of part of the Chinese absorber, and its 4.2% inflation bears the trace, with tariffed goods getting dearer while Europe imports the discount. We showed in our breakdown of American import prices how much reading those indices demands separating fuel from the underlying; the same care applies here. The wall did not stop the flow, it deflected it: China's surplus with the United States holds around $29 billion a month, while the overflow takes the road to Rotterdam and Hamburg. Anti-involution, or the art of moving the problem Beijing no longer denies the diagnosis, and that is new. The "anti-involution" campaign has become official policy: an amendment to the Pricing Law banning below-cost selling, two-year plans for ten key industries, output targets revised down. The logic is to break the spiral in which every producer cuts prices to move a capacity nobody wants to close, a spiral that has turned whole swaths of industry, from solar panels to express delivery, into margin-destruction machines. The short-term paradox still deserves a hard look: domestic price floors without capacity destruction do not remove the overflow, they redirect it. If the factory can no longer discount in Canton, it discounts abroad; the first-half figures, domestic EV sales down 13% and exports doubled, trace exactly that communicating vessel. Anti-involution, 2026 edition, looks less like a detox cure than a change of clientele, and the most sceptical analysts recall that previous attempts foundered on local governments, for whom closing a factory is an immediate political cost against a diffuse national benefit. The other reading Honesty requires running the scenario in which Beijing is winning. The deflator is positive for the first time in three years, the Pricing Law is starting to bite, and if domestic consumption finally takes over, 2026 will be remembered as the year of the turn, not of the disguise. The record surpluses also have a less aggressive reading than dumping: June's Chinese imports jumped 36% to a record $286.8 billion, driven by AI components, the sign of an economy buying massively from the world what its own tech needs. And the flight-forward strategy has a built-in limit: exporters that double their volumes while destroying their margins cannibalise themselves, three profitable carmakers out of dozens being not a model but a countdown. The question is not whether this race stops, but who stops it first: Chinese margins, or Europe's protectionist patience, already dented by the 2024 duties on electric vehicles and brought closer to breaking point by every record surplus. The arbiters A few appointments will settle the readings, and scenarios they remain. Monthly Chinese data first, producer prices and trade: a reflation that held without the support of the wartime barrel would change the nature of the turn. The European trade response next: every anti-subsidy investigation and every record surplus brings Brussels closer to a choice between its industry and its inflation, because taxing the Chinese absorber in the middle of an oil shock would mean inflicting the American fate on itself. The ECB's September projections again, the first that must explicitly arbitrate between the barrel pushing and China pulling. The profitability of the price war finally: the day China's big exporters stop losing money while gaining market share, or the reverse, the dynamic changes regime. The overall thread joins that of our emerging markets double squeeze, published this morning: the 2026 oil shock does not hit a uniform world, it passes through filters. The importing South takes it head-on, America doubles it with a tariff wall that makes everything dearer, and Europe has it cushioned by the Chinese factory, at its industry's expense. Three geographies, three inflations, and in the middle, an exporter that has learned to turn its domestic weakness into a trade weapon. The textbooks called this a symmetric shock; 2026 makes it a revealer of architectures. --- Primary sources: National Bureau of Statistics of China, June 2026 Consumer Price Index and GDP releases; ECB, blog "China-US trade tensions could bring more Chinese exports and lower prices to Europe" (30 July 2025) (the 0.3-point estimate) and box "Where do the costs of higher US tariffs fall?"; Eurostat, June 2026 HICP. Analysis: T. Rowe Price on the anti-involution policy; BNP Paribas AM, "Involution, deflation and structural reform"; China Leadership Monitor on involution and local strategies; CEPR, "The Great Wall of Chinese goods"; Tax Foundation, Tariff Tracker (the IEEPA ruling, ~30% effective rate). Press and data: CNBC (9 July 2026) and CNN (14 July 2026); RTE, Fortune, China Global South and Mitrade for June trade; Trading Economics, trade balance; BigGo Finance on the deflator; TechTimes on carmaker profitability; Courthouse News on the 2025 European deficit; Bloomberg on electric vehicle exports. Figures and dates checked against the sources cited; the monthly Chinese data are those published on 10 and 14 July 2026. ============================================================================ ANALYSIS: The emerging markets double squeeze: strong dollar, wartime barrel URL: https://l0g.fr/en/analysis/emerging-markets-double-squeeze-dollar-oil/ Canonical French source: https://l0g.fr/posts/double-peine-emergents-dollar-fort-baril-de-guerre/ Date: 2026-07-20 (reviewed 2026-07-20) Topics: macro, dollar, debt, geopolitics, emerging markets ---------------------------------------------------------------------------- On Friday 17 July, Brent closed at $88, up 13% on the week after the attack on a desalination plant in Kuwait. The same day, the dollar sat near a thirteen-month high, propelled by markets that have started betting on a Federal Reserve hike again. For an energy importer of the South, those two quotes are one and the same: the oil bill swells in volume and in currency at once. The last time this vice closed, in 2022, it ended at the International Monetary Fund for Pakistan, Egypt and Sri Lanka. Here it is being reassembled, piece by piece, in front of emerging debt markets settled at multi-year lows. The numbers of the pincer fit in three lines. Oil first: Brent at $88.10 on 17 July, up 27% over a year, after a full round trip in three weeks, having fallen below $75 in late June before the war relit the premium. The dollar next: an index up about 2.5% in June alone, close to a thirteen-month peak, supported by Fed rate-hike bets fed by the oil surge, with US inflation at 4.2% closing the door on any quick easing, a mechanism we dissected in our analysis of the dollar rebound. Emerging currencies finally: the MSCI Emerging Market Currency Index erased the whole of its 2026 gains in early July, its lowest since April. Three quotes, one single move: everything the South imports costs more, in a currency that is appreciating, financed at rates that are climbing. The vice closes The mechanics deserve to be laid out, because there is nothing abstract about them. Oil is invoiced in dollars; when the barrel and the greenback rise together, the emerging importer suffers a double, multiplicative increase, price times currency. The energy bill widens the trade deficit, the deficit weighs on the currency, the sliding currency makes the bill dearer still: the loop feeds itself, and it bites first where the cushions are thinnest. The IMF wrote it in black and white as early as March: energy importers across Africa, the Middle East and Latin America are absorbing heavier import bills on top of already limited fiscal space and external buffers, with the fuel, fertiliser and food trio widening deficits and pressing currencies. Then comes the financial layer. Seen from the South, American tightening is not a piece of domestic policy news: it is the refinancing price of a stock of dollar debt the BIS puts at about $4.3 trillion for emerging and developing economies at end-2025, up 35% in ten years, more than half in the form of securities. This debt in other people's money, the emerging world's old "original sin", runs through the offshore plumbing we mapped in our piece on eurodollars: when the Fed hardens and the dollar climbs, servicing that debt gets dearer without any parliament of the South having voted on anything. The map of fragilities The shock does not hit a homogeneous category but a precise geography. In the front line, energy importers with thin reserves. Egypt piles up the wounds: diminished Suez Canal and tourism receipts, a pound slipped to 50.55 per dollar, and an IMF drip whose latest tranche of about $2.3 billion was unlocked in late February. Pakistan has seen its food import bill jump to $9.1 billion in fiscal 2026 while the Gulf war has produced fuel shortages from Bangladesh to Nigeria by way of Vietnam and Zimbabwe. Nigeria illustrates the cruelty of the rankings: a crude exporter but a refined-products importer, it pockets the rent on one side and pays for the shortage on the other. On the other side of the vice, the winners collect. Gulf producers bank the war premium, Brazil sells its crude and its grain into a world short of both, and commodity exporters watch their terms of trade improve exactly as the importers' deteriorate. The dividing line no longer runs between "emerging" and "developed": it runs between those who sell molecules and those who buy them, a grid the global food shock makes even harsher for the poorest countries. The World Bank keeps a rather clement baseline, a rise limited to 2.5% for its food price index in 2026, but itself lists the risks that would derail it: a prolonged war, El Niño, export restrictions. The IMF's record window The system's safety net already bears the marks of the previous cycle. The IMF's credit outstanding passed SDR 110 billion in early 2026, about $150 billion, a historic record. Argentina alone concentrates $60.2 billion, more than 10% of its GDP; Egypt follows with $10.7 billion, Pakistan with $10.5; the top ten debtors, Ukraine included, add up to $128 billion. Put simply: the world's lender of last resort approaches a possible second round with a balance sheet already loaded with the survivors of the first, and an unprecedented concentration on a handful of giant programmes. This photograph reads two ways. The reassuring one: the Fund proved in 2022-2023 that it could chain programmes, and its current big debtors are precisely the ones under supervision, quarterly reviews included. The worrying one: every extra dollar on the barrel degrades the arithmetic of programmes calibrated on tamer oil, and the countries not yet at the window, frontier Africa first, would arrive in a world where the big drawers are already open. The paradox of 235 points Facing this machinery, the market displays a calm that raises questions. The benchmark emerging sovereign spread, the EMBI, tightened by 53 basis points in the second quarter to end at 235 points, not far from multi-year lows. The sequence says everything about the market regime: a 35-point widening in the first quarter when the war broke out, then a rally through May and June as the US-Iran ceasefire framework unwound the geopolitical premium. The problem is the date: those 235 points photograph the end of June, before the Kuwait attack and the return of threats around Hormuz relaunched the barrel by 13% in one week. Three explanations coexist, and telling them apart is the whole point. Complacency: the market extrapolates June's unwind and has not yet repriced the re-escalation. Differentiation: the indices aggregate winning exporters and losing importers, and the average hides an internal gap that is, itself, working. Genuine resilience, finally, which deserves its own section. The answer will be read in the coming weeks where the three meet: if Brent settles above $90 and the EMBI stays at 235, the anomaly will become hard to defend. The resilience school The other side has serious arguments, and it dominates asset management. Its thesis: the emerging markets of 2026 are not those of 2013, still less those of the 1990s. Their central banks tightened before the Fed and are being rewarded for it; their local-currency debt markets have deepened, shrinking the original sin; their reserves were rebuilt after 2022. Emerging debt outperformed other bond markets in 2025, carried by resilient exports and receding inflation, and strategists see strengths that endure, supported by inflows and reduced net supply. Even the BIS, hardly suspect of complacency, documents emerging monetary responses to external shocks that have become markedly more orthodox. Monday's picture in fact proves both camps right at once, and that is its singularity. No major currency of the South is in free fall: the Pakistani rupee holds within a handkerchief around 279, the Egyptian pound slides without snapping. The stress of 2026 is not a panic, it is an erosion: a few tenths of reserves a month, a few points of energy bill, a few basis points of refinancing. The history of emerging crises teaches that erosion can last a long time, then stop being slow all at once, at the first refinancing accident that comes along. The fault lines to watch The scenarios will be settled on a handful of dated dials, and scenarios they are. The barrel first: Brent durably above $90 turns erosion into haemorrhage for fragile importers; a relapse toward 75, as in late June, closes the vice without major damage. The Fed next, as early as 28-29 July: every notch of higher-for-longer transmits to the South's $4.3 trillion of dollar debt, and an actual hike would change the shock's category. The EMBI again: its reaction, or lack of one, to the mid-July re-escalation will say whether the 235 points reflected differentiation or complacency. The IMF reviews finally, Egypt and Pakistan first, the earliest to fold a wartime barrel into their programme arithmetic, and the food front if one of the risks listed by the World Bank materialises. The thread to keep runs beyond the quarter. The global financial system has spent four years testing the resistance of its rich links, American regional banks, British Gilts, French debt. The closing vice now tests the poor links, with a safety net already committed at record levels and markets that have not yet put a price on that eventuality. The precedents of 2022 were called Colombo, Cairo and Islamabad; the starting points were exactly today's. --- Primary sources: IMF, blog "How the War in the Middle East Is Affecting Energy, Trade, and Finance" (30 March 2026), April 2026 World Economic Outlook and credit outstanding (treasury data); BIS, international banking statistics and global liquidity indicators at end-December 2025 (dollar credit to EMDEs ~$4.3 trillion) and March 2026 Quarterly Review on emerging monetary responses; World Bank, "When risks stack up: Threats to global food markets in 2026". Market and analysis: State Street, EMD Commentary Q2 2026 (EMBI at 235 points) and Q1 2026; PineBridge, "2026 EMD Outlook"; MetLife IM, "Strengths Endure"; Convera, July 2026 FX Outlook. Press and data: Bloomberg, "EM Currencies Erase 2026 Gains" (1 July 2026) and the Egyptian IMF tranche (26 February 2026); CNBC (9 July 2026); Trading Economics, Brent and the Egyptian pound; Dawn, Pakistan's food import bill; Daily Pakistan, 18 July rates; Zee News and The Economy Pakistan for the IMF debt rankings; Wikipedia, economic impact of the 2026 Iran war. Figures and dates checked against the sources cited; spreads, currencies and prices move continuously, levels quoted are those of 17-20 July 2026. ============================================================================ ANALYSIS: The end of Bund scarcity URL: https://l0g.fr/en/analysis/the-end-of-bund-scarcity/ Canonical French source: https://l0g.fr/posts/fin-de-la-rarete-du-bund/ Date: 2026-07-19 (reviewed 2026-07-19) Topics: europe, bonds, rates, markets, liquidity ---------------------------------------------------------------------------- In early July, the Finanzagentur launched its new ten-year line, maturing August 2036, six billion euros to start, close to forty at completion. A routine issuance, except the routine itself is new: four syndications in the year, an unprecedented twenty-year line, a €511 billion programme. Germany, which spent a decade rationing its debt, is now industrialising it. And the market has already delivered its verdict, discreet but historic: the Bund's yield has moved above the equivalent swap rate, a first since these two curves have existed. Europe's most sought-after asset is turning into a bond like any other. The technical detail that sums it all up fits on a trading screen. On 17 July, the ten-year Bund yielded 3.14%, while the euro swap rate of the same maturity traded below 3%. This negative swap spread, the German state's yield above the fixed interbank rate, would have sounded absurd to any operator of the 2010s: the gap historically ran the other way, and its width measured the premium investors accepted to pay for holding the risk-free asset par excellence. The reversal, initiated in late 2024 and entrenched since, means one simple thing: the Bund's scarcity premium is gone. Our guide on interest rate swaps details the mechanics of this spread; here is the story of its death. Where the scarcity came from The Bund of the 2010s and early 2020s was not merely a bond: it was a rationed commodity. The rationing had two mutually reinforcing sources. On the supply side, the constitutional debt brake and the cult of the "schwarze Null", the balanced budget, kept net issuance close to nil: some years Germany repaid more than it borrowed. On the demand side, the Eurosystem vacuumed up the stock: the purchase programmes removed such a share of federal paper from the free float that Bundesbank research documented a repo "specialness" premium, where borrowing a specific Bund cost more than the cash it secured, including for bonds merely eligible for purchases and never actually bought. The German bond served as the ultimate collateral for Europe's entire plumbing, the machinery we describe in our analysis of the repo market and collateral, and that unfindable-asset status carried a price: yields crushed below swaps, dips under zero, auctions oversubscribed no matter what. That world rested on a precise political equilibrium: a state that refuses to borrow, a central bank that buys everything. Both pillars fell in under three years. The tap opens The first pillar gave way on 21 March 2025, when Germany reformed its debt brake to exempt defence spending above 1% of GDP and house €500 billion of infrastructure in a special fund, a turn we framed in our guide on European sovereign debt. The translation into paper is immediate: the Finanzagentur's 2026 issuance programme plans about €511.5 billion of securities against €362.4 billion of redemptions, a net supply in the region of €150 billion, a volume raised by a further €8 billion along the way. Total federal borrowing needs, special funds included, approach €174 billion, more than triple two years earlier, financing among other things a military budget now above €100 billion. The toolkit follows: two new ten-year lines in the year, four syndications including an unprecedented twenty-year line, €16 to 19 billion of green securities. The issuer that used to play hard to get has become an industrial producer of debt. The second pillar withdrew in silence: the ECB reinvests nothing anymore and its quantitative tightening returns about €500 billion of securities to the market each year. The price-insensitive buyer disappeared at the precise moment supply tripled. The OECD puts a number on the result: since mid-2022, sectors other than the central bank have had to absorb about €430 billion of additional German federal debt, taken up by investors far more price-sensitive, funds, insurers, households, foreigners. The consequence reads on the curve: long-dated European risk-free rates gained more than 40 basis points in 2025 alone, a steepening the ECB itself attributes to the combination of supply and central bank withdrawal. The chain of consequences A regime change in the reference asset never stays confined to that asset, and three shifts are already visible. The first concerns how sovereign spreads should be read: the celebrated European convergence, Italy back at 70 basis points, owes part of its existence to a moving denominator. When the Bund cheapens under the weight of its own supply, the gap with everyone else compresses without the periphery doing anything more; a portion of the advertised "normalisation" of spreads is in reality a banalisation of Germany. A reading grid to keep in mind when interpreting the tightening of the BTP-Bund and OAT-Bund spreads. The second touches the valuation reference. When the safest sovereign yields more than the swap, the swap curve, anchored to the €STR, becomes the de facto pricing yardstick: it is already the curve on which European Union bonds are priced, along with agencies and a growing share of credit, a path American markets travelled before Europe, Treasury swap spreads having been negative for years. The Bund remains the hedging instrument and the futures underlying, but its monopoly on the risk-free rate is now shared. The third is, for once, good plumbing news: with German collateral abundant again, specialness premia evaporate, quarter-ends strain less, and the European repo market breathes more easily than in the days when every borrowed Bund was a treasure. Scarcity carried a discreet systemic cost; its end is a dividend of the same order. The other reading: scarcity does not die, it sleeps The picture deserves its counterpoints, because burying the Bund's haven status would be premature. First, the yardstick of stress: in every episode of tension, from June's oil shock to French fever spikes, money still flees toward Germany, and redenomination risk, were it ever to awaken, would make the Bund the most sought-after asset on the continent within hours; the scarcity premium is cyclical, the German insurance policy is not. Second, the arithmetic: even at €174 billion of annual borrowing, Germany started from debt of about 63% of GDP, the lowest of the large advanced economies, and its absorption capacity remains unmatched; the market is charging for supply, not doubting the signature. Third, depth: a larger, more liquid pool serves benchmark status over the long run, as the US Treasury has demonstrated for decades, having survived its own swap spreads' move into negative territory without damage. The optimistic version of the same phenomenon reads like this: the Bund stops being a collector's item and becomes a genuine asset class, and Europe gains the deep bond anchor it lacked, a natural complement to the common debt pool under construction. Between the two readings, one point of agreement exists: the old regime is not coming back. Neither the fundamentalist version of the debt brake, buried by geopolitics, nor massive ECB purchases, with the balance sheet still shrinking, will make the Bund a rationed asset again on any foreseeable horizon. The arbiters What follows will be tracked on a handful of screens. The ten-year swap spread first: a deeper plunge would say absorption is still forcing the discount, a drift back toward zero that the market has digested the new volume. German auctions next, cover ratios and tails, read with the same grid as American auctions: the day a German syndication struggles, the regime will have shifted another degree. The slope of the curve again, the German ten-to-thirty segment pricing the compensation demanded for long duration once supply settles in. And Thursday's ECB meeting finally, since any further tightening stacks on top of this supply shift: the mix of rate hikes, QT and tripled German issuance is precisely the configuration our analysis of the 2011 remake flags for watching. Germany spent ten years proving that a state can be too lightly indebted for markets to function well. It will spend the coming decade testing the converse. --- Primary sources: Deutsche Finanzagentur, 2026 issuance outlook (18 December 2025) and Q3 update (25 June 2026): €511.5 billion of issuance, €362.4 billion of redemptions, new lines and syndications; ECB, blog "Sloping up: the repricing of euro area yields in 2025" (16 January 2026): long risk-free rates up more than 40 basis points in 2025; Bundesbank (research), "The Eurosystem's asset purchase programmes, securities lending and Bund specialness": specialness and eligibility premia in German repo; OECD, Global Debt Report 2026, investor base chapter: €430 billion absorbed outside the central bank since mid-2022. Market and analysis: Trading Economics, Bund at 3.14% on 17 July 2026; Blue Gamma, euro swap rates; TwentyFour AM on negative swap spreads; Amundi Research, "Swap Spreads: Analysis & Outlook"; Bloomberg on the borrowing programme increase (14 November 2025); Finance Unlocked on euro pricing conventions; Bruegel on the debt brake reform. Figures and dates checked against the sources cited; the swap spread moves continuously, levels quoted are those of mid-July 2026. ============================================================================ ANALYSIS: The ECB faces a 2011 remake: tightening into an oil shock URL: https://l0g.fr/en/analysis/ecb-2011-remake-oil-shock/ Canonical French source: https://l0g.fr/posts/bce-remake-2011-choc-petrolier/ Date: 2026-07-19 (reviewed 2026-07-19) Topics: europe, central banks, macro, rates, energy ---------------------------------------------------------------------------- On Thursday 23 July, the ECB's Governing Council meets for the first time since its 11 June rate hike, the first since 2023, decided in the name of inflation pressures born of the war in the Middle East. The market expects a pause, and keeps its eyes on September. A ghost will float through the room that nobody will name: 2011, the year the ECB raised rates twice into an overheating oil market, months before a crisis that nearly took the euro down. Fifteen years later, the ingredients are reassembling one by one. Whether the recipe produces the same dish remains to be seen. Thursday's central scenario is barely debated: about 88% of the market prices a hold at 2.25%, all the more so as July is a meeting without fresh staff projections, a poor vehicle for a change of course. The real battle is September: close to 70% of analysts expect another hike before year-end, which would take the deposit facility to 2.50%, against a backdrop of resurging energy costs. Thursday's stake is therefore not the move but the language: every word of the statement will be weighed against a question the ECB knows by heart, having settled it once in pain. Should you tighten into inflation imported by the barrel? 2011, the mistake that made the textbooks The precedent deserves telling with precision, because it structures the whole current debate. In the spring of 2011, euro area inflation runs above target, driven by oil pushed past $120 by the war in Libya. Jean-Claude Trichet, whose term ends that autumn, wants to lock in his legacy as guardian of prices: the ECB raises rates in April, then again on 7 July 2011, explicitly invoking the need to prevent second-round effects of the oil shock, the contagion from energy prices to wages and domestic prices. What followed belongs to monetary history. The euro area economy, already fragile, stalled; the summer of 2011 saw Italian and Spanish spreads blow out, forcing the ECB to reactivate its bond purchases through the SMP in emergency mode; and Trichet's successor, Mario Draghi, cancelled both hikes within his first weeks, in November and then December 2011, before arriving, the following year, at "whatever it takes". The retrospective diagnosis is unanimous, inside the institution included: tightening into an imported supply shock, in a monetary union with a fragile sovereign link, was a textbook error. It is, in large part, the founding trauma behind today's anti-fragmentation tools, including the Transmission Protection Instrument we describe in our guide on European sovereign debt. The barrel replays the scene, the thermometer hesitates The paradox of the moment sits in two crossing curves. On one side, official inflation is cooling: after May's peak at 3.2%, euro area prices fell back to 2.8% in June, with energy decelerating from 10.8% to 8.7% year on year and services from 3.5% to 3.2%. The June hike was thus followed, calendar irony, by a clearly softer inflation print, confirmed by Eurostat on 17 July. On the other side, the barrel replayed a full war cycle in three weeks. Having fallen below $75 in late June, back to pre-war levels, Brent regained more than 13% last week to close at $88 on Friday 17 July, a one-month high, after the attack attributed to Iran on a desalination plant in Kuwait and renewed threats around the Strait of Hormuz. Diesel refining margins hit an all-time record, and commercial inventories run about 6% below their seasonal average: a tight market where every incident is paid for in cash, as we documented in our analysis of strategic petroleum reserves and the second round. The September ECB will thus be looking at an oil price the June ECB had not yet seen, in one direction or the other. The resemblances that worry Set side by side, the two configurations share four traits. An imported supply shock first: in both cases, inflation comes neither from wages nor from domestic demand but from a wartime barrel, exactly the kind of price rise a policy rate cannot treat, except by crushing domestic demand. An identical rhetoric next: Trichet's "second-round effects" of 2011 are, word for word, the argument of the June 2026 statement on energy prices feeding into food, goods and services. A fragile sovereign link again: the periphery of 2011 was called Greece, Italy, Spain; the tension point of 2026 is called France, downgraded by four agencies in a year, with a spread around 80 basis points, and every turn of the screw mechanically inflates an interest bill already projected at €59 billion this year. A double brake finally, and this is worse than 2011: the rate increases come on top of a quantitative tightening still withdrawing about €500 billion of liquidity a year, a combination no textbook recommends during an external shock. The hawks' defence Fairness requires presenting the other side's case, and it is stronger than in 2011. First argument: the reference trauma has switched sides. The ECB of 2026 does not come out of a decade of over-reaction but out of the 2021-2022 episode, when it durably labelled "transitory" an inflation that ended in double digits; for the current council, the reputational risk is under-reacting, not the reverse. Second argument: the starting level. At a 2.25% deposit facility against 2.8% inflation, the real rate is still negative; this is far from restrictive tightening, and calling it monetary austerity stretches the language. Third argument: the architecture has changed. The TPI exists, born of the memory of 2011, excess liquidity still exceeds €2.3 trillion, banks are capitalised on another scale, and the sovereign contagion mechanism of 2011 has no automatic equivalent today. Fourth argument, finally: if the war in the Middle East settles in, the oil shock will not be a transitory spike but a durable shift in the level of energy prices, and anchoring expectations is then worth a pre-emptive tightening. Seen from Frankfurt, the June hike was "robust across a range of scenarios"; the Sintra turn, which we decoded in early July, said the same in softer tones. The weak point of the 2011 parallel deserves stating too: the economy of 2011 was already in creeping recession when Trichet tightened, while that of 2026, without shining, is so far absorbing the shock. Comparing calendars is no substitute for comparing conjunctures. The arbiters Neither the ghost of 2011 nor the hawks' confidence will settle this: a handful of data points will, and they all come with dates. The barrel first, the one variable Frankfurt does not control at all: Brent durably above $90 validates the persistent-shock reading and arms September; a relapse towards 75, as in late June, turns the June hike into an excess of caution. July inflation next, whose flash estimate lands on 1 August: if June's cooling is confirmed despite the oil, the second-round argument weakens. The OAT-Bund spread again, thermometer of the fragile link: its behaviour through the tightening will say whether the 2011 comparison is an analogy or a hyperbole. The September projections finally, the first full exercise integrating the new oil regime. Three scenarios emerge for what follows, to be read as scenarios and not forecasts. The durable pause: oil recedes, inflation converges, the June hike remains a one-way trip, and 2026 will have had only the shiver of 2011. The assumed tightening: the shock settles in, September takes rates to 2.50% and beyond, and the test of the French link becomes the real story of 2027, TPI on standby. The full remake: tightening, cyclical downturn, sovereign tensions, forced reversal, the 2011 sequence replayed with France in the lead role; it is the least likely scenario, the institution having been tooled up against precisely this, but its cost would dwarf the other two. On Thursday, the ECB will choose none of these yet. It will choose its words, and in 2011 too, everything began with words. --- Primary sources: ECB, monetary policy decision of 11 June 2026 (25 basis point hike, inflation scenarios) and press conference of 7 July 2011 (the 2011 hike, second-round effects); Eurostat, euro area inflation at 2.8% in June 2026 (17 July 2026) and flash estimate of 1 July. Market and analysis: Morningstar, "ECB Rate Decision: What to Expect on July 23" (hold priced at ~88%); Investing.com, ECB expectations (~70% for a hike by end-2026); MUFG Research, "Happy to hold, for now"; PitchBook on the 2011 hikes and their cancellation; Crux Investor on the oil war premium. Oil: Trading Economics, Brent at $88.10 on 17 July 2026; CNBC, Brent below $75 on 24 June; Al Jazeera on the return to pre-war levels in late June; EIA, Short-Term Energy Outlook. Figures and dates checked against the sources cited; market pricing moves continuously, the probabilities quoted are those of mid-July 2026. ============================================================================ ANALYSIS: Europe borrows without debt: the stack of common borrowing that appears in no ratio URL: https://l0g.fr/en/analysis/europe-borrows-without-debt/ Canonical French source: https://l0g.fr/posts/europe-s-endette-sans-dette-emprunts-communs/ Date: 2026-07-19 (reviewed 2026-07-19) Topics: europe, debt, bonds, macro ---------------------------------------------------------------------------- At the end of June 2026, the first disbursement of the European Union's €90 billion loan to Ukraine left for Kyiv. A few weeks earlier, in Brussels, the negotiation on how to repay the common debt already accumulated was sinking into deadlock. The two events tell the same story from both ends: the European Union borrows better and better, and still does not know who will pay. In between, over six years, a financial object the treaties never envisaged has taken shape: a de facto federal debt, invisible in every member state's accounts, whose outstandings will approach one trillion euros by year-end. Debt-to-GDP ratios drive Europe's fiscal news cycle: France at 116%, Italy at 137%, Brussels surveillance, rating actions. None of those figures contains one euro of the debt the Union issues itself. Yet that debt has changed scale: from a few tens of billions before 2020, the EU's outstanding bonds reached €827 billion by mid-2026, heading for about €1 trillion by the end of the year. By that yardstick, the Commission now borrows more than most member states, and its paper trades, sits in central bank reserves and gets compared to the Bund. The opacity mechanism is not concealment: everything is public, down to the semi-annual funding calendar. It is architectural, as so often in Europe: this debt has no assigned place in the accounting that structures the public debate. The stack: a method more than a plan The common debt stems from no founding decision. It built up by sedimentation, each layer voted as exceptional, temporary and capped. The SURE programme opened the way in 2020: €98.4 billion of loans to finance pandemic short-time work schemes. Then came the quantum leap, NextGenerationEU, the recovery plan: up to €806.9 billion in current prices, of which about €637 billion should be raised by the end of 2026. In 2024, the Union added its share of the G7's ERA loan to Ukraine, collateralised on the revenues of immobilised Russian assets. In May 2025, the Council adopted SAFE, €150 billion of defence loans, whose first commitments, €38 billion for eight states, were cleared in February 2026; Poland takes the lion's share with €43.7 billion. And at the end of 2025, the European summit settled Ukraine's financing for 2026-2027 through a €90 billion loan backed by the Union's budget, whose genesis we recounted in our investigation into Euroclear and the Russian assets. Five programmes, five emergency justifications, one signature. The method has an obvious political advantage: none of these texts is called a eurobond, none requires treaty change, and each can be presented to frugal electorates as a parenthesis. It also has a cumulative effect nobody formally decided: the Union now runs a permanent issuance programme, a yield curve, short-term bills, green bonds, and a global investor base. The exception has become an infrastructure. Why no ratio sees it The accounting mechanics deserve to be laid out, because there is nothing fraudulent about them: this is law. Debt issued by the Union is that of an international organisation, not of its members. In national accounts, a state's public debt captures the liabilities of its own administrations; the Commission's borrowings are not part of it, any more than those of the European Investment Bank. When Italy receives NGEU loans, only the on-lent fraction reappears in its national debt; the grants weigh on nobody in particular. And the €90 billion Ukraine loan rests on the budget's headroom: the gap between the ceiling of resources member states have committed to provide if needed and actual spending. A guarantee, not a debt; a contingent commitment, off every balance sheet. Every euro the Union borrows is therefore backed, ultimately, by the taxpayers of the 27, while appearing on none of their liability sides. Investors are not fooled: it is precisely this diffuse collective guarantee that earns the EU one of the continent's very best ratings. The paradox is the edifice's strength and its weakness: solidary enough to borrow in the Bund's neighbourhood, not enough to be called federal debt. Our guide on European sovereign debt places this quasi-safe asset within the wider architecture, a currency with no common debt to face it. A giant in the market, a dwarf in the indices The market, for its part, has settled half of the debate. The spread between EU bonds and the German Bund has narrowed from about 70 basis points in 2022 to around 40 on average in 2025: European paper now prices between Germany and France, as a top-tier issuer. Central banks, pension funds and sovereign wealth funds find in it the large pool of high-grade euro assets that was missing, to the point that the European Systemic Risk Board openly argues for expanding the supply of euro safe assets rather than restraining it. Recognition nonetheless stops at the index door. The major providers still classify EU bonds as supranational debt, alongside the EIB and KfW, not as sovereign: consultations were held, the status quo prevailed. The nuance sounds technical; it is heavy. Inclusion in sovereign indices would trigger tens of billions of passive buying, futures contracts, benchmark liquidity. Its absence sustains a structural discount. The index providers' reasoning is, at bottom, the same as the lawyers': a temporary, capped debt without unlimited joint and several guarantee is not a sovereign. The Union finds itself in an in-between of its own making: too big to be a niche issuer, too ambiguous to be a benchmark. 2028, the due date without a payer The calendar turns this ambiguity into a countdown. Repayment of NGEU principal starts in 2028 and stretches to 2058: about €13.9 billion of principal a year, plus interest peaking around €10.8 billion in 2030, an annual charge in the €25 billion range, and a total debt service estimated between €582 and €715 billion through 2058 according to Bruegel. The Commission built the constraint into its proposal for the 2028-2034 budget framework782646EN.pdf): out of nearly €2 trillion of budget over seven years, €168 billion is reserved for repaying NGEU. One-twelfth of the European budget, absorbed by yesterday's debt before financing anything of tomorrow. The recipe remained to be found. The new own resources promised as early as 2020, a carbon border levy, a share of the carbon market, a corporate contribution, were meant to yield about €58.5 billion a year on the Commission's estimates. Six years on, none has been adopted: the decision requires unanimity plus ratification in every member state. In the spring of 2026, the Council presidency noted that every avenue, higher national contributions, rescheduling the debt, new resources, appeared deadlocked. Arithmetic will not wait: without a dedicated revenue, repayment will be paid for in cuts to common policies or in higher national contributions, which is to say, either way, in political conflict. The other reading: an exemplary debt Honesty requires turning the argument around, because the hidden-debt indictment has its limits. Nothing is less concealed than the Union's borrowing: semi-annual funding plans published in advance, outstandings detailed bond by bond, parliamentary hearings, public ratings. Next to Europe's real blind spots, national guarantees, unfunded pensions, off-balance-sheet vehicles, the common debt is probably the best-documented public liability on the continent. Its size remains modest against national liabilities: a trillion euros is less than a third of France's negotiable debt alone, and about 6% of the Union's GDP, against the euro area's 88% average public debt. The deeper reproach also deserves its counterpoint: if this debt enters no national ratio, it is because it commits no state individually, and that is precisely its purpose. The advocates of a genuine European safe asset, from the European Systemic Risk Board to the economists who see it as the condition for a reserve-currency euro, are not asking for less common debt but for more, owned, permanent and properly institutionalised. In that reading, the problem is not the stacking of programmes: it is the refusal to draw the consequences, by giving the issuer a revenue of its own and a clear status. The opacity denounced here is not that of the figures, available to the cent; it is that of the political unsaid surrounding them. Three exits, one unsaid From here, three trajectories emerge, and these are scenarios, not forecasts. The first is silent rescheduling: refinancing redemptions rather than repaying them, as the rollover issuance planned until 2058 already allows, and as the capitals keen to spread the bill quietly suggest. Taken to its end, this path makes the common debt perpetual in fact, without any parliament ever voting on the principle. The second is the institutional leap: real own resources, credible repayment, and eventually the sovereign reclassification the indices refuse today. It is the most coherent path and the least likely near term, unanimity standing in the way. The third is the purge: repaying out of the existing budget, sacrificing common policies, at the risk of turning yesterday's debt into the enemy of tomorrow's priorities, defence included. The signals to watch are concrete: the endgame of the 2028-2034 framework negotiation, due before the end of 2027; any decision on own resources, each one made significant by the very unanimity that makes it improbable; the classification reviews of the major bond indices; and the EU-Bund spread, the best arbiter of this paper's real status. Meanwhile, the machine keeps running: the Commission will raise more than €150 billion again this year, the Ukraine loan is disbursing, SAFE is arming Polish and Romanian orders. Europe has invented a borrower without a state, backed by everyone and booked by no one, and it lives with it very well, as long as nobody demands to know who, in the end, signs the 2028 cheque. --- Primary sources: European Commission, EU as a borrower and NextGenerationEU (outstandings, ceilings, funding calendar, €637 billion expected raised by end-2026), next generation of own resources; Council of the EU, adoption of SAFE (27 May 2025) and finalisation of the €90 billion loan to Ukraine (23 April 2026); Commission (DG DEFIS), first wave of SAFE funding (January-February 2026); European Parliament, EPRS briefing on the 2028-2034 financial framework782646EN.pdf) (€1,984 billion, including €168 billion of NGEU repayment); ESRB/ECB, "Expanding the supply of euro safe assets" (22 April 2026). Analysis: Bruegel, "What will it cost the European Union to pay its economic recovery debt?" (€13.9 billion of principal a year, interest, total service of €582-715 billion); Morningstar, "What Are EU Bonds and Can They Become a Safe-Haven Powerhouse?" (outstandings, spread to Bund, index classification); OMFIF, "Outlook 2026"; Jacques Delors Institute on the budget package. Press: Agence Europe, on the financing deadlock (May 2026); Euronews on the SAFE timeline. Figures and dates checked against the sources cited; outstandings evolve with each issuance, orders of magnitude are as of mid-July 2026. ============================================================================ ANALYSIS: Euroclear: the opaque vault at the heart of Europe URL: https://l0g.fr/en/analysis/euroclear-the-opaque-vault-of-europe/ Canonical French source: https://l0g.fr/posts/euroclear-chambre-forte-opaque-avoirs-russes/ Date: 2026-07-18 (reviewed 2026-07-18) Topics: europe, geopolitics, sanctions, central banks, systemic risk, markets ---------------------------------------------------------------------------- On 17 July 2026, a Russian appeals court upheld a ruling ordering Euroclear to pay more than 18 trillion roubles, over €200 billion, to the Central Bank of Russia. The judgment is unenforceable outside Russia, and the Brussels depository does not recognise the court's jurisdiction. But it captures the essence of the moment: Europe's largest securities infrastructure, a house the general public has never heard of, has become the epicentre of a €200 billion legal, financial and geopolitical standoff. To inaugurate our European coverage, a deep dive into a vault whose opacity is not an accident but an architecture. A quarter of a century ago, a book about a Luxembourg securities depository triggered one of the longest judicial sagas in the history of the French press. In 2026, no investigation is needed to find out where the frozen reserves of the Central Bank of Russia sleep: they sit in Brussels, at Euroclear, their amount is disclosed in quarterly press releases, and their fate keeps foreign ministries busy. Yesterday's secret is now published at press conferences, without having lost any of its explosive charge. Forty-four trillion under custody Start with the scale, because it conditions everything else. Euroclear is not a bank in the everyday sense: it is a central securities depository (CSD), the layer of the financial system where bonds, equities and funds are registered, safekept and delivered against payment. At the end of March 2026, the group held close to €44 trillion in assets under custody for its clients, and the value of transactions settled in its books reached €1,390 trillion over the year 2025, roughly twelve times world GDP. When a Dutch pension fund buys a French government bond or a Japanese bank delivers a eurobond, odds are the trade settles inside this machine. History explains the geography. Euroclear was born in Brussels in 1968, inside the local branch of the American bank Morgan Guaranty, to settle eurobonds, those securities issued outside any national jurisdiction, whose market was then exploding. Two years later, rival banks created Cedel in Luxembourg, which would become Clearstream. The two houses, competitors to this day, share the same function and the same discretion: they are international central securities depositories (ICSDs), the crossroads of the global bond market. Their regulation has thickened considerably since, with the European CSD Regulation of 2014 (Regulation (EU) 909/2014) and, for Euroclear Bank, supervision by the National Bank of Belgium. An infrastructure of this size does not live in a regulatory grey zone. Its opacity lies elsewhere. Opacity by design The central mechanism is the omnibus account. A depository like Euroclear does not open accounts for savers or companies: its clients are banks and custodians, which pool the securities of their own clients into collective accounts. Each link in the chain knows only the adjacent link. Euroclear sees the custodian bank, the bank sees the asset manager, the manager sees the fund, and the final beneficiary appears nowhere in the books at the top. This nesting-doll arrangement was not designed to hide anything: it mutualises costs, enables netting, and makes it possible to settle hundreds of thousands of transactions a day. Without omnibus accounts, there is no global bond market at this price point. The trade-off is structural: economic ownership becomes invisible to the very infrastructure that safekeeps it. Hence the entire complexity of the Russian file. Of the roughly €210 billion of Central Bank of Russia assets immobilised in the EU779267), the bulk is lodged at Euroclear: €183 billion at the end of 2024 according to the group's own accounts, an amount that moved above €200 billion in the first quarter of 2026 as securities matured and turned into cash. Untangling what belongs to the central bank, to sanctioned entities or to ordinary non-sanctioned Russian investors means climbing back up custody chains that the system was never designed to expose. Révélation$, the archaeology of suspicion This architecture has a polemical history, and it is as French as it is Luxembourgish. In February 2001, journalist Denis Robert and former Cedel executive Ernest Backes published Révélation$, an investigation accusing Clearstream, Euroclear's Luxembourg rival, of having maintained a system of unpublished accounts that could serve tax evasion and money-laundering circuits. A decade of defamation trials followed, dozens of proceedings, and finally, on 3 February 2011, three rulings by the French Court of Cassation vindicated the journalist in the name of the public-interest debate and the seriousness of his investigation, on the basis of Article 10 of the European Convention on Human Rights. Two points of fairness are in order. First, the so-called Clearstream 2 affair, the fake account listings used to smear French political figures, has nothing to do with the original documentary work: it durably blurred the public perception of the case. Second, Euroclear is not Clearstream, and nothing in the 2001 investigation targeted the Brussels house. If we summon this history, it is for a precise reason: the mechanism questioned at the time, chains of accounts that no regulator or journalist could unwind end to end, is the same one that today makes the Russian assets both possible to immobilise and so hard to seize cleanly. In twenty-five years, the question has simply changed hands: it is no longer put by a journalist to an infrastructure, but by states to their own financial system. The 2022 freeze, an accidental rent On 28 February 2022, four days after the invasion of Ukraine, the European Union prohibited all transactions with the Central Bank of Russia. The reserves Moscow had parked in European securities, mostly safekept at Euroclear, were immobilised: neither confiscated nor returned. Then market mechanics did their work. Bonds matured, coupons fell due, and all that cash piled up in the books of Euroclear Bank, which redeposited it with central banks. At the end of March 2026, Euroclear Bank's balance sheet stood at €237 billion, of which €200 billion related to sanctioned Russian assets. A settlement infrastructure found itself, despite itself, with a balance sheet that is six-sevenths Russian. That cash earns money. Interest on the Russian assets reached €4.4 billion in 2023, €6.9 billion in 2024, then €5 billion in 2025, down 26% as rates eased, and another €2.3 billion in the first half of 2026. In total, about €17 billion of interest between 2022 and 2025, which along the way generated more than €4 billion in tax for the Belgian state. Europe organised the capture of this rent without touching the principal. In February 2024, the Council required depositories to ring-fence these extraordinary revenues; in May 2024, the Council earmarked most of them for Ukraine, under the name of windfall contribution. Euroclear paid a first instalment of €1.5 billion in late July 2024, then €1.6 billion in July 2025, with a further €1.4 billion announced for 2026, bringing the total to €6.6 billion. These flows also service the $50 billion loan extended by the G7 in June 2024, the ERA, collateralised on the future revenues of the immobilised assets. The legal boundary is constantly restated by the institutions: under the European reading, the interest belongs to no one, while the principal remains Russian. On this ridge line, Europe is financing a war with the proceeds of a freeze, without ever uttering the word confiscation. The reparations loan, chronicle of a deadlock In the autumn of 2025, the Commission decided to change scale. Its idea, the reparations loan: mobilise not the interest but the cash balances themselves, lending them to Ukraine through a structure in which Kyiv repays only if Russia one day pays war reparations. The proposal formalised in December 2025 amounts to €165 billion of new support, €115 billion for defence and €50 billion for Ukraine's budget, plus €45 billion to repay the G7's ERA, for a total of €210 billion mobilised779267EN.pdf). Seven member states, from Finland to Poland, signed a letter of support, while the Commission, which also tabled a classic joint-borrowing option, openly pushed the asset-backed loan. The blockage came from the host country. Belgian Prime Minister Bart De Wever demanded political, legal and financial guarantees: were the freeze ever lifted, or were Russia to win an arbitration, Belgium and Euroclear would face a €200 billion claim alone. The Commission proposed a system of national guarantees distributed by gross national income, but several capitals balked, and the commitments remained, in Brussels-the-Belgian's eyes, insufficiently binding. At the same moment, the 28-point American peace plan of November 2025 crashed into the scheme: it envisaged investing $100 billion of frozen Russian assets in a reconstruction vehicle led by Washington, with half the profits going to the United States, the remainder feeding a joint US-Russian investment vehicle. For the Europeans, who discovered the text without having been consulted, it was confirmation that their main negotiating lever could slip from their hands. The European Council of 18 December 2025 produced a scaled-back compromise: €90 billion of support for 2026-2027, financed by EU borrowing backed by the headroom of the European budget, with no recourse to the Russian assets. The text, backed by 25 heads of state and government, accommodates Czechia, Hungary and Slovakia through enhanced cooperation, explicitly reserves the right to use the Russian assets to repay the loan, and asks that technical and legal work on the reparations loan continue. A week earlier, a quieter move had locked the mechanism in place: on 12 December, the Council adopted Regulation 2025/2600, based on Article 122 of the Treaty, prohibiting any transfer of Central Bank of Russia assets back to Moscow and making contrary Russian judgments unenforceable in the Union. The immobilisation, until then hostage to the unanimous renewal of sanctions every six months, became de facto indefinite, and adopted by qualified majority. Some jurists see a bold innovation, others a bypass of the common foreign policy's own rules; both readings can be true at once. The framework of the €90 billion loan was settled by the Council on 4 February 2026, finalised on 23 April, and the June summit endorsed a first disbursement before the end of June 2026. The war of the courts The Russian response unfolded on every front at once, and 2026 turned it into a judicial war of attrition. In December 2025, the Central Bank of Russia filed a claim before a Moscow arbitration court to recover its assets, and announced it would seek damages from European lenders. On 27 February 2026, it challenged the immobilisation regulation before the General Court of the European Union, invoking the sovereign immunity of its assets and the contestable choice of qualified-majority voting. On 15 May, the Moscow court ordered Euroclear to pay more than 18 trillion roubles in damages, around $250 billion. Euroclear, which faces more than a hundred proceedings in Russia, does not recognise the court's jurisdiction and counter-attacked on 30 June before the Brussels enterprise court, asking a European judge to declare the Russian claim groundless. On 17 July, the Russian judiciary rejected Euroclear's appeal and upheld the ruling. None of these Russian judgments is enforceable in Europe, and Regulation 2025/2600 was written precisely so that they never become so. Their function lies elsewhere: to build a legal claim that can be executed against Western assets still present in Russia, where the holdings of Euroclear clients trapped in type-C accounts run into the billions, and to weigh on any future peace negotiation. Each side is methodically manufacturing its own legal reality, in mirror image. The outcome will most likely be decided neither in Moscow nor in Brussels, but by the balance of power that determines which of the two legal orders applies to the other side's assets. The precedent and the currency There remains the question that goes beyond Euroclear: what is a reserve currency worth if its assets can be withheld indefinitely by political decision? The European Central Bank has held a consistent line: no lender-of-last-resort role in any structure resembling monetary financing, which the treaties prohibit, and a demand for legal clarity that Christine Lagarde repeated as late as her press conference of 18 December 2025, since confidence in the euro is also played out on this ground. The Commission itself long rejected outright confiscation, on the grounds that it would violate sovereign immunity and expose the Union to retaliation. The underlying argument is familiar: if reserves held in euros become seizable, third-country central banks, from Beijing to Riyadh, will reallocate part of their holdings towards gold, whose record accumulation by central banks we have documented, or towards jurisdictions perceived as neutral. The de-dollarisation debate has shown how slow and often overstated such shifts are; it has also shown that they accelerate precisely after shocks to confidence. The most concrete risk is nonetheless more prosaic, and it lives in the plumbing. Euroclear is a systemic counterparty: a CET1 capital ratio of around 57% in the first quarter of 2026 makes it one of the best-capitalised institutions in Europe, but its balance sheet remains 85% immobilised Russian assets. Confiscating the principal would turn a depository position, neutral by construction, into a €200 billion debt to a hostile creditor, with a precedent invocable against any central bank client. The group itself has publicly warned the Union about the consequences of using the assets: loss of confidence among international investors, cascading litigation, and the weakening of an infrastructure on which the settlement of European debt depends, the same collateral plumbing whose mechanics and breaking points we have described on the American side. Overblown risks? The case made by the loan's supporters Fairness requires laying out the opposite reading, championed by part of Europe's economists and lawyers. For the Centre for European Reform, the legal risks of the reparations loan are largely ill-founded: Russia has already lost the Uniper v. Gazprom arbitration in 2024 and Krymenergo v. Russia in 2025, it refuses to appear before the very fora it invokes, and the December 2025 regulation neutralises the enforcement of its judgments inside the Union. Liquidity buffers exist, the authors add: force majeure clauses, regulatory grace periods, and, as a last resort, emergency liquidity assistance from central banks. Christine Lagarde herself judged the latest version of the scheme legally more solid than its predecessors, a notable shift in the ECB's position. As for the precedent for the euro, the loan's supporters point out that the freeze has lasted four years without any measurable flight of official reserves out of the euro area, and that the alternative, making the European taxpayer pay rather than the aggressor, creates a risk of its own, a political one. This debate is not settled, and our editorial protocol requires saying so: nobody has a solid empirical basis for quantifying the effect of a confiscation on the euro's reserve status, because the event would be unprecedented at this scale. Both camps reason in scenarios, not in data. Three paths for a vault As we publish, three paths remain open, and these are scenarios, not forecasts. The first is the rentier status quo: indefinite immobilisation keeps producing a few billion in interest a year, shrinking as rates fall, which Europe captures without touching the principal. This is the default trajectory, the one set by Regulation 2025/2600 and the €90 billion loan. The second is the revival of the reparations loan: the December conclusions demand in black and white that technical and legal work on the instrument continue, and the file will return if Ukraine's needs overflow the 2026-2027 envelope or if the guarantees demanded by Belgium are finally assembled. The third is a negotiated settlement of the conflict in which the assets become a bargaining chip, along the lines sketched by the American plan of November 2025; paradoxically, this would be the most destabilising scenario for Europe, which would watch the lever change hands. The signals to watch are identifiable: Euroclear's quarterly results, which every three months photograph the Russian balance sheet and the rent; the EU General Court's ruling on the Central Bank of Russia's challenge, the first judicial test of the immobilisation regulation; the Brussels proceedings opened on 30 June; and every European Council, where the question of financing Ukraine beyond 2027 will mechanically return. An infrastructure designed for the shadows now spends its quarters in the spotlight, with a European regulation written for it, a $250 billion Russian judgment bearing its name and foreign ministries tracking its cash balances release after release. Twenty-five years after Révélation$, nobody is searching for the money anymore: it is located to the nearest billion, in public quarterly reports. The fight is now over who gets the right to use it. --- Primary sources: Euroclear, results press releases for 2023, 2024, 2025 and Q1 2026, and the update on Russian sanctioned assets (May 2026); Council of the EU: ring-fencing of extraordinary revenues (12 February 2024), earmarking of net profits for Ukraine (21 May 2024), Regulation (EU) 2025/2600 of 12 December 2025, European Council conclusions of 18 December 2025, position of 4 February and finalisation of 23 April 2026 of the €90 billion loan, European Council of 18-19 June 2026; European Commission: two financing options (December 2025), first transfer of €1.5 billion (July 2024), €1.4 billion transfer (2026); ECB, press conference of 18 December 2025; European Parliament, EPRS briefing "Financing Ukraine in 2026 and 2027"779267); Regulation (EU) 909/2014 (CSDR), EUR-Lex. Analysis: Centre for European Reform (Tordoir and Paduano, 18 December 2025); CEPR VoxEU; Just Security; Institut Jacques Delors; Lawfare. Press: La Libre (17 July 2026) and 15 May 2026; L'Avenir (30 June 2026); RTBF; The Moscow Times (16 May 2026) and 18 December 2025; Euronews (3 March 2026) and 21 November 2025; Axios, full text of the 28-point plan; Euractiv. On the history of the Clearstream affair: Wikipedia and Ouvertures.net on the Court of Cassation rulings of 3 February 2011 (in French). Figures and dates checked against the sources cited; rouble and dollar amounts are orders of magnitude converted at current rates. ============================================================================ ANALYSIS: The chip relapse: a bear market on record profits URL: https://l0g.fr/en/analysis/chip-relapse-bear-market-record-profits/ Canonical French source: https://l0g.fr/posts/rechute-des-puces-bear-market-profits-records/ Date: 2026-07-17 (reviewed 2026-07-17) Topics: ai, markets, semiconductors, macro, valuations ---------------------------------------------------------------------------- In Friday trading, the Philadelphia semiconductor index shed as much as 5.7%, taking its fall past 20% from its late-June record, the technical threshold of a bear market, after a 105% surge between its March low and its peak. The day before, TSMC, the AI world's factory, had posted record quarterly revenue of $40.2 billion, up 33.7% year on year, and earnings per share up 77.4%, its eighth consecutive quarter above expectations. The stock fell anyway. That is this correction's singularity, and the reason to dwell on it: profits are not faltering, estimates are rising, and the market is nonetheless paying less and less for each dollar of earnings. What is deflating is not the E in the P/E, it is the P: a derating, in the strict sense. Records, sold The Nvidia case sums up the moment. The stock has lost about $1 trillion of market value in under two months, down 16% from its 14 May record, reached days after becoming the first company in history to cross $5.5 trillion. Over the same stretch, according to data compiled by Bloomberg, analysts raised their profit estimates by 13% in three months, and the consensus projection for fiscal 2027 stands at $228 billion of profit on $393 billion of sales. The mechanical result: Nvidia now trades at 18 times expected earnings, its lowest multiple since early 2019, that is, since before the AI boom. More striking still: that is less than the S&P 500, above 20 times, and the Nasdaq 100, at nearly 23 times. The emblematic stock of the AI revolution trades at a discount to the index it carried. At TSMC, the same grammar. The 16 July results beat the top of the company's own guidance, and the sanction came anyway: the stock fell, and the selling spread to Asia. Strategist Andrew Jackson of Ortus Advisors summed up the market's verdict in the morning press: results judged not strong enough to justify another leg higher, and growing concerns over the sector's excessive spending. When beating expectations no longer lifts a stock, the problem is not in the accounts: it is in the price. The catalysts of the derating The correction has not one trigger but four, spread over three weeks, and their nature reveals this market's particular fragility. The first carries the most meaning. On 1 July, Reuters and then Bloomberg revealed that Meta is preparing "Meta Compute", a cloud business designed to sell its surplus AI computing capacity to outside customers, raw or packaged with its Llama models. The decisive word is surplus: Meta is spending $125 to $145 billion on AI infrastructure in 2026, against about $72 billion last year, and has just expanded its CoreWeave contract by roughly $21 billion, while building enough to compete with that same CoreWeave and the big clouds. A hyperscaler looking to resell its surplus validates, from the inside, the hypothesis the sector feared most: overcapacity. The second catalyst touches demand. On 16 July, Alphabet lost more than 4% after Bloomberg reported the delay of Gemini 3.5 Pro, its most advanced model. If model roadmaps slip, the chip orders that serve them slip too. The third is positioning: the rally's most speculative names, Marvell, ARM and Intel, have lost more than 30% from the peak, the classic mechanics of a saturated trade unwinding. The fourth is macro-financial, and we dissected it as it happened: Kevin Warsh's first FOMC moved the median rate projection from 3.4% to 3.8% for end-2026, with nine members now projecting at least one hike and the 2026 PCE inflation forecast raised from 2.7% to 3.6% under the oil shock. Dallas Fed president Lorie Logan argued on 16 July for further hikes. Tech valuations are long-duration assets: when the discount rate climbs, the most stretched multiples compress first. The link between the Hormuz blockade, imported inflation and Nvidia's P/E is indirect, but it is real. Meta Compute, the detail that changes the thesis Among these catalysts, Meta's surplus sale deserves a freeze-frame, because it threatens three floors of the AI financial scaffolding we have been documenting for months. The first floor is the scarcity narrative: the sector's entire valuation premium rests on the idea that compute is scarce and will remain so. A surplus put up for sale by one of the world's largest buyers says otherwise, at least at the margin, and markets are set at the margin. The second floor is the price of compute: if Meta discounts its excess capacity, neocloud pricing and the revenue assumptions backing their debt compress with it. The third floor is the most fragile: the residual value of GPUs, on which a growing share of AI infrastructure credit rests. A secondary market for compute fed by hyperscaler surpluses is exactly the scenario that tests those guarantees. Add a structural irony that readers of our work on AI's circular financing will recognise: Meta expands its CoreWeave contract by $21 billion while preparing to become its direct competitor. The same capex dollar feeds the supplier's order book and the future supply that will weigh on its prices. This is not fraud, it is a loop; but loops amplify in both directions. Rotation is not exit Reading this correction as the end of the AI bet would nonetheless be a misreading, and this is where the picture gets interesting. At the very moment logic chips deflate, memory is on fire. Micron has gained 229% in 2026 after 239% in 2025; SK Hynix, up about 248% this year, has joined Micron in the $1 trillion market cap club, with Samsung gaining some 165%. The engine is physical: high-bandwidth memory (HBM), the bottleneck of inference, is sold in advance, with Micron's capacity booked through 2027, and Bank of America projects the HBM market growing from $34.6 billion to $54.6 billion in 2026. Capital is not leaving AI: it is migrating from compute to memory, from Nvidia to Micron and SK Hynix, from the segment that got expensive to the segment still rationed. That migration has its own fragility, and it is historical: memory is the most cyclical corner of the entire industry, the land of repeated booms and busts, as fund managers were warning back in May. Investment plans announced at the top of the cycle, hundreds of billions of dollars of new fabs on the Korean and American side, replay the pattern that has always turned memory shortage into memory glut. The crowd is not leaving the theatre: it is moving to the part of the room where the floor has given way most often. The benign reading There is a benign interpretation of everything above, and it is defensible. A derating that happens while earnings rise is the least painful way to deflate a valuation excess: time and growth do the work a crash would otherwise do. Nvidia at 18 times earnings projected to grow 90%, cheaper than the S&P 500, is a statistical anomaly; of 82 analysts, 78 remain at buy with an average target 50% above the price. If the projected profits materialise, this correction will enter the textbooks as an entry point. The counter-argument, which we developed in the bubble within the bubble, fits in one question: what is the E worth on which this reasonable multiple is computed? Part of the sector's revenue is fed by circular flows between players financing one another, and by hyperscaler capex that Meta has just shown exceeds its own needs. A P/E of 18 on peak-cycle, partly self-fed earnings is not necessarily cheap; it is in fact the classic configuration of valuation traps in cyclical industries. The market is not choosing between these two readings, it is pricing both: hence a sector worth a fifth less than in June on record accounts. The gap between the two readings will close on numbers, not on narratives. Markers for what comes next Five appointments will separate a healthy correction from a turn. Hyperscaler capex in the late-July earnings round: a downward revision, even cosmetic, would change the episode's nature, turning Meta Compute from an isolated signal into the start of a series. Nvidia's results in late August, the first test of the $228 billion FY2027 profit consensus. TSMC's monthly revenue, the fastest thermometer of real demand. Memory prices and 2027 HBM contracts, which will say whether the rotation rests on durable scarcity or on the umpteenth top of a cycle. And the spreads on the debt financing data centres, because that is where contagion would show, more than in equities: stocks absorb multiple compression, credit only absorbs defaults. As long as the latter do not follow the former, the chip relapse remains a story about price. The day the E joins the P, the story changes genre. Sources - Bloomberg, "Chips Stocks Sink Into Bear Market as 105% AI Rally Fizzles", 17 July 2026 (SOX -5.7% intraday, more than -20% from the late-June record, +105% March to June, Marvell/ARM/Intel -30%): https://www.bloomberg.com/news/articles/2026-07-17/chips-stocks-tumble-into-bear-market-as-105-ai-rally-fizzles - TSMC, second quarter 2026 results, 6-K of 16 July 2026 (revenue of $40.20bn, +33.7% year on year, EPS +77.4%): https://www.sec.gov/Archives/edgar/data/0001046179/000104617926000451/a2q26ewithguidancexfinal.htm - Bloomberg (via Yahoo Finance), "Nvidia's $1 Trillion Slide Sends Valuation to Pre-AI Boom Levels", 8 July 2026 (-16% since 14 May, 18x forward earnings versus over 20x for the S&P 500 and nearly 23x for the Nasdaq 100, estimates +13% in three months, FY2027 consensus, YTD performance, Micron +229%): https://finance.yahoo.com/markets/stocks/articles/nvidia-1-trillion-slide-sends-084308296.html - Forbes, "Nvidia Hits Record $5.5 Trillion Value", 13 May 2026: https://www.forbes.com/sites/antoniopequenoiv/2026/05/13/nvidia-hits-record-55-trillion-value-first-company-to-ever-reach-mark/ - CNBC Daily Open, 17 July 2026 (Asian contagion, Andrew Jackson quote, Ortus Advisors): https://www.cnbc.com/2026/07/17/cnbc-daily-open-trump-electoral-system-fraud-china.html - Bloomberg, "Meta Is Planning a Cloud Business to Sell AI Computing Power", 1 July 2026: https://www.bloomberg.com/news/articles/2026-07-01/meta-is-building-a-cloud-business-to-sell-excess-ai-compute - TechCrunch, "Meta, like SpaceX, looks to turn excess AI compute into cash", 1 July 2026 (2026 capex of $125-145bn versus ~$72bn in 2025): https://techcrunch.com/2026/07/01/meta-like-spacex-looks-to-turn-excess-ai-compute-into-cash/ - CoreWeave, 8-K (contract expansion with Meta of roughly $21bn): https://www.sec.gov/Archives/edgar/data/1769628/000176962826000154/ex991.htm - BigGo Finance, "Philadelphia Semiconductor Index Plunges Into Bear Market; Google Slumps Over 4% on AI Delay", 16-17 July 2026 (Gemini 3.5 Pro delay reported by Bloomberg, Lorie Logan remarks): https://finance.biggo.com/news/d7a2304a-c07f-43fa-8268-a258ba731152 - CNBC, "Fed holds interest rates steady", 17 June 2026 (2026 median raised from 3.4% to 3.8%, nine members projecting at least one hike, 2026 PCE raised to 3.6%): https://www.cnbc.com/2026/06/17/fed-interest-rate-decision-june-2026.html - Federal Reserve, Summary of Economic Projections, 17 June 2026: https://www.federalreserve.gov/monetarypolicy/files/fomcprojtabl20260617.pdf - Yahoo Finance, "SK Hynix joins Micron in $1 trillion club as AI memory chip rally accelerates" (SK Hynix ~+248% YTD, Samsung ~+165%): https://finance.yahoo.com/markets/stocks/article/sk-hynix-joins-micron-in-1-trillion-club-as-ai-memory-chip-rally-accelerates-024514610.html - IG, "Memory chip supercycle 2026" (HBM market from $34.6bn in 2025 to a projected $54.6bn in 2026, BofA estimates): https://www.ig.com/en/news-and-trade-ideas/memory-chip-stocks-rally-2026-260708 - The Motley Fool, "AI Data Centers Will Consume 70% of All Memory Chips in 2026" (HBM capacity booked, Micron sold out through 2027): https://www.fool.com/investing/2026/06/25/ai-data-centers-will-consume-70-of-all-memory-chip/ - CNBC, "Beware the boom and bust cycle of memory stocks", 25 May 2026: https://www.cnbc.com/2026/05/25/memory-stocks-cyclical-boom-bust-samsung-sk-hynix.html This article is journalistic analysis and does not constitute investment advice. The 17 July price moves are intraday data, subject to change by the close; the cited earnings projections are consensus estimates, not facts. Data as of the dates of the cited sources. ============================================================================ ANALYSIS: Truth API: the presidential word becomes a paid market data feed URL: https://l0g.fr/en/analysis/truth-api-presidential-speech-market-data-feed/ Canonical French source: https://l0g.fr/posts/truth-api-parole-presidentielle-flux-de-marche/ Date: 2026-07-17 (reviewed 2026-07-17) Topics: markets, us politics, regulation, conflicts of interest, microstructure ---------------------------------------------------------------------------- On 16 July, two pieces of news dropped a few hours apart, and their juxtaposition says something neither says alone. In the morning, the White House confirmed the suspension of the president's teleprompter operator, suspected of betting on the content of Trump's speeches via the prediction market platform Kalshi. In the afternoon, Trump Media & Technology Group announced Truth API, a paid data feed promising trading firms the fastest access to posts from Truth Social's ten most influential accounts, starting with the president's. In both cases the underlying is the same: the presidential word moves prices, therefore it is worth money. The difference is the status of whoever collects. The technician faces a settlement with the CFTC; the president's company, roughly 41% owned by him, turns it into a recurring revenue line. Milliseconds of head start on the president The product is described without embarrassment by its seller. Truth API, available on 1 August to institutional customers, will deliver posts from the platform's ten most influential accounts in "milliseconds", with an archive back to 2022, aimed at firms "most impacted by the cost of a delay in information", meaning algorithmic and high-frequency trading, for whom the latency race is the core business. "Markets already move on Truth Social posts," interim CEO Kevin McGurn states in the press release, which frames the feed as a "high-margin" product and a durable source of revenue. Customers had already signed before launch, according to Reuters, and the company warns that firms that have been scraping the data in violation of its terms of service will meet "a lot of friction". Subscription pricing has not been disclosed. The pool covered goes beyond the presidential account and its 12.9 million followers: the most-followed accounts also include FBI director Kash Patel and health secretary Robert F. Kennedy Jr., several sources of regulatory and health pronouncements capable of moving entire sectors. Demand, meanwhile, is anything but hypothetical: this very week, the Hormuz blockade announcement on Truth Social sent Brent jumping before the post had even been read in full. The precedent of 9 April 2025 Truth API's business case fits in one trading day. On the morning of 9 April 2025, minutes after the open, the president posted "THIS IS A GREAT TIME TO BUY!!! DJT" on Truth Social. Around 1:30pm, he announced a 90-day pause on tariffs: the S&P 500 finished up 9.5%, the Nasdaq up nearly 12% in its best session since 2008, and Trump Media stock up 22.7%. Democratic lawmakers demanded an investigation into who knew what and when; Time's analysis noted that no offence had been established, US insider trading law fitting poorly on a president announcing his own policy. Which is precisely the point: the episode publicly demonstrated, with numbers attached, the monetary value of a few hours', or a few milliseconds', head start on the presidential word. Truth API turns that demonstration into a catalogue. A company with $871,000 in quarterly revenue The financial context of the announcement explains its logic. Truth Social generated $871,000 in revenue in the first quarter of 2026, against a $405.9 million net loss, mostly non-cash; trailing-twelve-month revenue barely reaches $3.7 million for a market capitalisation around $2.6 billion. The media business does not monetise, the stock is down about 27% this year, and most of the balance sheet is a financial war chest of over $2 billion in assets, largely in crypto. Truth API is the group's first data-licensing product, and the first whose margin depends neither on advertising audiences nor on the price of bitcoin: it sells a raw material the company alone can legally produce, first access to the statements of the sitting president. The president owns about 41% of Trump Media through a revocable trust; every subscription dollar therefore indirectly remunerates the author of the posts. On announcement day the stock gained just 0.6%: for now the market sees a small product, not a pivot. The man who bet on the teleprompter The Kalshi affair shows the same rent worked from the other end of the chain. According to the CFTC investigation revealed by ABC News and detailed by CNBC, Gabriel Perez, teleprompter operator for Trump's speeches since the 2016 campaign and an employee of a company that has equipped the White House since the 1960s, allegedly pocketed more than $90,000 in gains on Kalshi's "Mentions" markets, contracts that pay out according to the words or phrases the president will utter in a scheduled speech. He knew the script before everyone else, and press reports say he sometimes cancelled his bets mid-speech when Trump strayed from the prompter. More than a dozen speeches over three months are said to be involved, from the Davos address in January to February's State of the Union. Kalshi's internal surveillance spotted the unusual trades as early as March, froze the account and most of the gains, and referred the case to the CFTC; the White House press secretary called the affair a "disgrace" and confirmed the suspension. The case is not isolated, and that is what makes it instructive. In April, a US Army special forces master sergeant was charged over Polymarket bets on the capture of Nicolás Maduro, an operation he had helped plan; in May, a Google employee was prosecuted for trading on his employer's "Year in Search" lists. Prediction markets, now regulated and sizeable, create a new category of insider: not those who know a company's accounts, but those who know an event's script. The teleprompter is the chemically pure example: the information was neither financial nor classified, it was the text of a speech, and it was worth $90,000. The closed loop Widen the lens, because Truth API does not arrive alone. In October 2025, Trump Media announced with Crypto.com's US derivatives arm Truth Predict, a prediction market embedded in Truth Social where users will bet on elections, Fed decisions or commodities. Donald Trump Jr. is a strategic adviser to Kalshi, where he received an equity stake of roughly $300,000, while his fund 1789 Capital invested tens of millions in Polymarket, the direct competitor, where he also advises. And the regulator of it all, the CFTC, is chaired by a man described by The American Prospect as an open promoter of prediction markets. The full value chain of the presidential word thus closes on itself: the president produces the information that moves prices; his company sells the fastest access to that information; the venues where that information is bet on count his family among their advisers and shareholders; and the referee is sympathetic. Each link, taken alone, has a precedent or a justification. The whole has none: the president's 2025 income, $2.2 billion according to his financial disclosures, including more than a billion from his crypto ventures, describes a presidency in which office and estate converge, a pattern we have documented on the CLARITY Act as well as on public investment in AI. Two weights, two speeds Most striking is the contrast with America's historical doctrine on market-moving information. For the major economic indicators, CPI or the jobs report, a federal directive has required since 1985 that release be organised to reach all users at the same time: strict embargoes, sealed-room procedures, sanctions for leaks. Equal access to sensitive public information is treated there as a public good, precisely because a few seconds' head start is worth fortunes. The presidential word of 2026 moves markets as much as a CPI print, often more, and it follows the opposite doctrine: the milliseconds of head start are not neutralised, they are invoiced, and the proceeds flow mostly to the source. No rule is broken, and that is exactly the point: the doctrine of material non-public information was built for the corporate insider and the wayward official, not for a sovereign selling first access to his own voice. The serious objections Three counter-arguments deserve an honest hearing. First, selling fast access to public feeds is an ordinary industry. X sells its API to traders, agencies like Bloomberg and Reuters have been charging for speed for a century, and exchanges sell colocation and market data at prices that draw the same criticism. A post, once published, is public information; accelerating its delivery is not insider trading, and formalising through licences what firms were obtaining by scraping may even level the field. Second, in the teleprompter affair the system worked. Kalshi's surveillance detected the bets, froze the gains and referred the matter to the regulator, and the platform tightened its employment-disclosure rules in June. Third, the market itself plays down the stakes, with Trump Media stock down 27% on the year and a 0.6% reaction to the announcement; at this stage Truth API is a revenue promise, not a demonstrated cash machine. The answer to these objections does not cancel them, it locates them: none of the precedents cited, not X, not Bloomberg, not an exchange, puts the producer of the information, the seller of the access and the economic beneficiary in the same hand, and that hand signs the executive orders. The checkpoints Five things will tell whether this story remains a curiosity or becomes infrastructure. The data-licensing line in Trump Media's next quarterly results, sole judge of Truth API's commercial reality. The outcome of the Perez case at the CFTC, the first settlement over prediction market insider trading built on a speech script, which will set the rules for "Mentions" markets. The actual launch of Truth Predict and its scope: bets on Fed decisions hosted by the president's own company would add another notch of conflict. Congressional moves on prediction markets, whose lobbying in Washington has intensified as the cases pile up. And Truth API's customer list, if it ever surfaces: knowing who pays to read the president early will say a lot about who believes that head start pays. In a market, an information asymmetry always ends up with a price. What is new in July 2026 is that it now has a rate card, and a beneficiary in the White House. Sources - TMTG, press release "Trump Media and Technology Group Launches Truth API", 16 July 2026 (1 August launch, ten accounts, milliseconds, archive back to 2022, Kevin McGurn quotes): https://www.stocktitan.net/news/DJT/trump-media-and-technology-group-launches-truth-api-a-new-licensed-ua4pyh02fjbe.html - Reuters (via NBC News), "Truth Social to sell Wall Street firms the 'fastest' access to Trump's posts", 16 July 2026 (customers already signed, scraping in breach of ToS, "friction"): https://www.nbcnews.com/business/media/trump-media-early-access-truth-social-posts-rcna587912 - CBS News, "Trump Media to sell faster access to top Truth Social accounts", 16 July 2026 (12.9 million followers, president's 2025 income of $2.2bn): https://www.cbsnews.com/news/truth-api-trump-media/ - CNN Business, "Truth Social will sell Wall Street quicker access to posts", 16 July 2026 (Patel and Kennedy accounts, ~41% via revocable trust per FactSet, stock +0.6%, down 27% on the year, over a billion in crypto income in 2025): https://www.cnn.com/2026/07/16/business/truth-social-data-wall-street - CNBC, "Trump suspends teleprompter operator over Kalshi bets allegations", 16 July 2026 (Gabriel Perez, "Mentions" markets, over $90,000 frozen, Kalshi referral to the CFTC, Van Dyke and Spagnuolo precedents, Robert DeNault quote): https://www.cnbc.com/2026/07/16/trump-kalshi-teleprompter-cftc-investigation.html - The Hill, "White House suspends teleprompter operator accused of placing bets on speeches", 16 July 2026 (mid-speech bet cancellations, Karoline Leavitt quote): https://thehill.com/homenews/administration/5972297-trump-teleprompter-operator-suspended-kalshi/ - PBS News, "Trump told investors to 'buy' on social media hours before his tariff pause", April 2025 (9 April timeline, S&P +9.5%): https://www.pbs.org/newshour/politics/trump-told-investors-to-buy-on-social-media-hours-before-his-tariff-pause-rose-stocks-raising-questions-about-manipulation - Time, "Breaking Down 'Insider Trading' Accusations Leveled at Trump", April 2025 (legal framework, no offence established): https://time.com/7276515/explaining-insider-trading-accusations-leveled-at-trump-tariffs-pause/ - CNBC, "Trump's 'buy' call nets huge returns for those who listened", 9 April 2025 (Nasdaq ~+12%, DJT +22.7%): https://www.cnbc.com/2025/04/09/trumps-morning-buy-call-nets-huge-returns-for-those-who-listened.html - Variety, "Trump Media Reports Q1 Sales of $871,000 and $405.9 Million Net Loss", May 2026: https://variety.com/2026/digital/news/trump-media-truth-social-q1-2026-earnings-sales-net-loss-1236742097/ - TMTG, first quarter 2026 results (over $2bn in financial assets, positive operating cash flow): https://www.stocktitan.net/news/DJT/trump-media-technology-group-reports-first-quarter-2026-6cyeappukc37.html - stockanalysis.com, DJT statistics (market cap ~$2.61bn, trailing revenue ~$3.7m): https://stockanalysis.com/stocks/djt/statistics/ - CoinDesk, "Prediction Markets Come to Trump Media", 28 October 2025 (Truth Predict with Crypto.com: elections, Fed, sports, commodities): https://www.coindesk.com/markets/2025/10/28/trump-media-taps-crypto-com-to-launch-prediction-markets-on-truth-social - Prediction News, "Trump family faces scrutiny over strategic stake in prediction market Kalshi" (Donald Trump Jr.'s ~$300,000 stake, 1789 Capital's Polymarket investment): https://predictionnews.com/story/trump-family-faces-scrutiny-over-strategic-stake-in-prediction-market-kalshi - The American Prospect, "The CFTC's Word Games Legalizing Gambling", 15 July 2026 (the regulator's posture on prediction markets): https://prospect.org/2026/07/15/cftcs-word-games-legalizing-gambling-prediction-markets-kalshi-polymarket/ - CNBC, "Kalshi, Polymarket lobby as insider trading, betting eyed by Congress", 15 April 2026: https://www.cnbc.com/2026/04/15/kalshi-and-polymarket-congress-regulation-washington-influence.html - OMB, Statistical Policy Directive No. 3, "Compilation, Release, and Evaluation of Principal Federal Economic Indicators", 1985 (simultaneous release of major indicators): https://obamawhitehouse.archives.gov/sites/default/files/omb/inforeg/statpolicy/dir3fr09251985.pdf This article is journalistic analysis and does not constitute investment advice. Facts relating to the CFTC investigation are allegations reported by the press at the investigative stage; no offence has been adjudicated. Data as of the dates of the cited sources. ============================================================================ ANALYSIS: Ghost tankers: in the Gulf, the meter runs at $100,000 a day URL: https://l0g.fr/en/analysis/gulf-ghost-tankers-the-meter-is-running/ Canonical French source: https://l0g.fr/posts/tankers-fantomes-golfe-le-compteur-tourne/ Date: 2026-07-17 (reviewed 2026-07-17) Topics: oil, energy, geopolitics, iran, supply chain, markets ---------------------------------------------------------------------------- The Gem No. 2 entered the Persian Gulf two days before the war began. Loaded with Saudi crude in March, the supertanker never left: as of 9 July, ship-tracking data compiled by Bloomberg made it the last big vessel still trapped inside with its cargo, out of the 109 counted at the height of round one. Five days later, the American blockade was reimposed, and the trap closed again on a fleet that had only just extracted itself. The primetime address on the evening of 16 July, billed as an update on Iran and on elections, fits a doctrine on display since the 13th: the United States proclaims itself the "GUARDIAN OF THE HORMUZ STRAIT" and intends to be paid for the role. For every ship caught on the wrong side of the chokepoint, a precise accounting has restarted: days of waiting, insurance premiums, and charterparty clauses. That accounting, rarely spelled out, is what this article quantifies. One figure in circulation, three different measurements The number looping through coverage since 14 July has it that around 230 loaded tankers are stuck inside the Gulf with nowhere to deliver. It deserves an autopsy, because it probably does not measure what it is being made to say. The most rigorous count available, Bloomberg's vessel tracking, told the opposite story last week: 109 big non-Iranian crude tankers trapped in late February when the strait closed, at least 50 of them out from 18 June onwards thanks to the ceasefire, and a single one left by 9 July, the Gem No. 2. CNBC had already put the crude evacuated by freed vessels at 35 million barrels by 24 June. The stock of trapped ships from round one was nearly cleared when round two began. The 230 most likely measures something else: tankers present in the Gulf at a given moment, which in normal times is unremarkable, since hundreds of vessels load and discharge there continuously. The question for round two is not how many are inside but how many will stay stuck, and it has no published answer yet. At the worst of round one, Saudi Aramco's CEO Amin Nasser spoke of more than 600 vessels immobilised inside and 240 waiting on the other side of the strait. The flow data, meanwhile, is already unambiguous: according to Kpler, Hormuz voyages fell 52% week on week around the weekend of 12 July, and Al Jazeera counted six passages in twelve hours on 11 July, against 18 to 22 a day before fighting resumed. Kpler analyst Muyu Xu sums up the shift: the question is no longer the size of the backlog, but who is still willing to go in and out. The meter: over $100,000 per ship per day An idle tanker is never free. When the wait occurs during the commercial operations set out in the contract, it is billed as demurrage, the daily indemnity owed to the shipowner; outside that framework, it counts as detention or plain hire. Round one's orders of magnitude are documented: in early March, with roughly 700 tankers clustered on both sides of the strait, maritime law firm Fortior Law put VLCC hire and demurrage rates at more than $100,000 per ship per day, potentially $70 million a day in waiting costs across the blocked fleet. The real battle is legal: who pays for those days? The answer sleeps in the charterparties, and it is less obvious than it looks. Fortior Law notes that under standard clauses, a state closing the strait can qualify as "restraint of princes", which shields the charterer; but if the contract carries a specific Hormuz clause with an agreed daily rate, the owner collects without even a duty to mitigate. Protection and indemnity clubs such as NorthStandard have been publishing entire FAQs since February to referee between owners and charterers. The financial translation: the cost of a closed strait never disappears, it migrates along the contractual chain, from charterers to refiners and then into product prices, with a dispute at every link. Our survey of the supply chain already documented the downstream floors; this is floor zero, the ship itself. The war premium, the strait's real gatekeeper Before paying demurrage, a ship must first have agreed to enter. The real regulator of traffic is neither the blockade nor the mines: it is the war risk insurance premium, requoted sometimes from one day to the next for every transit. In peacetime, the Lloyd's Market Association describes it as little more than notional. Before the war, the market quoted 0.15% to 0.25% of hull value for a Gulf voyage. In March, at the peak, Lloyd's List reported quotes of 2.5% to 5% for a Hormuz transit, 5% to 10% and beyond for vessels with a US, UK or Israeli nexus, meaning $10m to $14m asked for a VLCC worth $138m. After June's memorandum, the premium had eased back towards 2%. Since the attacks of early July it has settled around 5% of vessel value, described as the new market norm, according to market estimates reported by Xinhua; Neil Roberts, the LMA's head of marine, describes rates that move as the risk moves, softening after the memorandum, tightening after the ship attacks of the week of 7 July, the same ones logged in our inventory of vessels struck. Two caveats stop this from becoming a story of vanishing insurance. First, cover exists: in the LMA's March survey, 88% of underwriters were still writing hull war risk and more than 90% cargo, and the association insisted that ships were staying put because masters and owners judged the safety risk too high, not because insurers refused. Second, when the private market genuinely retreats, the public floor steps in: the World Economic Forum described as early as April governments becoming insurers of last resort for strategic transits. The war risk premium works, in effect, like a market toll: it does not close the strait, it prices it, and reprices it at every incident. The official toll on top Since 13 July an openly declared layer sits on top of that private stack. Announcing the reinstated blockade, effective 14 July at 4pm Washington time, Donald Trump declared the United States would collect 20% on all cargo shipped through the strait, as payment for the security provided. The International Maritime Organization replied that there is no legal basis for imposing a toll on transit through an international strait, whose right of passage may not be suspended or impeded. Tehran contested not the principle but the till: foreign minister Abbas Araghchi claimed the guardian's role for Iran, judging that "20% is of course too much" and promising a fair tariff. History's irony is that Iran already ran the experiment this spring, collecting its own Hormuz tolls in USDT on Tron to route around OFAC. Nobody knows today how an American levy would be collected, or on what base; but at 20% of a VLCC cargo, roughly $33m at current Brent prices, merely stating it is enough to weigh on chartering decisions, which is probably its primary function. The opposite reading The picture of a fleet caught in a trap invites three serious objections. First, round one's trap proved porous. Owners eventually extracted nearly every trapped ship, riding the truce but also sailing dark; Bloomberg was still observing, on 14 July, six sanctioned supertankers leaving the Gulf in a week with transponders off, carrying the equivalent of 12 million barrels, on the Iranian side this time. A blockade, even backed by the world's largest navy, filters more than it seals. Second, most of the costs described here remain a scenario until the contracts are unwound; part of the demurrage will never be paid, absorbed by exoneration clauses, settlements between parties or plain write-offs. Third, nothing says round two will match round one's scale. The market itself only half believes it, with Brent back up to $82-$87, far from May's peak of $114. If the diplomatic track reopens, the premium will fall as it did in June, and the episode will end as a few weeks of expensive freight rather than a physical rupture. On the other side, the White House promises to escalate strikes until Tehran yields on Hormuz: between those two paths, the meter keeps running. Dials to watch Four instruments will tell whether the Gulf's fleet becomes ghostly again or merely slowed. The trapped-vessel count from satellite tracking, Bloomberg's or Kpler's, which will give round two its true measure where the 230 figure was an ambiguous snapshot. War risk premiums and the Listed Areas of Lloyd's Joint War Committee, requoted at every incident, the best thermometer of risk as perceived by those who carry it. Daily Hormuz transits, whose collapse or normalisation reads in near real time. And the structure of the oil market itself: a deepening backwardation will say that the crude idling on the water is starting to be missed on land, precisely the scenario where the strategic reserves, already dented, become the last cushion. One party's voluntary floating storage is another's trap: the first waits for a better price, the second for a right of way. The difference between the two, again this week, is measured in tens of thousands of dollars per ship per day. Sources - Bloomberg (via Yahoo Finance), "Stranded Oil Ships Clear Out of Gulf as Owners Get Wary Again", 9 July 2026 (109 tankers trapped in late February, at least 50 out after 18 June, the Gem No. 2 last remaining, Muyu Xu quote, Kpler): https://finance.yahoo.com/energy/articles/stranded-oil-ships-clear-gulf-103918129.html - CNBC, "Oil tankers with 35 million barrels stuck in Persian Gulf exited Strait of Hormuz since Iran deal", 24 June 2026: https://www.cnbc.com/2026/06/24/oil-tanker-strait-hormuz-iran-deal.html - TheStreet, "U.S. blocks Strait of Hormuz: Here's what's next for oil prices", 14 July 2026 (the ~230 tanker figure, the 20% toll announcement, the IMO's position, Abbas Araghchi's reply, Brent at $114 on 4 May): https://www.thestreet.com/investing/us-blocks-strait-of-hormuz-what-next-for-oil-prices - Al Jazeera, "Oil prices jump as US and Iran trade attacks over Strait of Hormuz", 13 July 2026 (six passages in twelve hours on 11 July, against 18 to 22 a day): https://www.aljazeera.com/economy/2026/7/13/oil-prices-jump-as-us-and-iran-trade-attacks-over-strait-of-hormuz - Yahoo Finance, "Oil prices rise as US prepares to reinstate Hormuz strait blockade, charge 20% on all cargo shipped", 13 July 2026 (voyages down 52% week on week, Kpler data): https://finance.yahoo.com/markets/article/oil-prices-rise-as-us-prepares-to-reinstate-hormuz-strait-blockade-charge-20-on-all-cargo-shipped-140723529.html - Axios, "Trump Iran blockade in Strait of Hormuz starts Tuesday", 13 July 2026 (effective 14 July, 4pm ET): https://www.axios.com/2026/07/13/trump-iran-blockade-strait-hormuz - Fortior Law, "War, Blockade and Shipping: Who Bears the Cost of Delays in Hormuz?", 3 March 2026 (700 tankers clustered, over $100,000 per day per VLCC, ~$70m a day in total, restraint of princes analysis): https://fortiorlaw.com/news/war-blockade-and-shipping-who-bears-the-cost-of-delays-in-hormuz/ - NorthStandard, "Persian Gulf Hostilities - Charterparty FAQs" (contractual allocation of costs between owners and charterers): https://north-standard.com/insights-and-resources/resources/articles/persian-gulf-hostilities-charterparty-faqs - Lloyd's List, "Gulf war risk premiums topping double-digit millions of dollars per trip", 11 March 2026 (0.15%-0.25% pre-war, 2.5%-5% for Hormuz, 5%-10% and above for US/UK/Israel nexus vessels, $10m-$14m for a $138m VLCC): https://www.lloydslist.com/LL1156586/Gulf-war-risk-premiums-topping-double-digit-millions-of-dollars-per-trip - Lloyd's Market Association, "Safety concerns, not insurance availability, driving reduced vessel traffic in the Strait of Hormuz", 23 March 2026 (88% of underwriters active on hull war, over 90% on cargo): https://lmalloyds.com/safety-concerns-not-insurance-availability-driving-reduced-vessel-traffic-in-the-strait-of-hormuz/ - Xinhua (via GlobalSecurity), "War-risk insurance rates for Strait of Hormuz vessels rise amid renewed tensions", 11 July 2026 (premium around 5% as the new norm, ~2% after the MoU, 10% peak, Neil Roberts quotes, LMA): https://www.globalsecurity.org/wmd/library/news/iran/2026/07/iran-260711-pdo02.htm - World Economic Forum, "What stopping war-risk insurance in the Strait of Hormuz tells us", April 2026 (governments as insurers of last resort): https://www.weforum.org/stories/2026/04/how-middle-east-war-turning-governments-into-insurers-last-resort/ - RFE/RL (via GlobalSecurity), "10 Japan-Linked Ships Exiting Gulf After Being Stranded For Months", 6 July 2026 (Amin Nasser's 11 May statement: more than 600 vessels stuck inside, 240 waiting outside): https://www.globalsecurity.org/wmd/library/news/iran/2026/07/iran-260706-rferl05.htm - Bloomberg, "Iran Sneaks Out Tankers Via Hormuz as Trump Amps Up Threats", 14 July 2026 (six sanctioned supertankers out with transponders off, ~12 million barrels): https://www.bloomberg.com/news/articles/2026-07-14/iran-sneaking-out-tankers-via-hormuz-as-trump-amps-up-threats - Bloomberg, "Trump Pledges to Escalate Attacks Until Iran Relents on Hormuz", 15 July 2026: https://www.bloomberg.com/news/articles/2026-07-15/trump-pledges-to-escalate-attacks-until-iran-relents-on-hormuz - NBC News, "Oil hits $87 per barrel again after latest U.S. strikes on Iran and looming blockade" (Brent between $82 and $87 in mid-July): https://www.nbcnews.com/business/energy/oil-prices-us-iran-strikes-blockade-rcna587432 - The Hill, "Trump says he will give primetime speech Thursday amid flare-up with Iran", 13 July 2026 (16 July address devoted to elections and Iran): https://thehill.com/homenews/administration/5966299-us-iran-naval-blockade/ This article is journalistic analysis and does not constitute investment advice. Per-transit costs are orders of magnitude drawn from quotes reported at different dates of the crisis, flagged as such; the theoretical toll figure is an l0g calculation. Data as of the dates of the cited sources. ============================================================================ ANALYSIS: Strategic reserves: the state of the buffer before round two URL: https://l0g.fr/en/analysis/strategic-petroleum-reserves-buffer-round-two/ Canonical French source: https://l0g.fr/posts/reserves-strategiques-petrole-matelas-deuxieme-round/ Date: 2026-07-17 (reviewed 2026-07-17) Topics: oil, energy, macro, geopolitics, commodities, iran ---------------------------------------------------------------------------- A strategic reserve is only good once per crisis. The one for 2026's first shock has been used: a record 400-million-barrel release coordinated by the International Energy Agency in March, massive American destocking, Japanese and Korean reserves drawn upon. Four months later, the ceasefire has shattered, the blockade of Iranian ports was reimposed on 14 July, and Gulf exports have fallen back below half their pre-war level. Round two of the supply shock therefore opens with a question few commentaries quantify: how much is left in the emergency tanks, and whose are they? The US SPR at its lowest since 1983 The starkest figure comes from the EIA's weekly series: the US Strategic Petroleum Reserve (SPR) fell to 316.5 million barrels in the week of 10 July 2026, after 319.5 million the week before. You have to go back to April 1983, at 317.45 million, to find less, as Mansfield Energy notes. Against its design capacity of about 714 million, the reserve is 56 percent empty, per Rigzone. The origin of the drawdown is known: on 11 March, in response to the price surge triggered by the strikes on Iran, Energy Secretary Chris Wright announced the release of 172 million barrels over about 120 days, a programme ending precisely in these days, as Al Jazeera recalls. Washington promises a refill of about 200 million barrels within a year, presented as 20% more than the volume drawn and at no cost to the taxpayer. The promise has two documented limits: the same Energy Secretary estimated that a full refill would cost on the order of $20 billion and take years, and no one refills a reserve while the barrel is surging, except to worsen the surge. The refill timetable belongs to the optimistic scenario, not to the present situation. The largest release in IEA history The international layer was tapped at the same time. On 11 March, the 32 member countries of the IEA unanimously decided to make 400 million barrels of their emergency reserves available, the largest release in the agency's history, with Asia-Oceania stocks mobilised immediately and those of the Americas and Europe from late March. The order of magnitude dwarfs the precedent: in 2022, the coordinated response to the invasion of Ukraine had mobilised about 240 million barrels in two waves. The base this release drew from was documented before the crisis: about 1.25 billion barrels of government stocks in the OECD, plus 600 million of industry stocks held under public obligation, figures we cited as early as our April situation report. A rough subtraction, which public data does not allow refining to the week, therefore puts remaining OECD government stocks around 850 million barrels, before even counting the needs of round two. The rule underpinning the system, the 90-days-of-net-imports obligation, remains met by most members, but it measures peacetime coverage: it says nothing about the ability to absorb two major shocks in the same year. The buffer still full is Chinese There is one large reservoir that round one barely touched, and it does not belong to the IEA system. China's onshore crude stocks were estimated at about 1.24 billion barrels in the spring, most likely the largest reserve in the world, complemented by some 166 million barrels of Iranian crude in floating storage in Asian waters. We documented the strategy that goes with it in the Chinese inventory capping prices: Beijing buys at a discount when the barrel falls, stops buying when it surges, and lives off its tanks during crises. The nuance is decisive for what comes next: this buffer is real, but it is private in the geopolitical sense. China takes part in no coordinated mechanism, publishes no levels, and its destocking serves its own import bill, not the world's balance. Its stabilising effect exists, we measured it in round one, but it operates by withdrawing demand, not by supplying the open market. In other words, the biggest remaining cushion cushions China first, the world by ricochet, and no one controls its trigger except Beijing. Round two starts dented The calendar is the cruel fact of this file. The American 120-day programme ends in mid-July, at the exact moment the memorandum's clocks break: blockade reimposed on 14 July, strikes expanded, and, per the Goldman Sachs estimate reported by Reuters, Gulf exports back below 50% of their pre-war level in the week to 15 July, after recovering above 80% during the truce. The first shock found full reserves and a market confident in their use. The second finds an SPR below its 1983 floor, an OECD base cut by a third, and governments that now know the political cost of destocking: whoever fires their last rounds tells the market so, and the market prices the nakedness that follows. This is the most important mechanism to grasp: a strategic reserve works as much through its level as through its existence. While it is full, its mere presence calms risk premiums. Once dented, each barrel released reassures less and each barrel remaining counts double. Emergency stocks are a non-linear shock absorber, and the non-linear zone is precisely the one being entered. The shock absorbers that remain An honest inventory does not stop at the tanks, because the market of 2026 has other cushions, and they are substantial. The first is OPEC+ spare production capacity: the spring quota increases and the 188,000-barrel-a-day adjustment confirmed for August, with the explicit option to accelerate, pause or reverse, make up a tap that 2022 did not have at this scale. The second is demand: soft, with a China in retreat that had already cut its imports by 20% year on year at the worst of the first shock. The third is American: the United States being a net exporter, the 90-day rule is met there even with a low SPR, and the country produces at record levels; the insurance function of its reserve has changed in nature since 1983, which puts the historical comparison brandished by both sides into perspective. These shock absorbers nonetheless share one trait that sets them apart from stocks: they produce barrels over time, not immediately, and they assume the infrastructure works. Saudi spare capacity is worthless if the terminals are under fire or if the strait no longer lets tankers and carriers through. The stock, by contrast, is already ashore, on the right side of the chokepoints. It was spent first for that precise reason, and for the same reason its current level remains the true measure of the safety margin. The gauges to watch Five gauges track the state of the buffer continuously. The SPR weekly series published by the EIA, every Wednesday, to check whether the destocking stops or resumes. The IEA's follow-up statements on the collective action, which will say whether a second release is contemplated and how deep. Signals of Chinese destocking or restocking, readable in imports and satellite tank tracking. The OPEC+ decisions of August, the first test of the acceleration option. And the stocks at Cushing, already scraping their operating floor, which measure physical tension where the American price is set. Round one of 2026 proved that strategic reserves work: they cushioned the largest supply shock in the history of the oil market. It also consumed a large share of their power. If round two stays short, the remaining cushions will probably suffice, helped by OPEC+ and soft demand. If it settles in, the world will discover the difference between a market that fears running out of oil and a market that knows it has already used its insurance. That is not the same price for a barrel. Sources - EIA, Weekly U.S. Ending Stocks of Crude Oil in SPR, series WCSSTUS1 (316.5 million barrels as of 10 July 2026, 319.5 on 3 July): https://www.eia.gov/dnav/pet/hist/LeafHandler.ashx?n=PET&s=WCSSTUS1&f=W - Mansfield Energy, "Strategic Petroleum Reserve Falls to Lowest Level Since 1983" (April 1983 floor at 317.45 million): https://mansfield.energy/2026/07/02/strategic-petroleum-reserve-falls-to-lowest-level-since-1983/ - Rigzone, "USA Strategic Petroleum Reserve is 56 Percent Empty", 13 July 2026 (design capacity ~714 million): https://www.rigzone.com/news/usastrategicpetroleumreserveis56percentempty-13-jul-2026-184121-article/ - Al Jazeera, "Why the US Strategic Petroleum Reserve matters amid US-Iran tensions", 9 July 2026 (release of 172 million barrels over ~120 days announced 11 March by Chris Wright, ~200-million refill promise, ~$20bn cost of a full refill): https://www.aljazeera.com/economy/2026/7/9/why-the-us-strategic-petroleum-reserve-matters-amid-us-iran-tensions - IEA, "IEA Member countries to carry out largest ever oil stock release", 11 March 2026 (400 million barrels, 32 countries, Asia-Oceania then Americas-Europe sequencing): https://www.iea.org/news/iea-member-countries-to-carry-out-largest-ever-oil-stock-release-amid-market-disruptions-from-middle-east-conflict - IEA, "Update on IEA collective action decision of 11 March 2026" (implementation and member contributions): https://www.iea.org/news/update-on-iea-collective-action-decision-of-11-march-2026 - Congressional Research Service and IEA (OECD base: ~1.25 billion barrels of government stocks + ~600 million under industry obligation, cited in our April situation report): https://www.iea.org/reports/oil-market-report-march-2026 - Reuters (via Investing), Goldman Sachs estimate: Gulf exports back below 50% of pre-war in the week to 15 July, after more than 80% during the truce: https://uk.investing.com/news/commodities-news/oil-prices-rise-2-as-hostilities-worsen-in-the-middle-east-4771073 - OPEC, statement of 5 July 2026 (188,000 b/d adjustment for August, option to accelerate, pause or reverse): https://www.opec.org/pr-detail/1835609-5-july-2026.html - OilPrice (Chinese onshore stocks ~1.24 billion barrels, ~166 million of Iranian crude in floating storage): https://oilprice.com/Latest-Energy-News/World-News/China-Boosts-Oil-Stockpiles-Despite-Import-Plunge.html This article is journalistic analysis and does not constitute investment advice. Remaining OECD stock levels are calculated orders of magnitude, flagged as such. Data cited as of the date of its sources. ============================================================================ ANALYSIS: NAV loans: the leverage no one sees URL: https://l0g.fr/en/analysis/nav-loans-the-hidden-fund-level-leverage/ Canonical French source: https://l0g.fr/posts/nav-loan-levier-cache-fonds-credit-prive/ Date: 2026-07-16 (reviewed 2026-07-16) Topics: private credit, private equity, leverage, systemic risk, valuation, liquidity ---------------------------------------------------------------------------- When a fund can no longer sell, one option remains: borrow against what it is not selling. That is the principle of the NAV loan, the loan secured on the net asset value of an entire private equity or private credit portfolio. The tool had existed for years in the discretion of "fund finance" desks. The exit drought made it change scale: about $70 billion deployed in 2025 according to 17Capital, a market that a consensus of projections sees tripling or more by 2030. This leverage stacks on top of the already heavy debt of the portfolio companies, and it remains largely invisible to the end investor. This article takes apart the mechanics, the use that causes offence, and what this market says about the real state of private assets. The drought that creates the need The NAV loan thrives on a simple problem: the money no longer comes out. According to Allianz's analysis, annual distributions from private equity funds have been stuck between 8 and 10% of portfolio value for three years, against a historical norm of 20%, ever since the rise in rates froze the exit channel. The stock piles up: about 28,000 companies held by funds are awaiting a buyer worldwide, against 19,000 in 2019, some $3.2 trillion of unrealised value. This breakdown in distributions has changed the hierarchy of metrics. DPI, the capital actually returned relative to the capital paid in, has dethroned the internal rate of return as institutional investors' first criterion: according to the surveys cited by Pipeline Road, 74% of LPs now make it their primary re-up criterion. A manager who wants to raise the next fund must therefore show cash returned, precisely when sales no longer generate it. The NAV loan was born big in this vice. Borrowing against the whole portfolio The mechanics are best understood by contrast with the funds' other credit line. Early in its life, a fund uses a subscription line: a bridge facility secured by its investors' uncalled commitments, repaid as capital calls come in. The NAV loan comes at the other end of the fund's life, when the capital is deployed and there is nothing left to call. The collateral is no longer the LPs' promise, but the value of the portfolio itself: per the reference description by Moonfare, a diversified pool of holdings, structured as share pledges or assignments of distribution rights, with a loan-to-value typically capped between 5 and 25% of net asset value. This low loan-to-value and the cross-collateralisation over every line of the portfolio are what make the lender safe: the whole set of holdings would have to lose enormous value before the loan is threatened. They also explain why agencies agree to rate these facilities, an exercise whose limits our guide on reading a credit rating recalls when the pledged value is itself a model-based estimate. Because everything rests there: the "V" in NAV is a net asset value computed by the manager, not a market price. One borrows against an opinion. The use that causes offence: the borrowed distribution What the proceeds are used for makes all the difference, and this is where the controversy formed. Used to support a portfolio company or seize a bolt-on acquisition, the NAV loan is a management tool. Used to pay a distribution to investors, it becomes something else: the fund borrows to hand LPs money the portfolio has not yet earned, and the distribution inflates DPI without any value having been realised. The investor receives, in effect, an advance on their own supposed performance, whose cost and risk stay in the fund. The criticism drew blood. According to the history compiled by Wikipedia, the use of NAV loans to fund distributions fell 90% in the second half of 2023, under investor pressure. The Fund Finance Association now estimates that about 80% of facilities serve to reinvest in the portfolio and 20% to distribute. The proportion has cleaned up; the precedent remains: when DPI pressure returns, the borrowed-distribution tool is available, documented, and road-tested. Leverage on leverage The systemic point is not any single use, but the stacking. The companies in a private equity portfolio already carry their acquisition debt, that of the LBOs and direct loans whose defaults and liability management we track. The NAV loan adds a layer of debt at the fund level, pledged on the net value of those same indebted companies. And upstream, the fund's investors, insurers first, sometimes carry their own leverage. Moody's explicitly files NAV loans, along with PIK, under the hidden leverage proliferating in US leveraged finance, outside the balance sheets where one looks for it. The Financial Stability Board goes further: its May 2026 report finds that private credit's reported leverage is understated and that true leverage, all layers included, would approach 7 times EBITDA. The stack's fragility lies in its circularity. The NAV loan is pledged on a value the manager computes itself, that terrain of model-based valuations we have documented: as long as you do not sell, the NAV stays smooth. Yet it is precisely because one does not want to sell that one borrows. Should valuations be marked down, the loan-to-value would rise mechanically, triggering margin calls or early repayments, at the worst moment. Fund-level leverage thus turns a valuation correction, an accounting event, into a liquidity need, a very real one. The ILPA framework, a late guardrail Institutional investors have obtained the beginnings of a framework. The guidance published by ILPA, the LP association, recommends that managers obtain the prior consent of the investor committee before putting a NAV facility in place, and provide all LPs with standardised disclosures: use of proceeds, size, structure and vehicles used, costs, induced obligations. As Mayer Brown notes, this guidance prohibits nothing: it demands transparency. That it had to be demanded is the information. Until this guidance, a fund could add a layer of debt on its investors' portfolio without informing them by name. The market, meanwhile, is institutionalising at speed: the Fund Finance Association sizes the market around $100 billion and projects $600 billion in 2030, while 17Capital and Oaktree retain a path from $44 billion in 2023 to $145 billion in 2030. The ranges diverge, the direction does not. A legitimate tool, on three conditions Fairness requires giving the defence its full strength, because it is solid. First, a well-used NAV loan is often the least bad option: rather than dumping an asset into a closed exit market, the fund borrows at a modest loan-to-value to hold out for better conditions, or to defend a portfolio company that needs capital. The forced sale would destroy more value than the loan costs. Second, the lender's risk is genuinely contained: a 5 to 25% loan-to-value on a diversified, cross-collateralised pool requires a generalised collapse to be dented, which justifies these facilities finding sophisticated lenders and ratings. Finally, the rebalancing of uses, 80% toward investment, shows that LP discipline has bitten: ILPA did not ban the tool, it brought it out of the shadows, and that is largely enough when investors do their job. The defence holds on three conditions: that the pledged NAV be honest, that the use remain investment rather than cosmetic distribution, and that LPs know what is done in their name. None of the three is guaranteed by construction. All three rest on the quality of governance, in a market where the value is declared by the one borrowing against it. The dials to watch Five dials will say whether fund-level leverage stays a tool or becomes a symptom. The share of facilities funding distributions, whose rise would signal the return of borrowed DPI. The level of loan-to-values granted, because a market drifting from 15 toward 30% of NAV changes nature. The gap between model-based valuations and actual secondary-market transactions, which tests the honesty of the "V". The recovery, or not, of real distributions in 2026, which Allianz sees climbing back toward 17 to 19% in its central scenario: if it materialises, the need deflates by itself. And the application of the ILPA guidance, measurable by the number of funds informing their LPs before rather than after. The NAV loan is the chemical developer of private markets: it exists at this scale only because exits are jammed and valuations refuse to fall. A market that borrows against its own estimates to wait it out is making a coherent bet as long as the wait ends in sales. If it does not, the added leverage will not have bought time: it will have added a floor to what must, one day, be reconciled with prices. Sources - Allianz Research, "Private equity in transition: from distribution drought to selective recovery" (distributions at 8-10% of NAV against a 20% historical norm, 17-19% projection for 2026, NAV lending path from $44bn in 2023 to $145bn in 2030 per 17Capital/Oaktree): https://www.allianz.com/en/economicresearch/insights/publications/specialsfmo/260220-private-equity.html - Partners Capital / Private Markets Insights (~28,000 portfolio companies awaiting exit against 19,000 in 2019, ~$3.2tn of unrealised value): https://www.privatemarketsinsights.com/post/the-liquidity-drought-forces-a-reset-private-markets-midyear-review - Pipeline Road, "Private Equity Returns Statistics" (DPI the primary re-up criterion for 74% of LPs): https://pipelineroad.com/blog/private-equity-returns-statistics - Moonfare, "What is NAV lending" (typical loan-to-value of 5 to 25% of NAV, cross-pledged collateral, distinction from the subscription line, ~$70bn deployed in 2025, $700bn TAM): https://www.moonfare.com/blog/what-is-nav-lending - Wikipedia, "NAV lending" (90% fall in distribution-funding use in H2 2023, ~80% of facilities oriented to investment per the Fund Finance Association): https://en.wikipedia.org/wiki/NAVlending - Private Debt Investor, "NAV loans are the next frontier of private credit's growth" (market ~$100bn, $600bn projection for 2030 per the Fund Finance Association): https://www.privatedebtinvestor.com/nav-loans-are-the-next-frontier-of-private-credits-growth/ - Moody's, "Will CLO performance and leveraged finance trends diverge or align in 2026?" (hidden leverage via PIK and NAV lending increasingly prevalent): https://www.moodys.com/web/en/us/creditview/blog/leveraged-finance-and-clo-2026.html - Financial Stability Board, "Report on Vulnerabilities in Private Credit", 6 May 2026 (reported leverage understated, true leverage close to 7x EBITDA): https://www.fsb.org/uploads/P060526.pdf - ILPA, "New ILPA Guidance Encourages LP-GP Dialogue, Transparency around NAV-based Facilities" (LP committee consent, standardised disclosures): https://ilpa.org/news/new-ilpa-guidance-encourages-lp-gp-dialogue-transparency-around-nav-based-facilities/ - Mayer Brown, "NAV Facilities: The ILPA's New Guidance" (scope and limits of the guidance): https://www.mayerbrown.com/en/insights/publications/2024/09/nav-facilities-the-institutional-limited-partners-associations-new-guidance This article is journalistic analysis and does not constitute investment advice. Market sizes are estimates, cited as of the date of their sources. ============================================================================ ANALYSIS: The broken clocks of the US-Iran memorandum URL: https://l0g.fr/en/analysis/the-august-clocks-us-iran-memorandum-deadlines/ Canonical French source: https://l0g.fr/posts/les-horloges-d-aout-mou-usa-iran-echeances/ Date: 2026-07-16 (reviewed 2026-07-16) Topics: geopolitics, oil, iran, macro, markets, us politics ---------------------------------------------------------------------------- The memorandum signed on 17 June between Washington and Tehran had turned the war into a calendar. The US naval blockade was meant to disappear within 30 days, commercial passage through the Strait of Hormuz was to remain free of charge for 60 days, and a final agreement was to be negotiated within the same period. That reading is no longer valid. Donald Trump declared the memorandum and ceasefire over on 8 July, strikes resumed, and the United States reimposed its blockade of Iranian ports on 14 July. On 16 July, the Associated Press was still reporting an expansion of US targets and new Iranian retaliation. The question is therefore no longer whether the written deadlines will be met as if the text still governed events. The useful question is what its former clocks allow us to measure: the gap between commitments and facts, the mediators' ability to restore a framework, and the risk that the shock spreads from Hormuz to the rest of the economy. What the memorandum actually said The memorandum text, reproduced by the American Presidency Project, separated three commitments. Military operations were to stop immediately. Washington was to begin removing its blockade upon signature and finish within 30 days. Tehran was to arrange safe commercial passage without charge for 60 days while discussing the strait's future administration with Oman. Finally, a definitive agreement covering nuclear issues, sanctions and enriched material was to be negotiated within 60 days, extendable by mutual consent. These clauses contained an important asymmetry. The 30-day deadline required the United States to lift the blockade, while the 60-day clause did not settle what would follow the no-charge period. That ambiguity became a conflict of interpretation, then an operational conflict. It explains why the 17 June memorandum remains useful as a reference document even though it has ceased to be a credible roadmap. Hormuz is neither simply open nor simply closed The strait cannot be reduced to a map on which one actor controls everything. The AP reports that crossings fell by about 52% between the Friday and Monday preceding 14 July, with roughly 14 ships on Sunday versus nearly 130 a day before the war. The UK's UKMTO had received six reports of attacks against ships near the Omani route since 25 June. Mines, escorts, Iranian registration demands and fear of attack therefore fragment passage without producing stable, exclusive control. Washington and Tehran each claim a capacity for control. International passage nevertheless remains the legal reference, and the Trump administration ultimately abandoned its announced plan to levy 20% on cargo. The reimposed blockade targets Iranian ports and flows, while Iran threatens regional energy exports. This military and legal overlap is more dangerous than a simple toll because it multiplies opportunities for miscalculation. The energy shock has changed in nature The earlier draft described a risk premium rebuilding as a deadline approached. That formulation is no longer defensible. The dominant factor is now the actual resumption of hostilities. Reuters estimates that about a fifth of global oil and liquefied natural gas flows passed through Hormuz before the war. It also reports a Goldman Sachs estimate that Gulf exports, which had recovered above 80% of pre-war levels after the memorandum, fell back below 50% in the week to 15 July. Outside supply provides a buffer, not a geopolitical solution. OPEC confirmed a production adjustment of 188,000 barrels a day for August while retaining the option to increase, pause or reverse it. That flexibility can soften a global shortfall. It cannot repair attacked ships, mines, marine insurance or a contested corridor. Three paths from 16 July These paths are working bounds, not quantified probabilities. A negotiated return to the framework. Qatar, Pakistan and other mediators secure de-escalation followed by renewed technical talks. Axios was still reporting coordinated attempts to revive the memorandum after its breakdown. The 17 June text could serve as a base, but a new agreement would need explicit terms for the blockade, passage routes and monitoring. Conflict contained around the strait. Strikes and attacks on ships persist without tipping into total regional war. Flows continue at low capacity, under escort and with elevated insurance costs. This is an unstable status quo in which every incident can reduce traffic before any political decision. Regional expansion. Attacks increasingly reach energy infrastructure, US bases or other maritime corridors. The risk would then extend beyond Hormuz to several trade arteries and to Gulf producers' ability to export. The useful dashboard now The memorandum's dates are no longer sufficient. The relevant indicators are AP and military reports on strikes, UKMTO vessel alerts, the number and composition of crossings, the exact reach of the US blockade, Qatari and Pakistani mediation, and the next OPEC+ meeting on 2 August. The oil curve remains useful, but it must be read as the consequence of physical flows and operational risk, following the method in our guide to reading the oil market. The memorandum bought time. Renewed strikes consumed it before the written deadlines. Analytical integrity therefore requires abandoning mid-August as an automatic decision date. The operative calendar is now event-driven: attack, retaliation, mediation, partial reopening or regional expansion. Sources - American Presidency Project, full text of the Islamabad Memorandum of Understanding, 17 June 2026: https://www.presidency.ucsb.edu/documents/islamabad-memorandum-understanding-between-the-united-states-america-and-the-islamic - Axios, "Regional mediators push to save US-Iran nuclear deal after strikes", 9 July 2026: https://www.axios.com/2026/07/09/us-iran-nuclear-deal-mediators-qatar-pakistan - Associated Press, "A look at US and Iranian claims of control over the Strait of Hormuz", 14 July 2026: https://apnews.com/article/trump-iran-strait-of-hormuz-8df557699c900b29fb33172e6da7f3e9 - Associated Press, "US reimposes its blockade on Iran after Tehran's attacks on ships", 15 July 2026: https://apnews.com/article/iran-us-hormuz-strait-war-july-14-2026-abd060c55feea216625689e57d8f76be - Associated Press, "US expands strikes into northern Iran and disables ship trying to run blockade", 16 July 2026: https://apnews.com/article/iran-us-hormuz-strait-war-july-16-2026-f98ff56554de2336f0e85bb5fdcae769 - Reuters, "Oil prices rise as hostilities worsen in the Middle East", 15 July 2026: https://uk.investing.com/news/commodities-news/oil-prices-rise-2-as-hostilities-worsen-in-the-middle-east-4771073 - OPEC, statement on the August production adjustment, 5 July 2026: https://www.opec.org/pr-detail/1835609-5-july-2026.html This analysis is current as of 16 July 2026. It is not investment advice. The paths described are working scenarios, not predictions. ============================================================================ ANALYSIS: Life insurers: retirement savings, the silent fuel of private credit URL: https://l0g.fr/en/analysis/life-insurers-retirement-savings-private-credit-bermuda/ Canonical French source: https://l0g.fr/posts/assureurs-vie-epargne-retraite-credit-prive-bermudes/ Date: 2026-07-16 (reviewed 2026-07-16) Topics: insurance, private credit, systemic risk, bermuda, regulation, retirement ---------------------------------------------------------------------------- An American retiree who buys a lifetime annuity believes they are entrusting their savings to one of the most prudent trades in finance. They are right about the history, less surely about the present. A growing share of these annuities is now backed by private credit loans, housed in Bermuda reinsurers controlled by the asset managers themselves, and rated by specialised agencies whose grades are not public. Each link in this chain has its own logic. It is their stacking that manufactures something new: a life-insurance balance sheet whose contents neither the saver, nor at times the home regulator, can quite see any more. This article completes our private credit series, after the defaults and the gating, the redemptions facing the mega-IPOs and the migration of credit risk out of regulatory sight. The link treated here is the largest of all: the one that connects retirement savings to shadow credit. A third of the balance sheet in private assets The orders of magnitude set the scale. According to American Banker's investigation, US life insurers have allocated nearly a third of their $5.6 trillion in assets to private credit, about $1.9 trillion of exposure. The IMF uses a similar figure in its April 2026 financial stability report: private credit accounts for about 35% of North American insurers' portfolios. The same report highlights a divide within the sector: insurers backed by private-equity firms hold nearly twice as many illiquid assets as the others. This shift did not happen by chance. For a private credit manager, a life insurer is the ideal partner: it brings a long, stable liability, fed by regular premiums, exactly the kind of permanent funding that traditional fundraising no longer supplies. For the insurer, private loans offer extra yield in a business whose margins have compressed. Apollo showed the way with Athene, and KKR, Ares, Brookfield and Carlyle each have their insurance vehicle. The annuity trade has become, for the giants of alternative asset management, a liability factory. The Bermuda triangle, insurance edition The second stage of the structure plays out offshore. Rather than carrying the commitments in the US entity, regulated state by state, the insurer cedes its reserves to a reinsurer, very often affiliated with the same group and domiciled in Bermuda, where the prudential and accounting regime is more accommodating. This is asset-intensive reinsurance. According to Bloomberg's investigation of the sector, US life insurers ceded $2.4 trillion of reserves in 2024, of which more than $1.1 trillion went to offshore jurisdictions, Bermuda first among them. Athene's case illustrates the degree of integration. According to the same investigation, 96% of its $200 billion of reinsurance came from Bermuda, and in 2024 all of its $192 billion of reinsurance support came from its own affiliate. The loop is closed: the insurer cedes its reserves to itself, under another flag, and the manager who controls the whole invests the assets in its own funds and its own loan origination. The American Academy of Actuaries devoted an entire brief to the risks of this offshore ceded reinsurance, from reinsurance leverage to concentration in affiliated assets. The private rating, key to regulatory capital There remains the task of fitting illiquid loans into a regulated balance sheet. The tool is called the rated feeder note: a feeder vehicle invests in the private credit fund and issues debt securities carrying a rating, most often a private one, disclosed only to the subscriber. As Troutman Pepper explains, holding a rated note rather than an unrated fund interest reduces the capital the insurer must set aside. The economic exposure is the same; its regulatory cost is not. The rise of these private ratings is spectacular. The IMF, cited by Institutional Investor, notes that about 7,000 securities were rated by specialised agencies in 2023, against 2,000 in 2019, a near-quadrupling. And the quality of these grades raises questions: according to the research relayed by Alternative Credit Investor, when a security leaves the in-house assessment of the NAIC's securities valuation office for a private rating, it is upgraded more than four times as often as it is downgraded; the same move to a public rating produces as many upgrades as downgrades. The choice of rating channel then looks less like a measure of risk than an optimisation of capital. Regulators have begun to react. The NAIC adopted rules, applicable in 2026, that allow it to override ratings deemed too favourable and impose heavier capital charges. Capstone reports that the US Treasury itself is engaging with state regulators on the stability of this market, a sign that the subject has left the circle of specialists. 777 Re, the dress rehearsal What can go wrong is no longer a hypothesis: it has a name. 777 Re, the Bermuda reinsurer of the 777 Partners group, had accumulated on its balance sheet affiliated assets, invested in its shareholder's own businesses, from a football club to an airline. On 8 October 2024, the Bermuda Monetary Authority cancelled its registration, having found an excess of affiliated assets, deficient governance and insufficient capital contributions. Upstream in the chain, the American insurer A-CAP, which had ceded $1.7 billion of reserves to 777 Re according to the regulators' petition relayed by the Retirement Income Journal, saw its rating cut by AM Best in February 2024, the agency citing high reinsurance leverage and the deteriorating quality of its counterparties. The episode stayed contained, and that is good news. But it validates the contagion pattern in miniature: American annuities, an affiliated Bermuda reinsurer, illiquid assets tied to the shareholder, and a state regulator discovering the problem from the end of the chain. The difference between 777 Re and the sector's large players is one of scale and asset quality, not of structure. The long liability, the model's defence The argument of the model's defenders deserves to be stated at full strength, because it is serious. An annuity is not a bank deposit: the liability is long, predictable, and early surrenders are curbed by contractual and tax penalties. A holder of twenty-year commitments is precisely the actor best placed to carry illiquid assets, far better than a semi-liquid fund open to quarterly redemptions, whose repeated gating we have documented. Asset-liability matching is the heart of the insurance trade; private credit can find a legitimate place in it. The data is, moreover, partly reassuring. The liquidity ratios of Bermuda life reinsurers comfortably exceed the regulatory minimums, according to the regulator's figures relayed by the trade press, and the Bermuda Monetary Authority has tightened its review of asset strategies. The IMF itself, in its April 2026 press briefing, judges the exposure of insurers and pension funds to private credit "fairly manageable to date", while calling on supervisors to keep watching it closely. The reservation holds in three points. First, the liability is long only as long as surrenders remain discouraged: a sharp rise in rates, which makes old annuities uncompetitive, can accelerate exits at the precise moment illiquid assets are hardest to sell, the run scenario the IMF described as early as its work on private equity and life insurers. Next, the model-based valuation of private assets makes the balance sheet hard to challenge from the outside, the opacity problem at the centre of the silent contagion we have been documenting for months. Finally, generalised affiliation, where the same group originates the loans, manages the funds, controls the insurer and the reinsurer, concentrates the conflicts of interest that 777 Re's structure displayed on a small scale. The signals to watch Five indicators will say whether this structure ages well. The pace of reserve cessions to Bermuda, first, which the American Academy of Actuaries and the NAIC now track closely. The share of affiliated assets on reinsurers' balance sheets, the criterion that brought down 777 Re. The effective enforcement of the NAIC's new rules on private ratings, and the number of grades overridden. The surrender rate on annuities in a rate move, the only real test of the long liability. And the transparency of Bermuda's regulators, whose credibility has become a component of America's financial-stability apparatus. Life insurance has always invested long savings in long-term assets; that is its function. The novelty is not there. It lies in the concentration of roles in the hands of the same groups, the shifting of reserves toward more accommodating jurisdictions, and ratings whose discretion suits everyone except the person who, at the end of the chain, collects the annuity. The risk has not left the system: it has settled where the saver never thinks to look for it, on the balance sheet of their own insurer. Sources - American Banker, "Is private credit a $2 trillion-dollar insurance timebomb?" (nearly a third of US life insurers' $5.6tn of assets in private credit, about $1.9tn): https://www.americanbanker.com/news/is-private-credit-a-2-trillion-dollar-insurance-timebomb - IMF, Global Financial Stability Report, April 2026 (private credit ~35% of North American insurers' portfolios; private-equity-backed insurers ~2x more illiquid assets): https://www.imf.org/en/publications/gfsr/issues/2026/04/14/global-financial-stability-report-april-2026 - Bloomberg, "Apollo and Wall Street Private Equity Firms Bet on America's Life Insurance" ($2.4tn of reserves ceded in 2024, of which more than $1.1tn offshore; Athene: 96% of $200bn of reinsurance in Bermuda, $192bn entirely affiliated in 2024): https://www.bloomberg.com/graphics/2025-america-insurance-part-1/ - American Academy of Actuaries, "Asset-Intensive Reinsurance Ceded Offshore" (risks of asset-intensive reinsurance, leverage, affiliated assets): https://actuary.org/wp-content/uploads/2024/02/risk-brief-bermuda-reinsurance0.pdf - Troutman Pepper, "Private Credit Fund Investments: How New NAIC Rules Could Affect Insurance Companies" (mechanics of rated feeder notes, NAIC rules applicable in 2026): https://www.troutman.com/insights/private-credit-fund-investments-how-new-naic-rules-could-affect-insurance-companies/ - Institutional Investor, "The IMF Is Raising the Alarm on Insurance Investments in Private Credit" (~7,000 securities rated by specialised agencies in 2023 against ~2,000 in 2019): https://www.institutionalinvestor.com/article/imf-raising-alarm-insurance-investments-private-credit - Alternative Credit Investor, "Insurers and private credit: Ratings under the microscope" (asymmetry of revisions: more than four upgrades per downgrade on moving to a private rating), December 2025: https://alternativecreditinvestor.com/2025/12/04/ratings-under-the-microscope/ - Capstone, "Insurers' Increasing Exposure to Private Credit Attracts Regulators' Scrutiny" (Treasury-state regulator dialogue, withdrawn NAIC report): https://capstonedc.com/insights/insurers-increasing-exposure-to-private-credit-attracts-regulators-scrutiny/ - Bermuda Monetary Authority, notice of cancellation of 777 Re Ltd's registration, 8 October 2024 (excessive affiliated assets, governance, capital): https://www.bma.bm/viewPDF/documents/2024-10-08-12-44-33-Notice---Cancellation-of-Registration---777-Re-Ltd.pdf - Retirement Income Journal, "Double Trouble in the Bermuda Triangle" (A-CAP: $1.7bn of reserves ceded to 777 Re, rating cut by AM Best in February 2024): https://retirementincomejournal.com/article/double-trouble-in-the-bermuda-triangle/ - Insurance Business (Reinsurance), "Bermuda life insurers' liquidity ratios soar past regulatory minimum" and "BILTIR signals continued scrutiny" (liquidity ratios above the minimums, reinforced BMA oversight): https://www.insurancebusinessmag.com/reinsurance/news/breaking-news/bermuda-life-insurers-liquidity-ratios-soar-past-regulatory-minimum-573904.aspx - IMF, GFSR press briefing transcript, April 2026 (exposure "fairly manageable to date", continued monitoring): https://www.imf.org/en/news/articles/2026/04/15/tr-04142026-press-briefing-transcript-global-financial-stability-report-spring-meetings-2026 - IMF, "Private Equity and Life Insurers", Global Financial Stability Note, December 2023 (mass-surrender scenario in a rate shock): https://www.imf.org/en/publications/global-financial-stability-notes/issues/2023/12/13/private-equity-and-life-insurers-541437 This article is journalistic analysis and does not constitute investment advice. Data is cited as of the date of its sources. ============================================================================ ANALYSIS: The residual value guarantee, the blind spot of the credit that finances AI URL: https://l0g.fr/en/analysis/residual-value-guarantee-ai-infrastructure-credit/ Canonical French source: https://l0g.fr/posts/valeur-residuelle-garantie-credit-infrastructure-ia/ Date: 2026-07-15 (reviewed 2026-07-15) Topics: private credit, ai, data centers, systemic risk, valuation, leasing ---------------------------------------------------------------------------- The financing of artificial intelligence has changed in nature without the public debate really noticing. The tens of billions poured into compute centers no longer take only the form of classic corporate debt. They pass through dedicated vehicles that buy the hardware, lease it to the operator, and rest on a discreet but central promise: the residual value guarantee. This clause sets the floor value the asset will be worth at the end of the lease. It makes the structure financeable, takes it off the operator's balance sheet, and shifts to a guarantor the hardest risk to estimate in the whole chain: how much will a data center packed with chips that go obsolete in three years be worth in ten or fifteen? Leasing the brick, guaranteeing the wreck The best-documented case is Meta's. In October 2025, the group financed its Hyperion campus, in Louisiana, through a special-purpose vehicle named Beignet Investor LLC, which raised about $27.3bn of senior secured notes at 6.581%, maturing 2049, rated A+ by S&P. According to Bisnow, Blue Owl funds hold 80% of the joint venture and Meta 20%, with investors paid not by corporate debt but by the operating rent Meta pays the vehicle. The debt does not appear on the group's balance sheet. The pivot of the structure fits in one signature. According to Global Data Center Hub, Meta granted investors a residual value guarantee over sixteen years: if the campus value falls below an agreed threshold and Meta decides not to renew the lease, the group must repay the vehicle's investors. The guarantee is what lets creditors accept a very long-lived asset without directly bearing the risk that it depreciates faster than expected. The law firm Quinn Emanuel moreover ranks these structures among the new hotbeds of litigation in AI financing, precisely because of the uncertainty over the exit value. The SPV that buys then leases, the operator that guarantees the end-of-life value, the debt kept off balance sheet: the mechanism is nothing new. It is the old finance lease, applied to an asset whose true economic life no one knows. Private credit joins the loop What Meta does with Blue Owl, the compute operators do with private credit, pushing the logic a notch further: they back the debt directly with the chips. CoreWeave, the leading such neocloud, inaugurated in August 2023 the first loan collateralised by Nvidia H100 GPUs, arranged by Magnetar and Blackstone. The move has since changed scale. According to CoreWeave, its DDTL 4.0 facility of $8.5bn, rated A3 by Moody's and A (low) by DBRS, is the first financing backed by compute hardware to reach investment grade, secured by a $14.2bn contract with Meta. The falling cost of capital tells of the market's growing confidence. According to Quartz, CoreWeave borrowed against GPUs at about 15% in 2023; the fixed tranche of the DDTL 4.0 comes in around 5.9% in spring 2026. The same article recalls the two implicit bets of any chip-backed loan: that the hardware keeps enough value over the life of the loan, and that the utilisation rate stays high enough to service the debt. Yet a high-end GPU loses about half its resale value in three years. Forbes poses the question that follows logically: once this debt has gone investment grade, who really bears the risk? Private credit does not stop at lending to neoclouds. It also structures the biggest bets on AI. We documented this with the $35bn financing closed by Apollo and Blackstone for Anthropic, analysed here: a vehicle buys Google's TPU chips, leases them to Anthropic, and the whole is backed by residual value guarantees from Broadcom and payment guarantees from Google (Bloomberg). The same brick recurs everywhere: when the credit quality of the asset is not enough, you add a guarantor. The heart of the calculation: the depreciation curve The whole soundness of these structures rests on one accounting assumption: the period over which the hardware is depreciated. The longer it is, the lower the annual cost looks, the higher the reported profits, and the more a distant residual value seems credible. That is precisely where the consensus cracks. Investor Michael Burry brought the subject into the open in late 2025. According to CNBC, Google, Oracle and Microsoft depreciate their AI hardware over five to six years, when the renewal cadence of Nvidia chips would bring their real economic life down to two or three years. Burry puts at about $176bn the understated depreciation, and therefore the overstated profits, of the sector between 2026 and 2028. Nvidia responded with a memo to analysts disputing the calculation and rejecting any comparison with past accounting frauds. The divergence is already visible in practice: in 2025, Amazon shortened the useful life used on part of its servers, while Meta extended it further. The stakes go beyond the accounting quarrel. The sector plans about $1,000bn of AI spending over five years according to the same source. If the real useful life of the chips is closer to three years than to six, the residual value on which the guarantees rest melts well before the maturity of the leases that protect it. The guarantee shifts the risk, it does not cancel it A residual value guarantee does not make the depreciation risk disappear. It transfers it from the lender to the guarantor, most often the operator itself, a chip supplier, or a big technology name. The creditor gets a floor; the guarantor inherits a conditional commitment that triggers only in the bad scenario, when the asset's value collapses. The weakness of the arrangement is correlation. A credit guarantee works when defaults are independent of one another. A residual value, by contrast, depends on a common factor: the chip generation. A technological rupture, a new architecture that halves the value of the installed hardware, does not hit an isolated data center but the whole fleet at the same time. The guarantees would then trigger simultaneously, at guarantors often exposed to the same shock, since they are the very actors of AI. The risk has been made invisible on the balance sheet, it has not been diversified. A precedent the market prefers to forget The residual value guarantee is not an invention of the AI era. It has structured aircraft and rail finance leases for decades, where the lessee commits to cover the gap between the asset's resale price and an agreed value. Accounting moreover requires provisioning that gap as soon as the expected value falls below the guarantee, a warning signal that off-balance-sheet treatment tends to delay. The most useful reminder comes from autos. In 2008, the collapse of vehicle resale prices blew up the residual value assumptions of captive lenders. According to its annual report, GMAC recorded $1.2bn of impairments on the residual values of its leases in 2008, and its auto-finance division swung from a $1.3bn profit to a $753m loss in nine months. The ill was not new: WardsAuto recalls that the sector had already accumulated nearly $20bn of residual losses in the early 2000s, after artificially inflating those values to boost leasing volume. A floor set too optimistically, subscribed en masse, becomes a mass loss when the used market turns. The other reading One must avoid mechanically transposing 2008. Several arguments plead for the robustness of these structures, and it would be dishonest to leave them out. First, the economic life of the chips could be longer than Burry says. Earlier generations do not end up scrapped: they move down to less demanding inference tasks, a real second-hand market absorbs part of the fleet, and the appetite for compute exceeds supply for now. Nvidia makes this case in its response to analysts. Next, the quality of the guarantors and the contracts matters: when the rent is owed by a top-tier tenant, like the CoreWeave contract backed by Meta, the immediate credit risk is low, which partly justifies the investment-grade ratings. Finally, a residual value guarantee triggers only on non-renewal of the lease; as long as the operator needs the campus and pays its rent, the guarantee remains a dormant clause. The balance is therefore subtler than a mere omen of crisis. The question is not whether these structures are fraudulent, they are not, but how to measure a risk that has been moved out of sight, concentrated among a small number of correlated actors, and pegged to a life-span assumption the market itself does not settle. The signals to watch Four indicators will say whether the promise holds. Revisions to the depreciation period of servers at the hyperscalers, first, because a general shortening would validate the fast-depreciation thesis. The formation of a real secondary market for GPUs, next, the only tangible proof of an observable rather than assumed residual value. The release cadence of new Nvidia architectures, which sets the pace of obsolescence of the installed fleet. And the scale of the guarantee commitments subscribed by the big actors, compared with their capacity to honour them if several triggered at once. The credit that builds AI has found, in the residual value guarantee, the tool that makes infrastructure financeable at scale. The same tool concentrates a depreciation risk no one knows how to quantify, on assets whose useful life remains in dispute, at guarantors who are also the first exposed to the turn. The debt has left the balance sheets. The bet has stayed on them, whole. Sources - Bisnow, "Meta Pushes Its Largest Data Center Project Off Its Books With $27B JV" (Beignet Investor SPV, $27.3bn of notes at 6.581%, Blue Owl 80% / Meta 20%, operating rent): https://www.bisnow.com/national/news/data-center-capital-markets/meta-pushes-its-largest-data-center-project-off-its-books-with-27b-joint-venture-131490 - Global Data Center Hub, "Meta + Blue Owl's $27B Bet" (sixteen-year residual value guarantee, repayment of investors on non-renewal): https://www.globaldatacenterhub.com/p/meta-blue-owls-27b-bet-is-this-the - Quinn Emanuel, "AI Data Center Financing and Litigation Risks" (legal risks of off-balance-sheet structures and the exit value): https://www.quinnemanuel.com/media/4dzkfccz/client-alert-ai-data-center-financing-and-litigation-risks.pdf - CoreWeave, press release on the $8.5bn DDTL 4.0 facility, rated A3 / A (low), first investment-grade GPU-backed financing, $14.2bn Meta contract: https://investors.coreweave.com/news/news-details/2026/CoreWeave-Closes-Landmark-8-5-Billion-Financing-Facility-Achieving-First-Investment-Grade-Rated-GPU-backed-Financing/default.aspx - Quartz, "GPU-collateralized debt explained" (cost of capital from 15% in 2023 to ~5.9% in 2026, markdown of about 50% in three years, implicit bets on value and utilisation): https://qz.com/gpu-collateralized-debt-ai-neocloud-coreweave-financing-risks-050526 - Forbes, "GPU Debt Has Gone Investment Grade. Here's Who Holds The Risk" (June 2026): https://www.forbes.com/sites/daraabasiita/2026/06/09/gpu-debt-has-gone-investment-grade-heres-who-holds-the-risk/ - CNBC, "The question everyone in AI is asking: How long before a GPU depreciates?" (14 Nov 2025: 5-6 year depreciation vs 2-3 year life, Burry estimate of $176bn understated depreciation 2026-2028, Nvidia memo, Amazon shortens / Meta extends, ~$1,000bn of AI capex): https://www.cnbc.com/2025/11/14/ai-gpu-depreciation-coreweave-nvidia-michael-burry.html - Bloomberg, "Apollo Wraps Up $35 Billion Debt to Buy AI Chips for Anthropic" (TPU vehicle, residual value guarantees from Broadcom, payment guarantees from Google): https://www.bloomberg.com/news/articles/2026-06-05/apollo-wraps-up-35-billion-debt-to-buy-ai-chips-for-anthropic - GMAC LLC, annual report (Form 10-K) FY2008, SEC ($1.2bn of impairments on residual values in 2008, loss at the auto-finance division): https://www.sec.gov/Archives/edgar/data/0000040729/000119312509039567/d10k.htm - WardsAuto, "The Rise and Fall of Automotive Leasing" (nearly $20bn of residual losses in the early 2000s, inflated residual values): https://www.wardsauto.com/finance-insurance/the-rise-and-fall-of-automotive-leasing This article is journalistic analysis and does not constitute investment advice. Market data is cited as of the date of its sources. ============================================================================ ANALYSIS: Basel III in reverse: US regulators hand capital back to the banks URL: https://l0g.fr/en/analysis/basel-iii-rollback-us-regulators-bank-capital/ Canonical French source: https://l0g.fr/posts/bale-iii-rollback-regulateurs-us-capital-bancaire/ Date: 2026-07-15 (reviewed 2026-07-15) Topics: banks, regulation, systemic risk, private credit, fed, us politics ---------------------------------------------------------------------------- A prudential reform rarely ends up as the opposite of its stated intent. Yet that is the fate of the final leg of the Basel III accords in the United States. Designed after 2008 to strengthen the capital of the largest banks, the "Basel III endgame" has been turned into an instrument of capital relief, after a change of leadership at the Fed and sustained pressure from the industry. The move is not confined to this text. It extends to the leverage ratio and the stress tests, forming a coordinated loosening of the constraints that weighed on systemic banks. From +19% to net relief The trajectory is striking. The first proposal, published in July 2023 under Fed Vice Chair for Supervision Michael Barr, aimed to raise the capital of the largest banks by about 19%, as Brookings recalls. Facing an outcry from the industry and Congress, a September 2024 re-proposal had already brought the increase down to around 9%. The arrival of Michelle Bowman at the Fed's supervision post, in June 2025, tipped the file. On 19 March 2026, the federal agencies unveiled an entirely recalibrated version. Officially, it is meant to be "capital-neutral" and better aligned with actual risk. In practice, according to the law firm Simpson Thacher, it delivers net relief of about $87.7bn in common equity, and cuts the aggregate requirement of the largest banks by roughly 6%. The text passed on a Fed vote of six to one. The dissenting voice is that of Michael Barr, architect of the original version, who flagged, according to Freshfields, more than twenty material downward deviations from the international Basel standard. The consultation runs until 18 June 2026, for finalisation expected in late 2026 and entry into force in 2027. The leverage ratio loosened: the eSLR reform Risk-weighted capital is not the only lock eased. The enhanced leverage ratio, the eSLR, which imposes on systemic banks a capital floor independent of risk weighting, was reformed in parallel. In a proposal of 25 June 2025, the Fed, presented by Michelle Bowman, proposed recalibrating the eSLR buffer to set it at half of each bank's systemic surcharge, instead of a fixed 2% flat rate. The rule was finalised in late 2025, for application on 1 April 2026. The scale of the relief shows in its distribution. According to the FDIC, the recalibration reduces required Tier 1 capital by about 1.4%, or $13bn, at the holding-company level, but by 27% on average, or $213bn, at the level of the bank subsidiaries. The stated aim is to make the eSLR a backstop rather than a binding constraint, and to give banks back capacity for activities deemed low-risk: intermediation of the Treasury market and repo financing. One nuance deserves noting, because it tempers the most alarmist reading: unlike the temporary regime of 2020, the reform does not exclude Treasuries or reserves from the ratio's denominator; it has merely put that exclusion out for comment. Stress tests with less bite The third pillar of the loosening touches the stress tests, which set the stress capital buffer and, in turn, the banks' ability to pay dividends and buy back shares. Following a lawsuit filed in late 2024 by the industry, the Fed agreed to open its scenarios and models, long opaque, to public comment. The proposal of late October 2025, welcomed by the Bank Policy Institute, also plans to smooth the results over time to reduce the volatility of requirements from one year to the next. In parallel, a revision of the systemic surcharge, planned by Bowman according to Sullivan & Cromwell, is set to lower requirements modestly further. Each brick, taken in isolation, looks technical. Added together, they hand banks a substantial capital margin, partly destined for shareholders. The regulators' bet The official justification is not baseless, and it deserves to be taken seriously. The first argument is competitive neutrality. By raising the cost of capital on certain exposures, the 2023 version mechanically pushed activity toward the less-regulated non-bank sector. Easing the constraint would let banks stay in the credit game, notably mortgage credit, rather than ceding ground to funds. The second argument concerns the Treasury market. A leverage ratio that penalises holding safe assets discourages banks from intermediating US debt, at the risk of thinning liquidity when it is most needed. The March 2020 episode, when the Treasury market dislocated for want of dealer balance-sheet capacity, serves as the reference for this reasoning. Recalibrating the eSLR to free up that capacity aims to avoid a repeat of such a freeze, an objective that even cautious observers deem legitimate. The objections The opposite reading is just as argued, and it begins inside the Fed itself. Barr's dissent is no mere formality: to flag more than twenty downward deviations from the Basel standard is to say that the reform departs from the international consensus built after 2008. The core criticism is the timing. Easing capital just as the credit cycle is mature, valuations are stretched and private-credit defaults are rising, is to remove a shock absorber right before it is needed. Procyclicality is the heart of the risk. Capital requirements that ease at the top of the cycle leave a thinner cushion when the turn comes, exactly when losses materialise. The same rules, now looser, then become hard to tighten in a hurry without amplifying the credit contraction. Today's loosening mortgages tomorrow's room for manoeuvre. The private-credit paradox The subtlest point lies in the interaction with the non-bank sector. The hard 2023 version was called, by the ABA Banking Journal, a gift to private credit: by raising the cost of bank capital, it pushed credit toward funds escaping the same oversight, a shift we documented in the migration of credit risk. The 2026 relief could, in theory, reverse part of that movement and bring activity back into the bank perimeter, better capitalised and better supervised. The effect remains ambiguous, however, because the two worlds are now linked. US banks had lent close to $300bn to the private-credit sector by mid-2025 according to Moody's, a subject developed in our stocktake of private credit at mid-2026. Handing capital back to banks does not cut this thread; it may even lengthen it, if the freed margin funds more lines to non-bank actors. The risk is not simply repatriated into a safer compartment: it circulates between the two, and the reform acts on only one end of the chain. The signals to watch A few markers will say what face this new regime takes. The final Basel III endgame text expected in late 2026, first, and the real scale of the relief once the consultation closes. The volume of the large banks' share buybacks next, which will measure how much of the freed margin is returned to shareholders rather than retained. The trajectory of financing lines extended to private credit, to see whether the returned capital feeds the non-bank sector. And the resilience of the Treasury market under stress, the only real test of the regulators' central argument. The bet is clear, even if it is not stated this way. The authorities are wagering on banks freer to move, able to intermediate sovereign debt and win back ground from shadow credit, without the lower cushions being paid for at the next shock. The opposite bet, that of Barr and part of the economics profession, is that a financial system stripped of its capital at the top of the cycle finds out too late, when the cushion is missing. Between the two lies precisely the definition of a shock no one knows how to date. Sources - Brookings, "What is bank capital? What is the Basel III Endgame?" (increase of about 19% targeted by the July 2023 proposal): https://www.brookings.edu/articles/what-is-bank-capital-what-is-the-basel-iii-endgame/ - Simpson Thacher, via CLS Blue Sky Blog, "Basel III Endgame Evolution" (19 March 2026 re-proposal, net relief of about $87.7bn, aggregate decline of about 6%, Fed vote 6-1): https://clsbluesky.law.columbia.edu/2026/04/24/simpson-thacher-discusses-basel-iii-endgame-evolution/ - Freshfields, "Basel III Endgame, Take Two" (Barr's dissent, more than twenty downward deviations from the Basel standard, consultation timeline): https://www.freshfields.com/en/our-thinking/blogs/a-fresh-take/basel-iii-endgame-take-two-8-key-takeaways-from-the-federal-banking-agencies-c-102mnm3 - Bloomberg Professional Services, "Fed remarks point to capital-neutral Basel III Endgame in 2026": https://www.bloomberg.com/professional/insights/financial-services/fed-remarks-points-to-capital-neutral-basel-iii-endgame-in-2026/ - Federal Reserve, statement by Vice Chair for Supervision Michelle W. Bowman on the eSLR proposal, 25 June 2025 (recalibration to half the systemic surcharge): https://www.federalreserve.gov/newsevents/pressreleases/bowman-statement-20250625.htm - FDIC, "Final Rule to Modify the Enhanced Supplementary Leverage Ratio" (reduction in required Tier 1 of about 1.4% / $13bn at the holding level and 27% on average / $213bn at the subsidiary level, capacity for Treasury intermediation): https://www.fdic.gov/news/speeches/2025/final-rule-modify-enhanced-supplementary-leverage-ratio - Federal Register, final eSLR rule, 1 December 2025 (Treasuries and reserves remain included in the denominator, unlike the 2020 temporary regime): https://www.federalregister.gov/documents/2025/12/01/2025-21626/regulatory-capital-rule-modifications-to-the-enhanced-supplementary-leverage-ratio-standards-for-us - Bank Policy Institute, "Fed Proposal Marks Progress in Improving Stress Test Transparency" (scenarios and models opened to comment, smoothing of results, 2024 lawsuit): https://bpi.com/fed-proposal-marks-progress-in-improving-stress-test-transparency/ - Sullivan & Cromwell, "Fed Vice Chair Bowman Previews Basel III, G-SIB Surcharge & Revised Standardized Approach Proposals" (modest decrease in the systemic surcharge): https://www.sullcrom.com/insights/memo/2026/March/Fed-Vice-Chair-Bowman-Previews-Basel-III-GSIB-Surcharge-Proposals - ABA Banking Journal, "The Basel III endgame proposal: Yet another gift to private credit funds" (November 2023, shift of credit toward the non-bank sector): https://bankingjournal.aba.com/2023/11/the-basel-iii-endgame-proposal-yet-another-gift-to-private-credit-funds/ This article is journalistic analysis and does not constitute investment advice. Regulatory data is cited as of the date of its sources. ============================================================================ ANALYSIS: Stablecoins, the marginal buyer of US Treasuries URL: https://l0g.fr/en/analysis/stablecoins-the-marginal-buyer-of-us-treasuries/ Canonical French source: https://l0g.fr/posts/stablecoins-acheteur-marginal-bon-du-tresor/ Date: 2026-07-15 (reviewed 2026-07-15) Topics: stablecoins, credit, treasury, dollar, systemic risk, crypto ---------------------------------------------------------------------------- A stablecoin is not really a cryptocurrency. It is a digital wrapper around a portfolio of government debt. Each token in circulation is supposed to be backed by a dollar of reserves, and since the summer of 2025 US law requires those reserves to take mainly the form of short-dated Treasury bills. The result went almost unnoticed: a sector born on the margins of the financial system has become one of its largest buyers of sovereign debt. At the end of July 2026, the stablecoin market weighed about $303bn, of which $184bn for Tether's USDT and $73bn for Circle's USDC, according to DefiLlama tracking data. Behind these tokens sits Treasury debt, and a lot of it. The law that anchors the token to short-dated debt The GENIUS Act, enacted on 18 July 2025, sets the first framework for payment stablecoins in the United States. It requires one-for-one coverage by safe, liquid assets. The law firm Arnold & Porter details the six permitted categories: coins and notes, insured bank deposits, Treasury bills, notes or bonds with a remaining maturity of no more than 93 days, repurchase agreements backed by those same securities, money market funds invested in those assets, and central-bank reserves. The 93-day cap is no detail. It is meant to exclude interest-rate risk: a long-dated security loses value when rates rise, and that latent loss can knock the token off its peg. As Spark notes, this is a direct lesson from the 2023 collapse of Silicon Valley Bank, whose balance sheet held long-dated debt at a loss. The law therefore forces reserves toward the shortest, most liquid segment of the curve, that of the Treasury bill. This constraint makes issuers heavyweight creditors. According to the reserve analysis relayed by Spark, Tether carries more than $141bn of exposure to US debt, which would place it among the twenty largest holders of Treasury securities worldwide, on a par with mid-sized countries. The retail token has become, without saying so, an instrument for financing the federal government. A quiet pillar of debt demand The Treasury itself has taken the measure of the phenomenon. In its April 2025 work, the Treasury Borrowing Advisory Committee estimated that about $120bn of Treasury bills already served as collateral for stablecoins, and that a market grown to $2,000bn by 2028 would mobilise more than $1,000bn of bills. Treasury secretary Scott Bessent himself has spoken of additional demand potentially reaching $2,000bn in the coming years. A Forbes projection goes further: stablecoins could overtake China among the holders of US debt as early as 2028. The interest for Washington is direct. While the Treasury issues short-dated debt massively to fund its deficits, a captive and growing source of demand for the Treasury bill eases the pressure on auctions and on yields. The calculation is the same as the one described in our reading of the GENIUS Act's bet on debt: to regulate stablecoins is also to buy oneself a buyer. The effect on short yields, measured The intuition of a demand that weighs on yields now has a measure. In a working paper published in August 2025, economists Rashad Ahmed and Iñaki Aldasoro, for the Bank for International Settlements, estimate the effect of stablecoin flows on three-month Treasury bill yields, on daily data from 2021 to 2025. A two-standard-deviation inflow into stablecoins lowers the three-month yield by 2 to 2.5 basis points. In periods of bill scarcity, when the available supply is thin, the same flow compresses the yield by 5 to 8 basis points, roughly double. The order of magnitude stays modest on the scale of a single session, but it is not zero, and it grows with the size of the sector. A demand of several hundred billion, set to double or triple, stops being a microstructure detail and becomes a factor in the formation of short yields, alongside Federal Reserve policy and the Treasury cash management described in our guide on net liquidity. The asymmetry that worries The sensitive point is not the average effect, it is its shape. The same BIS study reveals a clear asymmetry: outflows raise yields two to three times more than inflows lower them. An outflow of $3.5bn pushes the three-month yield up by 6 to 8 basis points, when an equivalent inflow pulls it down by only 2 to 2.5 points. The reason lies in the mechanics of redemption: an inflow invests without haste, an outflow forces the issuer to sell its bills quickly to honour the redemptions. The authors see in this a risk of forced sales in degraded market conditions, and an argument for strengthening issuers' liquidity risk management. The direction of the transmission deserves to be stated plainly, because it reverses the usual narrative. The customary worry is a shock coming from sovereign debt that would contaminate crypto. Here, the channel runs the other way: a panic on a retail token can propagate to the Treasury bill market, the core of the global financial system, through the liquidation of its reserves. The stablecoin becomes a bridge between two worlds once thought watertight. The precedent of March 2023 This scenario is not theoretical. It has already happened, on a small scale. On 11 March 2023, Circle revealed that $3.3bn of its reserves, about 8 percent of the total, were stuck at the failed Silicon Valley Bank. According to CNBC, USDC broke from its peg and fell to $0.87, a run on redemptions set in, and Circle had to suspend live redemptions over the weekend. The token only regained parity after the announcement of a public backstop and the resumption of redemptions on the Monday. A Federal Reserve note draws a simple lesson from the episode: a stablecoin backed by assets deemed safe can nonetheless suffer a run, because confidence in the token depends on the immediate liquidity of the reserve, not only on its quality. At the time, USDC weighed a fraction of its current size and the Treasury bill market felt nothing. The question posed today is one of scale: what happens when the issuer forced to sell in a hurry holds tens of billions of bills, in a market already tight on short-dated supply? The other reading The comparison with 2023, or with money market funds, calls for several nuances that argue for the robustness of the arrangement. First, the GENIUS Act's 93-day cap removes the interest-rate risk that felled SVB. A reserve in very short bills liquidates at or near par, with no valuation loss, which limits the gap between the exit price and the displayed value. Next, the Treasury bill is the most liquid asset in the world; even a forced sale finds a buyer, when a real-estate fund or a private-credit portfolio does not sell in a day. The captive demand of stablecoins is, finally, a real day-to-day stabiliser, absorbing a share of the Treasury's short-dated issuance and smoothing the formation of yields, as the BIS study notes for inflows. Yet the protection bears on the quality of the asset, not on the mechanics of the run. A stablecoin issuer is not a regulated money market fund: it holds neither the same liquidity buffers nor the ability to cap redemptions. In a panic, it sells, it does not gate. This is where the parallel with shadow banking reasserts itself: the risk has been lodged in a compartment that regulation is only beginning to equip. The signals to watch A few markers will say whether the promise of stability holds. The detailed composition of reserves, first, which the GENIUS Act requires to be published each month, with particular attention to the concentration at a single issuer, Tether. The state of short-dated bill supply, next, because it is in periods of scarcity that the effect of flows doubles. The effective rollout of the implementing rules, while the Federal Reserve had not yet proposed its own by spring 2026. And the presence, or absence, of redemption-management mechanisms in case of strain, the only real bulwark against a run. The stablecoin has pulled off a silent conversion: from speculative object, it has become a cog in the financing of the American state. This new respectability has a downside. By tethering hundreds of billions of dollars of tokens to the Treasury bill, a thread has been strung between the volatility of crypto and the stability of sovereign debt. As long as confidence holds, the thread supports the market. The day it snaps, it transmits the shock in the direction no one expected. Sources - DefiLlama, stablecoin market capitalisation (about $303bn as of 12 July 2026, USDT $184bn, USDC $73bn): https://defillama.com/stablecoins - Arnold & Porter, "Analyzing the GENIUS Act" (six categories of reserve assets, 93-day cap, one-for-one coverage): https://www.arnoldporter.com/en/perspectives/advisories/2025/07/new-stablecoin-legislation-analyzing-the-genius-act - Spark, "Inside Stablecoin Reserves" and "Treasury Bill Reserve Mechanics" (the SVB lesson behind the 93-day cap, Tether's exposure of more than $141bn, rank among the twenty largest holders): https://www.spark.money/research/stablecoin-treasury-bill-reserve-mechanics - U.S. Department of the Treasury, minutes of the Treasury Borrowing Advisory Committee, 29 April 2025 (about $120bn of bills backing stablecoins, more than $1,000bn if the market reaches $2,000bn in 2028): https://home.treasury.gov/news/press-releases/sb0122 - Forbes, "Why Stablecoins May Surpass China In U.S. Treasury Holdings By 2028": https://www.forbes.com/sites/jonegilsson/2025/05/05/why-stablecoins-may-surpass-china-in-us-treasury-holdings-by-2028/ - Rashad Ahmed and Iñaki Aldasoro, "Stablecoins and safe asset prices", BIS Working Paper no. 1270, August 2025 (two-standard-deviation inflow: -2 to -2.5 bp on the 3-month, -5 to -8 bp under bill scarcity; asymmetry: a $3.5bn outflow makes +6 to +8 bp, an equivalent inflow -2 to -2.5 bp; USDT contributes most): https://www.bis.org/publ/work1270.pdf - CNBC, "Stablecoin USDC breaks dollar peg after firm reveals it has $3.3 billion in SVB exposure" (11 March 2023, fall to $0.87, run on redemptions, weekend suspension): https://www.cnbc.com/2023/03/11/stablecoin-usdc-breaks-dollar-peg-after-firm-reveals-it-has-3point3-billion-in-svb-exposure.html - Federal Reserve, "In the Shadow of Bank Runs: Lessons from the Silicon Valley Bank Failure and Its Impact on Stablecoins": https://www.federalreserve.gov/econres/notes/feds-notes/in-the-shadow-of-bank-run-lessons-from-the-silicon-valley-bank-failure-and-its-impact-on-stablecoins-20251217.html - Brookings, Davidovic, Ghani and Moszoro, "The Rise of Stablecoins and Implications for Treasury Markets" (October 2025): https://www.brookings.edu/wp-content/uploads/2025/10/TheRiseofStablecoinsandImplicationsforTreasuryMarketsDavidovicGhaniMoszoro.pdf This article is journalistic analysis and does not constitute investment advice. Market data is cited as of the date of its sources. ============================================================================ ANALYSIS: Iran war: an inventory of the oil and gas ships and infrastructure struck URL: https://l0g.fr/en/analysis/iran-war-tankers-and-energy-infrastructure-inventory/ Canonical French source: https://l0g.fr/posts/guerre-d-iran-inventaire-des-navires-et-infrastructures-petroliers-et-gaziers-fr/ Date: 2026-07-14 (reviewed 2026-07-14) Topics: geopolitics, energy, iran, oil, gas ---------------------------------------------------------------------------- import InfographicBars from '../../components/infographic/InfographicBars.astro'; The energy dimension of this war reads not only in the price of the barrel or in the closure of the Strait of Hormuz, subjects we have covered elsewhere. It also inscribes itself in a concrete accounting, that of the facilities and hulls struck. This article proposes its inventory, at a given date and with a demand for method, because such a count is worth only as much as the quality and caution of its sources. A method, and its limits Two categories of target call for two levels of confidence. For infrastructure, the best sources are the investigations of Human Rights Watch, supported by satellite imagery and the statements of operators such as QatarEnergy, Shell or the National Iranian Gas Company. The degree of certainty is high there: you see the damage, you date it, you measure it. For ships, the material is more heterogeneous, and we have anchored it to first-order maritime sources. The reference authority is the UKMTO, the UK Maritime Trade Operations liaison office, which recorded at least 52 incident reports in the Arabian-Persian Gulf, the Strait of Hormuz and the Gulf of Oman by the end of May 2026, between attacks, suspicious activity and hijackings. The joint maritime information centre, the JMIC, rated the threat as critical at the height of the crisis, an attack being judged nearly inevitable, before lowering it to severe in June. The intelligence firm Ambrey and the protection and indemnity clubs complete this base. These sources are authoritative on the count and on the severity, but they often anonymise the vessels: identification by name, and above all by IMO number, the unique fingerprint of a hull, then comes from the specialist press and maritime intelligence, sometimes from belligerents' claims that no neutral investigator has verified. When these sources diverge, we favour the first order: thus the Skylight, which some compilations gave as two dead, is described by the UKMTO as having caused four injuries. Three cautions therefore apply. This inventory is one of what is documented, not a guarantee of exhaustiveness: some incidents may have escaped the count, others been counted twice under different names. The damage descriptions range from a lightly damaged vessel to a sunk ship, and this gradation matters as much as the raw count. Finally, attribution, who struck whom, is often claimed but rarely proven. We therefore list what is reported, weighting it by its source. The energy infrastructure The turning point came on 7 March 2026, with the first Israeli strikes on Iranian oil facilities, four depots around Tehran. The escalation peaked on 18 March with the attack on the South Pars gas field and the Asaluyeh facilities, then the Iranian retaliation on the Gulf's energy sites. South Pars, the world's largest gas field and the heart of the Iranian energy system, supplies about 80% of the country's gas and feeds nearly 79% of its electricity production. According to imagery analysed by Human Rights Watch and confirmed on the ground, several refineries in the complex were fire-damaged, refinery number 4, the largest, being almost entirely destroyed. The attack affected about 12% of Iranian gas production and interrupted gas deliveries to Iraq. In retaliation, Iran struck Ras Laffan, in Qatar, which alone provides nearly a fifth of world LNG. Human Rights Watch documented there, by satellite and via QatarEnergy and Shell, severe damage: liquefaction lines 4 and 6, 12.8 million tonnes a year and 17% of Qatari exports, and line 2 of the Pearl gas-to-liquids plant. Repairs are estimated at up to five years for the LNG. Other Gulf and Israeli sites were hit, as the table below summarises. | Site | Country | Date | Nature of damage (source) | |---|---|---|---| | Four oil depots, Tehran region | Iran | 7 March | First Israeli strikes on Iranian oil (Al Jazeera, HRW) | | South Pars gas field, refineries 3 to 7 | Iran | 18 March | Refinery 4 nearly destroyed, fire, ~12% of Iranian gas, ~100 Mm³/d halted (HRW, NIGC) | | Asaluyeh petrochemical complex | Iran | 18 March | Storage tanks and gas facilities damaged (Al Jazeera, HRW) | | Ras Laffan, LNG (lines 4 and 6) and Pearl GTL | Qatar | 18-19 March | Severe damage, 17% of Qatari exports, repairs up to 5 years (HRW, QatarEnergy, Shell) | | Two refineries | Kuwait | 18-19 March | Hit by Iranian fire (PBS) | | Gas facilities, Abu Dhabi | UAE | 18-19 March | Operations suspended after 13 missiles and 27 drones (Wikipedia South Pars) | | Haifa refinery | Israel | 18-19 March | Power outage of about 45 minutes (Wikipedia South Pars) | | Ports of Duqm and Salalah, fuel storage | Oman | March | At least one fuel tank damaged at Duqm (incident compilations) | The oil and gas ships The second front is maritime, and it was particularly intense from the opening of the conflict, when traffic had not yet been rerouted. Daily traffic in the strait fell from about 138 vessels to a handful per day by early June, according to the UKMTO. The table below lists the oil- and gas-related vessels, tankers, chemical, bitumen or liquefied-petroleum-gas carriers, reported struck, seized or disabled. The IMO numbers, the unique fingerprint of each hull, are shown when a first-order source confirms them. Container ships, bulk carriers and others, also numerous among those targeted, do not appear here, the inventory focusing on the subject's scope. | Vessel | Flag | Type | Date | Location and nature (source) | |---|---|---|---|---| | Skylight (IMO 9330020) | Palau | tanker | 28 February | Projectile 5 nm north of Khasab; 4 injured, evacuation (UKMTO) | | MKD Vyom (IMO 9284386) | Marshall Islands | tanker | 28 February | Projectile above the waterline, fire contained (UKMTO) | | Sea La Donna (IMO 9380532) | Liberia | oil/chemical tanker | 28 February | Attack under investigation (UKMTO) | | Hercules Star | Gibraltar | tanker | 1 March | Minor damage | | Ocean Electra | Liberia | tanker | 1 March | Minor damage | | LCT Ayeh | UAE | tanker | 1-2 March | Damaged | | Stena Imperative | United States | products tanker | 2 March | Port of Bahrain, struck twice; 1 worker killed | | Athe Nova | Honduras | bitumen carrier | 2 March | Struck by drones | | Libra Trader | Marshall Islands | crude tanker | 3 March | Minor damage | | Sonangol Namibe | Bahamas | tanker | 4 March | Sea drone near Mubarak Al Kabeer port (Kuwait); oil spill | | Prima | Malta | oil/chemical tanker | 7 March | Drone attack (IRGC claim) | | Louis P | Marshall Islands | tanker | 7 March | Drone attack (IRGC claim) | | Safesea Vishnu | Marshall Islands | tanker | 11 March | Set ablaze; 1 dead (Iranian navy claim) | | Zefyros | Malta | tanker | 11 March | Set ablaze, abandoned | | Gas Al Ahmadiah | Kuwait | LPG carrier | 17 March | Projectile east of Fujairah, minor damage | | Parimal | Palau | chemical tanker | 18 March | Fire; captain missing | | Al Salmi | Kuwait | supertanker (VLCC) | 31 March | Drone at Dubai port; fire | | Aqua 1 | Panama | tanker | 1 April | Projectiles north of Doha | | Barakah | UAE (ADNOC) | tanker | 3 May | Two drones north of Fujairah; empty | | Ocean Koi | Barbados | tanker | 8 May | Seized by Iran | | Sevan | Panama | propane-butane carrier | 25 April | Seized by the United States | | Marivex | Palau | tanker | 8 June | Disabled by the United States, Gulf of Oman | | Settebello | Palau | tanker | 9 June | Disabled by the United States; 3 sailors killed | | Jalveer | Guinea-Bissau | tanker | 10 June | Disabled by the United States, Gulf of Oman | | Kiku | Panama | tanker | 27 June | Projectile in the strait | To this count are added several Iranian tankers seized by the US Navy as part of the naval blockade, including the Deep Sea, the Dorena, the Sevin, the Derya, the Tifani and the Majestic X, captured between April and May off India, Malaysia and in the Indian Ocean. These seizures stem from a blockade logic more than destruction, and must be distinguished from vessels damaged in combat. The lessons of the count Three lessons emerge from this survey. The first is chronological: the maritime violence was concentrated on the first weeks, in March, before the rerouting of traffic mechanically reduced the number of targets. The second is the double-acting nature of the strikes on infrastructure: each side targeted the other's energy core, Iran losing part of South Pars, Qatar part of Ras Laffan, in a destructive symmetry where the weapon is world supply itself. The third is human, often forgotten behind the tonnages: several sailors and at least one port worker perished in the recorded attacks on ships, the exact tolls varying by source, and the crews, often from South Asia, paid the heaviest price. This material cost translates into economic cost, which we have quantified elsewhere. The strike on South Pars and Ras Laffan sent the barrel jumping from $103 to $108 and European gas up 7% in a few hours, and the Ras Laffan damage, repairable in five years at most, will leave a durable imprint, as our analysis of the supply chain after Hormuz shows. The inventory of hulls and sites is not merely a macabre accounting: it is the physical fabric of a shock whose price still reads on the markets. Sources 1. Human Rights Watch, "Israel, Iran: Unlawful March Attacks on Energy Infrastructure", 22 April 2026: satellite analysis and operator statements, South Pars damage (refineries 3 to 7, refinery 4 nearly destroyed, ~12% of Iranian gas) and Ras Laffan (lines 4 and 6, Pearl GTL, 17% of Qatari exports, repairs up to five years): https://www.hrw.org/news/2026/04/22/israel-iran-unlawful-march-attacks-on-energy-infrastructure 2. Human Rights Watch, "Iran: Israel's Oil Depot Strikes Endanger Environment, Health", 14 April 2026: strikes on oil depots, health and environmental risks: https://www.hrw.org/news/2026/04/14/iran-israels-oil-depot-strikes-endanger-environment-health 3. Wikipedia, "2026 South Pars field attack": date and course of the 18 March strike, South Pars supplying 80% of Iran's gas and 79% of its electricity, Gulf retaliation, Haifa refinery, price impact: https://en.wikipedia.org/wiki/2026SouthParsfieldattack 4. Al Jazeera, "Iran oil facilities hit for first time as war with US-Israel enters day 9", 8 March 2026: first Israeli strikes on Iranian oil depots: https://www.aljazeera.com/news/2026/3/8/israel-strikes-irans-oil-facilities-for-first-time-as-war-enters-ninth-day 5. PBS News, "Iran intensifies attacks on Gulf energy sites after Israel struck its key gas field": Iranian retaliation on Gulf energy sites, Kuwaiti refineries: https://www.pbs.org/newshour/world/iran-intensifies-attacks-on-gulf-energy-sites-after-israel-struck-its-key-gas-field 6. Wikipedia, "2026 Strait of Hormuz crisis": consolidated table of merchant ships attacked, seized or disabled, with flags, types, dates and human toll, aggregating press and maritime security advisories: https://en.wikipedia.org/wiki/2026StraitofHormuzcrisis 7. ABC News, escalation of attacks on commercial vessels in the Strait of Hormuz: https://abcnews.com/International/attacks-strait-hormuz-intensify-iran-targeted-commercial-ships/story?id=130962627 8. NPR, "Tanker set ablaze after being struck by projectile in the Strait of Hormuz", 7 July 2026: https://www.npr.org/2026/07/07/g-s1-132265/tanker-attack-strait-of-hormuz 9. PBS News, "U.S. fires on and disables 2 more Iranian tankers as tensions rise in the Strait of Hormuz": Iranian vessels disabled by the US Navy: https://www.pbs.org/newshour/world/u-s-fires-on-and-disables-2-more-iranian-tankers-as-tensions-rise-in-the-strait-of-hormuz 10. UKMTO (UK Maritime Trade Operations), recent incidents and 2026 advisories: reference authority on the incident count, raising of the threat level, GNSS/AIS/VHF jamming: https://www.ukmto.org/recent-incidents 11. Skuld (protection and indemnity club), Gulf and Hormuz maritime security update: at least 52 incident reports to the UKMTO by late May 2026, JMIC threat levels (critical then severe), named vessels with IMO number (Skylight IMO 9330020, MKD Vyom IMO 9284386, Sea La Donna IMO 9380532), traffic down from about 138 to a handful of vessels a day, Ambrey's role: https://www.skuld.com/topics/port/port-news/asia/maritime-security-update-gulf-region--strait-of-hormuz-and-red-sea/ 12. Euronews, 27 June 2026, raising of the threat level in the Strait of Hormuz by the UK maritime agency after a tanker reported being struck: https://www.euronews.com/2026/06/27/uk-maritime-agency-raises-strait-of-hormuz-threat-level-after-oil-tanker-reports-being-str ============================================================================ ANALYSIS: The 2026 Iran war: anatomy of a global economic and political earthquake URL: https://l0g.fr/en/analysis/the-2026-iran-war-economic-political-earthquake/ Canonical French source: https://l0g.fr/posts/la-guerre-diran-2026-seisme-economique-politique/ Date: 2026-07-14 (reviewed 2026-07-14) Topics: geopolitics, macro, energy ---------------------------------------------------------------------------- Introduction: the most important shock since 1973 On 28 February 2026, the United States and Israel launched an air campaign against Iran's nuclear programme and ballistic capabilities. Four days later, on 4 March, Tehran declared the Strait of Hormuz closed and began attacking the Gulf's oil infrastructure. Within a few weeks, the world tipped into what the International Energy Agency called the "largest supply disruption in the history of the global oil market". The rest of the episode, from the fragile April ceasefire to the reopening deal, is followed in the situation report on Hormuz then in the market normalisation scenarios. For nearly three months, this war has no longer been merely a regional conflict. It has become a systemic event that simultaneously touches energy, food, finance, the domestic politics of every Western democracy, and world geopolitics. Historical comparisons point to the 1973 oil shock, the 1979 crisis, and more recently the Russian invasion of Ukraine, but none of these analogies fully captures the singularity of this crisis, which combines the closure of a strategic strait, attacks on the infrastructure of four OPEC producers at once, and a reconfiguration of alliances between great powers. This article proposes as rigorous a mapping as possible of the crisis's global economic and political consequences, drawing on the analyses of central banks, multilateral institutions, and the main geo-economic think tanks. It closes with several argued prospective scenarios to late 2026 and 2027. Part one: the unprecedented energy shock A historic supply rupture The Strait of Hormuz is a chokepoint through which about 27% of world seaborne trade in crude oil and petroleum products transits, according to Congressional Research Service report R45281 published on 11 March 2026. Its effective closure since 4 March 2026 produced immediate and cumulative effects. According to the estimates collected by the Dallas Fed in its analysis of 20 March 2026, and confirmed by the Atlas Institute, the combined production of Kuwait, Iraq, Saudi Arabia and the United Arab Emirates fell by 6.7 million barrels a day on 10 March, then by 10 million barrels a day on 12 March. That is a supply shock representing approximately 20% of world oil supply removed simultaneously from the market. By comparison, the 1973 OPEC embargo represented about 7% of world supply. The current crisis is therefore, in pure magnitude, about three times more violent than the shock of the 1970s. QatarEnergy, which operates the world's largest liquefied natural gas production site, declared force majeure on all of its exports. According to Bloomberg, sections of the Qatari gas complex suffered missile damage whose repair is estimated at five years. This gas dimension, often eclipsed by media attention on oil, is probably the most structurally serious: unlike oil, LNG has no alternative overland routes or a spot market as fluid. Prices: between $100 and $200 depending on the scenario At the time of writing (mid-May 2026), Brent hovers above $100 a barrel, with an intraday peak of $119 on 23 March, the highest since 2008. West Texas Intermediate (WTI) settled in a range of $94 to $98 depending on the modelled scenarios. Three institutional models converge on the short-term trajectory: - The Dallas Fed estimates that a closure of the strait removing 20% of world supply in the second quarter of 2026 takes the WTI average to $98 over the quarter. - The CEPR (Kilian-Zhou model) forecasts in a median scenario a WTI peak at $94 in April-May, staying above $80 over the whole of 2026. - Bloomberg Economics, via its SHOK model, considers that a Brent around $110 is compatible with contained growth, but that above $170, the impact on inflation and growth would double, producing a genuine stagflationary shock. Bloomberg sources report that US government officials and Wall Street analysts are now envisaging a $200 scenario if the crisis becomes entrenched beyond the summer. That would be an unprecedented threshold. Part two: world inflation and stagflation The United States: inflation back to 4% The US CPI for April 2026 (published on 12 May) came in at 3.8% year on year, the highest since May 2023, with core CPI at 2.8%. Gasoline prices jumped 28.4% year on year and energy prices 17.9% (Bureau of Labor Statistics). This is the sequence tracked in our piece on the return of US inflation. The CEPR estimates, in its structural model published a few days after the outbreak, that a one-quarter closure of the strait would add 0.6 percentage point to total US inflation over 2026 and 0.2 point to core inflation. The OECD is more pessimistic: its revised forecast takes US 2026 inflation to 4.2%, 1.2 points above pre-war forecasts. The most immediate effect on the bond market: the 30-year Treasury yield reached 5.12% on 15 May, the highest since May 2025. Markets now price a 44% probability of a Fed rate hike by December 2026, against 22.5% a week earlier. This is a complete reversal from the start of the year, when expectations pointed to at least two cuts. The euro area between technical recession and stagflation Europe is structurally more exposed than the United States: less of an own energy producer, more dependent on gas and oil imports, and economically more oriented toward energy-intensive manufacturing. The figures collected by S&P Global and published via Euronews on 23 April 2026 paint a grim picture: - The euro-area composite PMI moved back into contraction territory in April (the weakest performance since November 2024). - Industrial input-price inflation reached a 3.5-year high in Germany and a 3-year high in France. - The IMF, in its April 2026 World Economic Outlook, revised euro-area growth down to 1.1% for 2026 (against 1.4% in 2025), with Germany bearing the most severe revision (-0.3 point). - The ECB, in its economic bulletin no. 2 of 2026, projects harmonised inflation at 2.6% in 2026, with a peak at 3.1% in the second quarter of 2026. On 19 March 2026, the ECB paused its rate-cutting cycle and held them at 2%. On 30 April, it confirmed this status quo, noting explicitly that "the upside risks to inflation and downside risks to growth have intensified". According to the prediction markets recorded by Goldman Sachs (economist Niklas Garnadt), the probability of an ECB rate hike in 2026 now reaches 72%, against only a few percent before the Hormuz closure. The most marked slowdown concerns the United Kingdom, designated by several analyses as the hardest-hit major economy. British inflation could exceed 5% in 2026 according to European Commission forecasts, the highest in Europe. The world food crisis: 45 million more people in insecurity This is probably the most underestimated dimension of the crisis, because it unfolds with a lag relative to the energy shock. Three mechanisms converge. First mechanism: fertiliser. According to the International Food Policy Research Institute (IFPRI), the Gulf region accounted for 29% of world ammonia exports between 2023 and 2025 and 36% of world urea exports. Iran itself is the Gulf's largest urea exporter according to International Fertilizer Association estimates. Prices have already responded violently: the FOB price of granular urea in Egypt went from $400-490 a tonne before the war to about $700 a tonne in late March 2026, a 50% rise in a few weeks (source CNBC, March 2026, citing Chris Lawson at CRU and Sarah Marlow at Argus). Ammonia rose about 20%. Second mechanism: agricultural fuel. Energy costs for agricultural producers exploded, already partly passed through to wholesale prices but with a lag of about four months on retail prices according to the World Food Programme. Third mechanism: logistics. Transport routes were reconfigured, with ships having to route around the Gulf via the Cape of Good Hope. According to the Stimson Center, Asia-Mediterranean spot rates jumped to as much as $8,500 per FEU (40-foot container), carriers imposing emergency war surcharges. The WFP (World Food Programme) estimates that if the war continues beyond June 2026 with oil held above $100, the number of people in acute food insecurity could rise by 45 million worldwide. This figure is captured in a Center for Strategic and International Studies (CSIS) analysis of 7 April 2026. Most vulnerable regions identified by the FAO: India, Bangladesh, Sri Lanka, Egypt, Sudan, and most of sub-Saharan Africa. Africa imports more than 90% of its fertiliser according to University of Texas at Austin data cited by CNBC, and its fertiliser use had already fallen 25% in 2022 following the Russian invasion of Ukraine. The new crisis could reproduce this pattern on a larger scale. Capital Newspaper's analysis estimates that if the crisis lasts more than six months, African GDP growth could be cut by 0.2 point in 2026. Nearly thirty African currencies have already lost value since March 2026, a classic signal of capital flight to safe-haven assets. Part three: the geopolitical recomposition The calculated bet of China and Russia One of the most striking observations of the conflict is the absence of direct military support from China and Russia for Iran, despite the 25-year cooperation agreement signed between Beijing and Tehran in 2021 (which provided for $400bn of discounted Iranian oil in exchange for Chinese investment). The Peterson Institute for International Economics (PIIE), in its analysis of 30 March 2026, states the thesis explicitly: "the measured response [of Moscow and Beijing] is not a mistake. It is a strategic calculation: why interrupt a war waged by the United States while they bog down in a costly quagmire in the Middle East?" Several factual elements confirm this reading: - Iran supplies about 13% of China's oil imports, at a discount. But Beijing has favoured diversifying its sources rather than direct intervention. - The Atlantic Council (report of 25 March 2026) documents that China continues to supply Iran with dual-use components (drones, components for solid rocket fuels) without direct military commitment. - Russia draws a direct net benefit from the oil shock: its energy exports (to China, India, Turkey) sell at high prices, which funds its war economy in Ukraine. Several analysts consider that Russia could be the main geo-economic beneficiary of the crisis. This Chinese posture nonetheless carries a growing cost. The PIIE notes that Europe absorbs 15% of Chinese exports. A prolonged energy shock that tipped Europe into recession would crush Beijing's export orders and worsen the domestic real-estate crisis. According to standard models, Chinese GDP would fall about 0.5% for each 25% rise in oil. China is therefore betting that the United States will yield before it does. The collapse of the Gulf Cooperation Council model This is perhaps the most structurally profound geopolitical consequence. The economic model of the Gulf Cooperation Council (Saudi Arabia, the Emirates, Kuwait, Qatar, Bahrain, Oman) rested on three pillars: 1. The export of hydrocarbons via Hormuz 2. Massive food imports (80% of calories consumed in the GCC countries transit the strait) 3. American security protection All three pillars are shaken simultaneously. According to the Atlas Institute (March 2026), a "food supply emergency" unfolded as early as mid-March, with 70% of the region's food imports disrupted, and consumer price rises of 40 to 120% on staple products. Lulu Retail (one of the main regional distributors) resorted to emergency air transport for essential goods, economically unsustainable at scale. European defence commissioner Andrius Kubilius stated on 6 March 2026 that US military costs are over-stretched, with a shortage of key missile stocks, making the United States unable to provide military aid simultaneously to its Gulf allies and to Ukraine. This is a European institutional acknowledgment of an American strategic limit, unprecedented since 1945. The world energy order in mutation Several reconfigurations are already observable: - The crisis accelerates China's energy decoupling from the Middle East. Beijing is investing massively in Russian and Venezuelan (paradoxically) hydrocarbons and in domestic renewables. - OPEC+ is mechanically disorganised by the temporary exit of four of its major members. The Atlas Institute speaks of an "OPEC endgame", a bold thesis but one that deserves consideration. - The United States emerges as a strategic exporter of LNG and oil. US production is at historic highs according to Treasury secretary Scott Bessent (CNBC statement, May 2026), and the Emirates' exit from OPEC has freed up capacity. - Venezuelan oil returns to the market as part of a Trumpian policy of partial reintegration, with the explicit aim of diluting OPEC's influence and weakening Iranian and Russian revenues (Russia Matters analysis, January 2026). Part four: the domestic political consequences In the United States: the war as a catalyst for the midterms The November 2026 midterms are at the heart of US political strategy, and the first signals are unfavourable to the Trump administration. According to Left Voice (analysis of 13 May 2026) and the Politico coverage cited in this analysis, Republican strategists openly fear losing the midterms if the war's economic effects drag on. The reported quote: "We lose the midterms" if inflation and instability persist. Domestic polls show that a majority of Americans want the war to end, and that voters attribute to the administration responsibility for inflation and energy prices. The Democrats appear fragmented. Their strategy has evolved from a demand for briefings and legal justifications toward an economic attack: the Republican inability to protect households against rising costs. This is more consensual ground and lets them avoid appearing militarily weak. The stakes are macro-historical. A Democratic victory in the midterms would produce: - A legislative blockade on Trumpian priorities (notably the extension of the One Big Beautiful Bill Act tax cuts) - Parliamentary pressure to end the war - A potential geopolitical repositioning toward 2028 In Europe: political fragmentation and the rise of populism The European effect is more diffuse but probably more durable. The combination of stagflation + food crisis + potential migration crisis (flows from North Africa and the Sahel could accelerate according to the Stimson Center) is the historic cocktail for the rise of populist forces. In France, the budgetary deterioration (deficit 5.4% of GDP in 2025, debt 117.4% of GDP, interest charge of €78bn projected for 2026) now combines with an exogenous inflationary shock. The 10-year OAT went from 3.4% in early 2026 to 3.81% on 15 May, its highest since 2009. The OAT/Bund spread holds at 70 basis points, up but far from the stress levels of 2011 (225 bp). The Bayrou government seeks to bring the deficit below 4.6% in 2026 and below 3% in 2029, targets the markets clearly do not take at face value. This dynamic is analysed further in French rates and the no-Frexit thesis. In Germany, the downward revision of growth (-0.3 point in 2026 and 2027 according to the IMF) could weaken the ruling coalition. German industry, already affected by the energy transition and Chinese competition, takes an additional shock to its input costs. In Italy, growth remains stuck at 0.5% annually over 2026 and 2027 according to the IMF, the weakest base in the euro area. Italian debt exceeds 140% of GDP, and any sustained rise in European long rates mechanically threatens fiscal sustainability. In the United Kingdom, the bond move of spring 2026 was particularly violent. 10-year gilts hit a high since 2008, and the country was described as the "worst-hit major economy" by several analyses, owing to its energy dependence and vulnerability to capital flows. The Global South and the risk of instability This is probably the dimension that could produce the most dramatic political effects in the medium term. The Philippines declared a state of emergency on 24 March 2026 owing to the combination of a fuel crisis and a transporters' strike. Zimbabwe, Pakistan, Bangladesh, Nigeria, and Vietnam face severe shortages according to the compiled sources (the Wikipedia 2026 Iran war fuel crisis page lists these declarations). Egypt is doubly exposed: as a massive food and energy importer, and as operator of the Suez Canal, whose traffic is disrupted. The historical precedents (Arab Spring 2011, triggered in part by food prices) are watched closely by chancelleries. India, the world's ninth-largest economy, is doubly exposed: it is the world's largest urea importer according to the IFPRI, and its energy depends significantly on the Gulf. The internal political consequences for the Modi government are monitored ahead of the critical 2026 state elections. Part five: projections to late 2026 and 2027 Any projection on an ongoing conflict is by nature speculative. We propose three argued scenarios, explicitly weighting the assumptions. Scenario 1 · Diplomatic resolution by summer 2026 (estimated probability: 30-40%) Assumptions: A Russo-Chinese compromise, with Pakistani or Omani mediation, produces a durable ceasefire, with a gradual reopening of the Strait of Hormuz by September 2026. US secondary sanctions on Iran are partly lifted in exchange for halting the military nuclear programme. Economic consequences: - Brent falls back toward $70-75 by the end of 2026 - US and euro-area inflation eases toward 2.5-3% in Q4 2026 - The Fed and the ECB resume a moderate cutting cycle in 2027 - The French 10-year OAT falls back toward 3.4-3.5% - Avoidance of a technical recession in the euro area - Gradual recovery of growth from mid-2027 Political consequences: - Trump capitalises politically on the return of stability (potential rebound in the midterms) - Europe comes out weakened but without systemic rupture - Iran comes out durably weakened regionally, but the regime survives Scenario 2 · Stalemate and prolonged stagflation (estimated probability: 40-50%) Assumptions: The conflict bogs down without clear resolution. Neither complete military victory nor compromise materialises. The strait stays partly blocked, Iran maintains asymmetric nuisance capabilities, and US secondary sanctions crumble in the face of European weariness and Asian opposition. Economic consequences: - Brent oscillates durably between $90 and $120 - US inflation settled at 3.5-4% over 2026 and 2027 - The euro area enters technical recession in Q3 or Q4 2026 (Germany and Italy in the lead) - ECB rate hikes in 2026 (72% probability per Goldman Sachs) - US 30-year around 5.5%, 10-year OAT at 4-4.3% - Prolonged food crisis affecting 45 million more people - Significant risk of sovereign defaults in several emerging countries Political consequences: - Midterm defeat for the Trump administration: a Democratic takeover of the Senate and the House becomes likely - Rise of populist parties in Europe: France (RN/LFI), Germany (AfD), Italy held on the hard right - Collapse of several vulnerable regimes in the Sahel and North Africa - Russia and China consolidate their position as geopolitical alternatives Scenario 3 · Escalation and systemic shock (estimated probability: 15-25%) Assumptions: Iran launches asymmetric operations against Saudi and Emirati energy infrastructure beyond current thresholds. The Shia arc (Hezbollah, Houthis, Iraqi militias) is fully activated. A crisis within OPEC+ breaks the price-cohesion arrangement. Mismanagement leads to an incident involving a US aircraft carrier or a large tanker in Hormuz. Economic consequences: - Brent toward $170-200 per Bloomberg Economics assumptions - US inflation above 5%, euro area toward 4-5% - Multiple Fed rate hikes (potentially +100 to 150 bp cumulative in 2026) - A synchronised global recession comparable to 2008-2009 - A stock-market collapse of 25-40% on developed indices - Debt crises in several emerging countries, possible sovereign defaults - Major humanitarian crisis: potential doubling of the number of people in food insecurity Political consequences: - Legitimacy crisis for the Trump administration - Possible revolutionary waves in the Global South - Accelerated European rearmament, possible activation of EU mutual-assistance clauses - Accelerated reorientation toward a multipolar world order with a durable weakening of the dollar as a reserve currency Conclusion: a revealing crisis more than an inaugural one Beyond its immediate effects, the 2026 Iran war reveals structural fragilities that the Western world had preferred to ignore for fifteen years. First revelation: the end of the "peace dividend". The 1991-2022 period was historically exceptional in its geopolitical stability and contained inflation. The return to a world of recurrent supply shocks (Covid, Ukraine, tariffs, Iran) is the new regime, not a passing anomaly. Economist Daleep Singh (PGIM, former deputy National Security Advisor under Biden) puts it bluntly: "we have had one supply shock after another for five years. These are shocks that stack up and suggest a structurally inflationary environment." Second revelation: the erosion of American strategic pre-eminence. European commissioner Kubilius's statement on "over-stretched American military costs" is probably the most significant institutional admission since 1945. The United States can no longer wage two major wars simultaneously. This is a fundamental strategic fact that will restructure alliances for the next decade. Third revelation: the centrality of the Global South in the geopolitical equation. The food and energy shock hits first the poorest countries, which bear no responsibility for the conflict. The resilience or collapse of these societies will determine the scale of migration flows, the evolution of political regimes, and the diplomatic orientation of dozens of countries. This is the terrain where China and Russia have been scoring major strategic points since 2022. Fourth revelation: the fragility of the European model. The euro area enters this crisis with high public debts, listless growth, an ageing population, and still-fragile energy credibility. Without a significant rebound in the coming months, Europe could come out of this decade as the main geopolitical loser, behind the United States, China, and even Russia in relative terms. The 2026 Iran crisis will probably not be the moment the world order tips over. But it will very probably be the moment we realise it had already tipped. --- Primary sources: - Federal Reserve Bank of Dallas, "What the closure of the Strait of Hormuz means for the global economy", 20 March 2026. - CEPR (Kilian, Plante, Zhou), "Quantifying the impact of the Iran war on US inflation", May 2026. - European Central Bank, "Economic Bulletin Issue 2, 2026", April 2026. - Peterson Institute for International Economics, "How Russia and China are winning the war in Iran", 30 March 2026. - Bloomberg Economics, "Iran War: How High Could Oil Prices Get with Strait of Hormuz Closure?", March 2026. - Atlantic Council GeoEconomics Center, "From drones to rocket fuel, China and Russia are helping Iran through supply chains", 25 March 2026. - International Food Policy Research Institute, "The Iran war's impacts on global fertilizer markets and food production", April 2026. - Center for Strategic and International Studies, "Iran, Fertilizer, and Food Security", 7 April 2026. - Stimson Center, "Impacts of the Iran War on North Africa, the Sahel, and the Mediterranean", April 2026. - Congressional Research Service, "Iran Conflict and the Strait of Hormuz", 11 March 2026. - U.S.-China Economic and Security Review Commission, "China-Iran Fact Sheet", 16 March 2026. - Euronews, "Iran war effects on Europe: Is a recession already unfolding?", 23 April 2026. - Reuters, "ECB keeps rates on hold and warns [...] ============================================================================ ANALYSIS: Hormuz: the supply chain takes the hit, the bill is already here URL: https://l0g.fr/en/analysis/hormuz-supply-chain-the-bill-is-already-here/ Canonical French source: https://l0g.fr/posts/ormuz-la-chaine-d-approvisionnement-encaisse-la-facture-est-deja-la/ Date: 2026-07-14 (reviewed 2026-07-14) Topics: geopolitics, energy, supply chain, macro, commodities ---------------------------------------------------------------------------- There are two clocks in a chokepoint crisis, and they do not tick at the same speed. The first is the price clock, financial, which calms the moment a deal is signed. The second is the clock of holds, tanks and quays, physical, which takes months to recover. The mid-June US-Iran agreement, analysed in our three reopening scenarios, stopped the first. The second keeps running. In early July 2026, the war premium has left the barrel, but the global supply chain is still digesting the largest logistical rupture in its recent history. A narrow strait, an outsized share of the world The Strait of Hormuz is only about fifty kilometres wide, with navigation channels barely a few kilometres across. Through this bottleneck passes a share of the world economy out of all proportion to its size. The International Energy Agency measured in 2025 an average flow of about 20 million barrels of crude and refined products a day, nearly a fifth of world oil consumption and about a quarter of oil carried by sea. On gas, nearly 20% of world liquefied natural gas trade depends on it, Qatar routing almost all of its exports through it. The dependence does not stop at hydrocarbons. According to the World Economic Forum, up to a third of world trade in fertiliser raw materials passes through this strait, not counting methanol, aluminium, sulphur or graphite. Hormuz is not only an oil tap: it is an artery of planetary agricultural and industrial production. It is this concentration that makes the strait indispensable: there is no plan B on the scale needed. The Saudi and Emirati pipelines that bypass Hormuz cap, according to the IEA, between 3.5 and 5.5 million barrels a day. Against a flow of 20 million, that leaves a potential deficit of 14 to 16 million barrels a day that no infrastructure can absorb. The point is developed in our state of the blocked chokepoint. February, the flash closure The sequence was brutal. On 28 February 2026, US and Israeli forces strike Iran. In less than forty-eight hours, Tehran threatens navigation and the strait de facto closes to commercial traffic. The planet's top four container-ship operators, Maersk, MSC, CMA CGM and Hapag-Lloyd, suspend their transits. From 5 March, hull-and-machinery insurers stop covering the passage, and tanker traffic falls to near zero. Over the duration of the blockage, about 95% of tanker transits and nearly 99% of LNG were rerouted, an order of magnitude confirmed by IEA director Fatih Birol, who calls it the largest supply rupture in the history of the oil market. The difficulty is aggravated by a second maritime front. For the first time in modern history, the two great Middle East corridors are blocked at the same time: the Red Sea, already disrupted, was running at only 49% of its pre-crisis capacity. Ships linking Asia to Europe or the US East Coast had to route around the Cape of Good Hope, adding 10 to 14 days to each rotation. The cost of the detour A two-week detour is paid not only in time. It is paid in fuel, insurance and immobilised capacity, and this bill has already fed into transport prices. The war-risk premium to cross the strait, around 0.125% of the vessel's value before the strikes, had already climbed to between 0.2 and 0.4% per passage in the days before 28 February, before coverage disappeared altogether. Behind the indices, there are crews. At the height of the blockage, in early May, more than 1,550 merchant vessels were immobilised in the zone, with some 22,500 seafarers stuck on board, according to maritime organisations' counts. Insurance is the real lock. Without hull-and-machinery cover or protection and indemnity, no owner commits a vessel worth several hundred million dollars, whatever the freight rate offered. It is this insurance impossibility, even more than the military risk itself, that emptied the strait: you can charter a ship to brave threats, you cannot sail it without an insurer behind it. Container freight followed. The benchmark index of rates from Shanghai, the Shanghai Containerized Freight Index, reached 2,572 points in the week to 30 May 2026, up 16% in a week and double its level of late February, just before the strikes, according to Lloyd's List. Shippers see the surcharges stack up: a war-risk surcharge of up to $1,500 per TEU on Gulf-linked routes, an emergency bunker surcharge triggered by the doubling of the price of very-low-sulphur marine fuel, and an emergency freight hike of $3,000 per FEU or more for Persian Gulf freight. The mechanism is simple to follow: carriers pass on to their customers the more expensive fuel and the extra days at sea. These extra costs end up in the price of imported goods, with the usual lag of a few weeks to a few months between a ship's deck and the shelf label. The gas wave reaches Europe No region illustrates the propagation better than the European gas market. Qatar, whose cargoes take Hormuz, suspended part of its production under force majeure, removing at a stroke nearly a fifth of world LNG supply. The European benchmark price, the Dutch TTF, jumped 35% in a single session to exceed €60 per megawatt-hour, and 76% over the week. The prolongation scenarios are dizzying. A three-month halt of Qatari exports would take the TTF to around €155 per megawatt-hour, triple the pre-crisis level near €50. A six-month blockage would raise fears, according to analysts cited by the European press, of a 2022-style squeeze or worse, with averages around €160 and possible spikes beyond €200. The IEEFA estimates that the Hormuz disruption alone puts about 10% of Europe's LNG imports at stake. A shock absorber exists: the new American terminals take North American production to a record and partly offset the Middle Eastern losses. But since LNG often sets the marginal price in Europe, the European benchmark price should still climb about 25% over 2026. The absorber limits the damage, it does not cancel it. Fertiliser and food: the shock that reaches the field The least visible consequence from Europe is perhaps the heaviest elsewhere. Since up to a third of fertiliser raw materials transit Hormuz, deliveries of ammonia and nitrogen compounds contracted at the worst moment of the agricultural calendar. In Bangladesh, the closure of several state fertiliser plants disrupted national production during the winter-rice season, creating immediate pressure on farmers. The United Nations warns that, if the crisis drags on, 9.1 million more people in Asia could fall into acute food insecurity. The timing is cruel: the disruption coincides with decisive planting windows. A farmer facing more expensive or unavailable fertiliser cuts inputs, sows less or switches crops, all decisions that will weigh on yields in the months to come. The energy shock thus turns, with a lag, into a food shock. Beyond oil and gas: chemicals and batteries The strait does not only carry energy and fertiliser. The World Economic Forum lists at least four other disrupted industrial links, often ignored because they are invisible to the end consumer. Methanol first, a base building block of countless plastics, paints and solvents, of which the Gulf is a major supplier. Aluminium next, whose regional flows feed industry, construction and packaging. Sulphur, that refining by-product used to make sulphuric acid and, at the end of the chain, phosphate fertilisers, closing the loop with the agricultural crisis. Graphite finally, a key material for lithium-ion battery anodes, whose scarcity hits the energy transition and electric-vehicle manufacturing head-on. Each of these links has its own propagation lag, from ship to finished product, but all tell the same mechanic: a transport shock concentrated on a single point diffuses, step by step, to value chains with no apparent connection to oil. A European battery plant, an Asian paint maker and an African fertiliser producer discover that they share, without knowing it, the same bottleneck. The shock is already in the figures The usual objection would be to say that all this remains theoretical as long as the truce holds. The data say the opposite: the bill is already partly paid. US inflation bears its mark. In May 2026, the consumer price index came in at 4.2% year on year, its highest since April 2023, driven by energy up 23.5%, a sequence described in our guide on reading the CPI. Fuel made more expensive by a supply shock propagates mechanically to consumer prices. The signal also appears upstream, in import prices, whose monthly rise surprised, as we analysed in the fine print of import prices. And it reads in the physical flows: Chinese crude imports fell to their lowest since mid-2022, China preferring to draw on its stocks rather than buy at high prices. The Asian bill of the shock and the copper-shortage risk extend the same wave. The crisis is not a future risk to monitor: it is a present cost already spreading among importers, industrialists and households. For central banks, this type of shock is the most uncomfortable there is. A supply shock pushes prices up and activity down at the same time, the very definition of stagflation. Tightening monetary policy to counter imported inflation worsens the brake on growth; loosening it to support activity lets price expectations slip. The Federal Reserve chaired by Kevin Warsh, whose first FOMC we described, inherited this dilemma at the precise moment it intended to normalise its policy. The Strait of Hormuz has, in a few weeks, taken from central bankers the luxury of a simple choice. A reopening that is not really one That leaves the question of the present. The mid-June truce did trigger a recovery, but it is partial and fragile, and maritime players speak of controlled access rather than a reopening. In early July 2026, traffic in the strait still runs at around a third of its normal level. On 27 June, the joint maritime information centre supervised by the US Navy widened a route near Oman to ease passage, but confidence has not returned: Iran for a time reclosed the strait, denouncing Israeli strikes it said contradicted the deal. The result is visible at anchor. About 3,200 vessels, including nearly 800 tankers and cargo ships, still wait west of Hormuz, and the major transhipment hubs like Jebel Ali in Dubai are congested by the diversions. Above all, a good part of the owners have already rebuilt their schedules, contracts and fuel procurement for the rest of 2026 around the Cape of Good Hope route. Undoing these arrangements takes time, and that is why, as academic research sums up, the strait can reopen without global shipping returning to normal for months. The other reading: the shock absorbers For the sake of fairness, one must carry the opposite reading, because not everything gave way. Several shock absorbers worked, and ignoring them would give a falsely apocalyptic picture. Strategic reserves allowed the large importers to hold without immediate rationing, China first, which largely lived off its stocks accumulated at low prices. The ramp-up of US LNG, to a record level, partly replaced the missing Qatari cargoes. Shale oil and the barrels OPEC+ put back on the market added supply when it was missing. And the diplomatic de-escalation came sooner than anticipated, before reserves ran out. The disaster scenario, that of a total closure prolonged over six months, with the barrel durably above $130 and the TTF at €200, therefore did not happen. Some analysts draw an optimistic conclusion from this: the system proved more resilient than feared, and the logistical adaptation, however costly, did its job. This reading is solid, and it deserves to be set against the previous one. But it only fixes the upper bound. To say the shock was absorbed is not to say it was painless: between the disaster averted and the return to normal lie precisely all the costs described above, which are very real. The assessment, on two horizons Two lessons stand out. In the short term, the gap between the two clocks is the real subject: oil prices have largely erased the conflict premium, falling back from their April peak above $106, but the logistical cost remains inscribed in freight, insurance premiums, European gas and fertiliser. Whoever reads only the Brent curve will wrongly conclude the crisis is behind us. The mechanics of the oil market and this deceptive reading are detailed in our guide on reading the oil market. Over the longer term, Hormuz recalls an uncomfortable truth about globalisation: its productivity rests on a handful of chokepoints, none of which has a genuine substitute. Concentrating a fifth of oil, a fifth of gas and a third of agricultural inputs in a channel a few kilometres wide creates a formidable efficiency in normal times and an equally formidable fragility in times of crisis. The ceasefire stopped the bleeding. It did not close the wound, and above all it did not make the system less vulnerable to the next shock. Sources 1. IEA, weight of Hormuz in oil and LNG, scale of the rupture (Fatih Birol, largest disruption in the history of the oil market), bypass pipelines of 3.5 to 5.5 Mb/d: https://www.iea.org/ 2. U.S. Energy Information Administration, Hormuz flows around 20 million barrels a day and rise in international LNG prices during the closure: https://www.eia.gov/todayinenergy/detail.php?id=67604 3. World Economic Forum, "Beyond oil: 9 commodities impacted by the Strait of Hormuz crisis": LNG, fertiliser (up to a third of world trade in raw materials), methanol, aluminium, sulphur, graphite: https://www.weforum.org/stories/2026/04/beyond-oil-lng-commodities-impacted-closure-hormuz-strait/ 4. UNCTAD, "Hormuz disruption deepens global economic strain across trade, prices and finance": implications for trade, prices, freight and developing countries: https://unctad.org/news/hormuz-disruption-deepens-global-economic-strain-across-trade-prices-and-finance 5. UN News, "Despite ceasefire, Hormuz tensions continue to throttle supply chains worldwide": estimate of 9.1 million more people in acute food insecurity in Asia, closure of fertiliser plants in Bangladesh: https://news.un.org/en/story/2026/04/1167365 6. SeaVantage, chronology of the crisis: 28 February strikes, closure in 48 hours, suspension by Maersk, MSC, CMA CGM and Hapag-Lloyd, cancellation of cover on 5 March, Red Sea at 49%, lengthening of 10 to 14 days: https://www.seavantage.com/blog/strait-of-hormuz-crisis-2026-shipping-disruption-timeline 7. Lloyd's List, rise in container freight rates: Shanghai Containerized Freight Index at 2,572 points in the week to 30 May 2026, up 16% in a week and double the late-February level: https://www.lloydslist.com/LL1157327/Hormuz-crisis-side-effect-a-sharp-rise-in-container-shipping-rates 8. Freightos, surcharges applied to shippers: war risk up to $1,500 per TEU, emergency freight hike of $3,000 per FEU, doubling of marine fuel: https://www.freightos.com/freight-industry-updates/market-updates/the-strait-of-hormuz-and-the-container-market-what-you-need-to-know/ 9. Euronews, European gas: Qatari suspension, TTF up 35% in one session above €60 and 76% over the week, scenarios at €155 then €160 to more than €200: https://www.euronews.com/my-europe/2026/03/26/europes-gas-prices-on-the-brink-as-qatari-lng-flows-stall 10. CNBC, gas and LNG surge on Middle East supply fears, Hormuz's share of world LNG: https://www.cnbc.com/2026/03/03/middle-east-war-gas-energy-lng-drone-qatar-strait-hormuz-price-shock.html 11. IEEFA, the Hormuz disruption puts about 10% of Europe's LNG imports at stake: https://ieefa.org/resources/strait-hormuz-disruption-would-jeopardise-10-europes-lng-imports 12. World Bank, "Strait of Hormuz disruption sends natural gas prices surging": https://blogs.worldbank.org/en/opendata/strait-of-hormuz-disruption-sends-natural-gas-prices-surging 13. The Conversation, "The Strait of Hormuz is reopening, but global shipping won't return to normal for months": controlled access, schedules rebuilt around the Cape of Good Hope for 2026: https://theconversation.com/the-strait-of-hormuz-is-reopening-but-global-shipping-wont-return-to-normal-for-months-285313 14. UK House of Commons Library, "Israel/US-Iran conflict 2026: Reopening the Strait of Hormuz": widened route from the joint maritime information centre on 27 June, partial and fragile reopening: https://researchbriefings.files.parliament.uk/documents/CBP-10636/CBP-10636.pdf 15. BLS, May 2026 CPI release: index at 4.2% year on year, energy at 23.5%, cited via our CPI reading guide: https://www.bls.gov/news.release/cpi.nr0.htm ============================================================================ ANALYSIS: Cushing: the Oklahoma terminal that sets the US oil price URL: https://l0g.fr/en/analysis/cushing-the-terminal-that-sets-wti/ Canonical French source: https://l0g.fr/posts/cushing-le-terminal-qui-fixe-le-prix-du-petrole-americain/ Date: 2026-07-14 (reviewed 2026-07-14) Topics: oil, commodities, macro, wti, china ---------------------------------------------------------------------------- The Cushing tanks, in Oklahoma, fell to about 19.7 million barrels as of 26 June 2026, a low since 2014, within a whisker of their operating floor. It is here, in a pipeline crossroads of 8,000 inhabitants, that the WTI price is physically set. Small plumbing, big consequences. One imagines the oil price forming on screens, in London or New York. It also forms, and first of all, in a network of steel tanks laid on the plains of Oklahoma. Cushing is the point where the abstraction of the futures contract touches metal: crude arrives there, is stored there, leaves from there. When these tanks empty fast, it is not a logistical detail, it is a tension signal that propagates up the whole price curve. In late June 2026, they emptied to the point of brushing their lower limit. Why a pipeline crossroads makes the price of the barrel The futures contract on WTI, the US oil benchmark, is not an abstract bet on an index. It is a commitment to physical delivery of crude, and the place of that delivery is written in black and white in the contract specifications: Cushing, Oklahoma. Whoever holds a WTI contract to expiry without unwinding it must take or make delivery of real barrels, in the Cushing tanks. The WTI price is therefore, literally, the price of crude at Cushing. This status is no geographical accident. Cushing sat at the crossroads of the first US pipelines at the start of the twentieth century, and the NYMEX made it the delivery point of its WTI contract in 1983. Today, some thirty pipelines connect to it, with a working storage capacity of about 76 million barrels. Crude from the Permian basin, from the Dakotas or from Canada transits there before descending to the refineries and export terminals of the Gulf of Mexico. This hub role makes the Cushing stocks a barometer: when they rise, the US market is awash with crude; when they melt, physical demand pulls barrels out faster than they arrive. This also explains an episode that stayed in memory. In April 2020, at the collapse of demand, the Cushing tanks were threatening saturation. For lack of anywhere to deliver, the May WTI contract went into negative territory, down to minus $37 a barrel: contract holders paid to get rid of crude they no longer knew where to put. The lesson works both ways. Tank filling is not anecdotal; it can make the WTI price diverge from any world-market logic. To place this crossroads within the whole of the oil machinery, our guide on reading the oil market details the link between stocks, paper and physical. The data: twelve weeks of drawdown Now to the figures published each Wednesday by the US Energy Information Administration (EIA). Its weekly series of Cushing stocks tells, over spring 2026, a nearly uninterrupted decline. The stock peaked around 31.5 million barrels in early April. It fell to 18.96 million on 19 June, its lowest point, before a slight refill to 19.67 million on 26 June. Over twelve weeks, nearly 12 million barrels left the terminal, at a pace close to one million a week. Two reading cautions are needed. First, these figures are weekly and revisable: an isolated point is not a trend, and the rebound of the last week recalls that the system rebalances. Second, a low stock is not in itself a shortage. It must be set against the terminal's capacity and above all against the threshold below which the terminal ceases to function normally. That is where the data takes on meaning. The operating floor, an invisible wall A storage tank never empties completely. At the bottom stagnates an unusable layer of sediment, water, paraffin and residue, what the trade calls the tank bottoms. Above it, enough volume must be kept for the pumps to hold their pressure, for transfers between tanks to remain possible and for outgoing pipelines to keep feeding refineries and terminals. Below a certain threshold, the machinery seizes up. For the whole Cushing complex, this operating floor is estimated at around 20 million barrels, an order of magnitude on which sector analysts converge, including Wood Mackenzie. The issue is therefore not the average filling, but the proximity to the threshold. At 26% of its capacity, Cushing is nowhere near overflowing; on the contrary, it is the opposite that worries. A terminal approaching its floor loses its shock-absorber function: it no longer has a cushion to absorb a mishap, whether an incident on an incoming pipeline or a spike in refinery demand. This scarcity of available volume comes at a price, and it is paid on the futures market. Why the tanks are emptying now Several currents pull crude out of Cushing at the same time. The most powerful is exports. The United States now ships a record volume of crude, on the order of 5.6 million barrels a day according to Kpler, and this demand draws barrels toward the Gulf of Mexico terminals rather than toward the Oklahoma tanks. To this is added a light maintenance season: US refineries ran at nearly 95% of capacity, consuming more crude than usual. Finally, supply disruptions in the north, with the threat of Canadian wildfires to oil-sands production, reduce incoming flows. The geopolitical backdrop matters too. The Iran shock of spring 2026 and the partial closure of the Strait of Hormuz tightened the world crude market and made US barrels all the more sought after for export. We described the bill of that episode in our article on the Hormuz supply chain. The paradox is worth underlining: the same tension that inflates world prices empties the Cushing tanks, because it pushes the United States to export more. The price signal: the curve in backwardation A terminal near its floor translates mechanically into the structure of the futures market. When the immediately available barrel becomes scarce, buyers pay a premium to obtain it right away rather than in a few months. The price curve then goes into backwardation: near-dated maturities trade higher than distant ones. It is the opposite of contango, which signals abundance and rewards whoever stores. Spring 2026 saw this slope steepen. According to Kpler, the spread between the first and third WTI maturities (the M1-M3 spread) rose toward $7 a barrel by mid-June, a marked backwardation that reflects the physical tension at Cushing. Another symptom, the price gap between Gulf of Mexico crude (Magellan East Houston) and Cushing compressed, falling to about $1 against 4 a month earlier: the sign that the market now wants to keep its barrels on site rather than systematically sending them to export. Backwardation is not a speculators' whim, it is the price translation of the geography of the tanks. Several possible trajectories Where does Cushing go from here? Three scenarios take shape, and they must be taken for what they are, analyst hypotheses, not forecasts. The first scenario is that of the rebuild. OPEC+ raised its production by nearly 600,000 barrels a day between April and June, then by a further 188,000 in July; if this additional crude materialises while export demand runs out of steam and refineries enter maintenance, Cushing can refill and the curve ease. The slight rebound of the last week of June points that way, without confirming it. The second scenario is that of prolonged tension. As long as exports run at full tilt and refineries pull hard, the tanks stay scraping the floor. The market then lives with durable backwardation and an acute vulnerability: the slightest pipeline incident or the shutdown of a supply source could force a disorderly unwind of positions at expiry, a physical squeeze where holders of short contracts struggle to find crude to deliver. The third scenario is the most interesting, because it comes from outside. We documented how China, the world's largest importer, played the role of a price ceiling in 2026 by cutting its purchases and living off its record reserves, in our article on the Chinese inventory that caps prices. What happens if Beijing starts buying again? Its imports had fallen to 9.25 million barrels a day in April, a near-three-year low. A simple return toward normal would add one to two million barrels a day of demand to the world market. Brent would rise, the Brent-WTI spread would widen, and this more favourable spread would encourage even more US crude to be exported. The mechanics are counter-intuitive: a reviving Asian demand does not fill Cushing, it empties it a little more, by pulling US barrels toward the export terminals. A Chinese awakening would be bullish for the world price and tightening for the Oklahoma terminal. The opposite reading It remains not to overstate the signal. The serious objection lies in the very evolution of the US market. Since the United States became a major exporter, the benchmark crude price is increasingly set on the Gulf Coast, around Houston, where the export terminals concentrate, and less and less at Cushing. The marginal US barrel is a barrel that gets exported, priced in relation to world Brent, not a barrel sleeping in Oklahoma. In this reading, a floor at Cushing becomes a partly local phenomenon, a logistical bottleneck that inflates backwardation on the near-dated WTI maturities without necessarily saying much about the world oil balance. This nuance has weight, and it invites not confusing a plumbing tension with a planetary shortage. It has its limits, however. Cushing remains the delivery point of the most traded contract in the world, and a delivery stress can trigger violent price moves that spill far beyond Oklahoma, as the 2020 episode showed. The terminal weighs less than before on the price level; it weighs as much as ever on its volatility at expiry. Following it closely, without making it say more than it says, remains the right discipline. Sources 1. U.S. Energy Information Administration, weekly Cushing crude stocks excluding SPR, series WEPC0SAXYCUOKMBBL (31.5 Mb on 3 April, 18.96 Mb on 19 June, 19.67 Mb on 26 June 2026): https://www.eia.gov/dnav/pet/hist/LeafHandler.ashx?n=PET&s=WEPC0SAXYCUOKMBBL&f=W 2. Kpler, "WTI flirts with physical squeeze as Cushing buffers evaporate" (10 June 2026), M1-M3 backwardation toward $7, US exports ~5.6 Mb/d, refineries at 95%, Houston-Cushing differential ~$1 against 4 a month earlier: https://www.kpler.com/blog/wti-flirts-with-physical-squeeze-as-cushing-buffers-evaporate 3. Energy News Beat / TankTerminals, "Cushing Oil Storage Hits Tank Bottom", capacity ~76 Mb, level ~26%, definition of tank bottoms: https://tankterminals.com/news/cushing-oklahoma-oil-storage-hits-tank-bottom-implications-for-energy-markets-consumers-and-investors/ 4. Transport Topics, "Oklahoma Crude Inventories in Cushing Fall Near Minimum", lowest since October 2014, operating floor of about 20 Mb estimated by Wood Mackenzie: https://www.ttnews.com/articles/oklahoma-crude-cushing-low 5. l0g, "Oil: the Chinese inventory that caps prices", Chinese imports at 9.25 Mb/d in April 2026, OPEC+ increase of 600,000 then 188,000 b/d: https://l0g.fr/en/analysis/oil-the-chinese-inventory-capping-prices/ 6. l0g, guide "Reading the oil market" (paper vs physical, contango and backwardation, data calendar): https://l0g.fr/en/guides/read-oil-market/ ============================================================================ ANALYSIS: The Hormuz crisis: Asia's bill for an energy shock, at the hour of the truce URL: https://l0g.fr/en/analysis/hormuz-crisis-asia-economic-toll/ Canonical French source: https://l0g.fr/posts/crise-ormuz-asie-impacts-economiques/ Date: 2026-07-14 (reviewed 2026-07-14) Topics: geopolitics, energy, asia, macro, oil, inflation ---------------------------------------------------------------------------- The Strait of Hormuz, barely 33 kilometres wide at its narrowest point, concentrates more than 80% of the crude and liquefied natural gas bound for Asia. Blocked from early March to mid-June 2026, it turned a regional war into a continental shock. The 17 June memorandum of understanding begins a reopening, immediately weakened by a new Iranian closure announcement on 20 June. This article draws up the cumulative assessment, sector by sector, on verified institutional sources, without extrapolating beyond the available figures. The sequence is now documented. The war launched on 28 February 2026 led Iran to close the strait in early March, followed by a US blockade of Iranian ports from April to late May. The conflict was suspended by a memorandum of understanding signed on 17 June at the Palace of Versailles, which implies a reopening of the strait and the lifting of the blockade. Commercial traffic recovered on 18 and 19 June, before Tehran announced a new closure on 20 June, disputed by Washington. The truce is therefore real but precarious, and the bill analysed below is largely cumulative: it does not vanish with the signing. An already colossal macroeconomic bill According to a Reuters tally, companies listed in the United States, Europe and Asia recorded at least $25bn in costs directly linked to the conflict, an amount presented as a starting point. At least 279 international companies took defensive measures: price hikes, production cuts, suspension of dividends or buybacks, short-time working, fuel surcharges or requests for emergency aid. The institutional revisions converge. The Asian Development Bank (ADB), in a special update of 29 April 2026, lowered its growth forecast for Asia-Pacific to 4.7% in 2026 and 4.8% in 2027, against 5.1% in both cases previously, and raised regional 2026 inflation from 3.6% to 5.2%, on an assumption of oil around $96 a barrel, against $69 before the conflict. The IMF, in its April 2026 outlook, cut world growth to 3.1% in 2026, against 3.4% in 2025, and raised world inflation to 4.4%, on the assumption of a 19% rise in energy prices; the euro area drops to 1.1%. The World Bank, finally, projects a 24% rise in energy prices in 2026, the highest level since the invasion of Ukraine, and a 16% rise across all commodities. The United Nations Development Programme (UNDP) estimates that the escalation could cost the region between $97bn and $299bn in lost output and push 8.8 million more people into poverty. Asian industry under the naphtha shock Beyond crude, it is the Asian industrial apparatus that was disrupted. The most emblematic rupture concerns naphtha, an oil derivative present in a dizzying range of products, from plastic films to industrial inks and medical devices. Japan and South Korea depend heavily on naphtha imported from Qatar and Kuwait, whose exports were hindered by the blockage of the Strait of Hormuz. The consequences were visible. In Japan, consumer-goods companies, for lack of stable supply, dropped the colours of their food packaging to save on ink. According to Oxford Economics, naphtha was one of the main channels through which Middle East supply shocks transmitted to the whole economy. Petrochemical plants across Asia cut their operating rates, threatening the chains of manufacturing, textiles, construction and packaging. The World Bank calls the episode the largest oil supply shock ever recorded, with an initial reduction on the order of 10 million barrels a day. The spectre of inflation and the powerlessness of central banks The surge in energy prices placed Asian central banks before a formidable dilemma. Brent exceeded $100 a barrel at the peak of the crisis, up some 65% in the single month of March according to the World Bank, before easing; the institution pencils in an average of $86 in 2026, against $69 in 2025. Nearly 80% of the blocked crude and LNG was bound for Asia, making the region exceptionally vulnerable. Central banks had to navigate between the sudden inflationary shock and structural headwinds: slowing growth, pressure on currencies. According to S&P Global Ratings, the room to ease monetary policy narrowed, which forced Asia-Pacific central banks into caution. Several importing economies saw their gasoline prices rise sharply, while Thailand and Indonesia contained the increases through subsidies and price controls, at a growing budgetary cost. Agriculture and food security on the front line The shock goes beyond energy. The World Bank projects a 31% rise in fertiliser prices in 2026, driven by a 60% jump in the price of urea, because about half of world urea and nearly a third of ammonia transit the strait. The surge threatens food security and the livelihoods of hundreds of millions of smallholder farmers in South Asia. A prolonged disruption would weigh not only on immediate food prices, but also on subsequent harvests, as food crops are traded off in favour of more profitable production. The World Bank sums up the dynamic in cumulative waves: energy prices first, food prices next, inflation and interest rates finally, which weigh on the debt service of the most fragile countries. India: growth maintained, fragilities exposed India illustrates the paradoxical situation of many Asian economies. On one side, it remains the fastest-growing large economy, around 6.3% to 6.5% in 2026 according to the institutions. On the other, its dependence on energy imports through the strait is considerable: about 55% of its crude imports and 90% of its LPG imports transit Hormuz. According to Moody's Analytics, Asia-Pacific economies entered 2026 on fragile foundations, weak domestic demand and slowing exports. The conflict adds "a new difficulty in the wheel of growth of the large economies like China, India, Japan and South Korea", with a "disturbing echo" of the inflation and supply shocks that followed the pandemic and the invasion of Ukraine. The substitution routes: a costly illusion Alternative routes exist, but remain limited and costly. Saudi Arabia increased the capacity of its East-West pipeline (Petroline) toward the Red Sea; the United Arab Emirates extended its pipeline to Fujairah. The EIA estimates that these routes could collectively carry on the order of 3.5 million barrels a day, about 20% of the strait's normal traffic. For industry, the detours via the Cape of Good Hope lengthen journeys by several weeks and push up freight costs. The price of a container between Asia and Europe jumped 20% in a few days, and Maersk warned that the extra costs would feed through to entire supply chains. The governments' response: subsidies and rationing Faced with the shock, Asian governments mixed price mitigation and incentives for restraint. In India, cuts to the fuel excise tax and a squeeze on the margins of state oil companies limited the transmission to retail prices, while LPG rationing prioritised households. In Japan, abundant strategic reserves, on the order of 228 days at the start of April 2026 including public and commercial reserves, allowed, together with price caps, the increases at the pump to be limited. South Korea also drew on its reserves. These measures carry a budgetary cost and create distortions, which pushes the ADB to recommend targeted, temporary support rather than general subsidies. A structural recomposition under way The crisis could accelerate transformations already begun. For manufacturing-intensive economies, the conflict could hasten a strategic recalibration long under way. Some governments, including South Korea, have begun to explore alternative energy sources in response to the disruptions coming from the Gulf. Asia's dependence on the strait is a brutal reminder that the energy transition, however fast, does not happen in a few weeks: one lives for a long time in two systems at once, building the next while paying the price of the previous one. Asia at the hour of the truce The shockwave will remain tangible even after the signing: fuel shortages, petrochemical chains under strain, heavier food baskets. The World Bank and the IMF say it plainly, most of the damage is already absorbed and will take months to dissipate, assuming the truce holds. And Iranian leverage over the strait remains intact, as the 20 June closure announcement recalled. For Asia, more than 80% dependent on this corridor for its energy, the lesson is less an immediate threat than a structural vulnerability now quantified, and impossible to ignore. --- Primary sources: Asian Development Bank, special forecast update (29 April 2026) and Asian Development Outlook (April 2026); IMF, World Economic Outlook: Global Economy in the Shadow of War (April 2026); World Bank, Commodity Markets Outlook (28 April 2026); UNDP, crisis cost estimates; Reuters, tally of costs for listed companies; S&P Global Ratings, Economic Outlook Asia-Pacific Q2 2026; Moody's Analytics, Asia-Pacific Outlook; Oxford Economics; U.S. Energy Information Administration, World Oil Transit Chokepoints; House of Commons Library, Israel/US-Iran conflict 2026: Reopening the Strait of Hormuz (June 2026); PBS NewsHour and RFE/RL for the chronology of the 17 June memorandum and the reopening (June 2026). The growth, inflation and price figures come from these sources and have not been recalculated; the lost-output and poverty estimates are ranges, sensitive to the duration of the blockade and the holding of the truce. ============================================================================ ANALYSIS: Hormuz reopens: three quantified scenarios for the normalisation of the oil market URL: https://l0g.fr/en/analysis/hormuz-reopens-three-oil-scenarios/ Canonical French source: https://l0g.fr/posts/accord-paix-ormuz-rouvre-scenarios/ Date: 2026-07-14 (reviewed 2026-07-14) Topics: geopolitics, energy, macro, oil ---------------------------------------------------------------------------- On 14 June, Donald Trump announced on Truth Social that the deal with Iran was "complete" and authorised the "toll-free" reopening of the Strait of Hormuz, lifting the US naval blockade of Iranian ports. Tehran confirmed the next day, specifying that implementation would only begin at the formal signing, set for 19 June in Geneva. Markets reacted immediately: Brent lost 4.7% on 15 June to close at $83.17, its lowest level since 10 March, and WTI fell 4.8% to $80.75. Except that, as I already noted in my April situation report, an announcement of reopening is not a reopening. And above all, the deal took the war premium out of prices, not the damage the war left behind. Understanding that gap is the whole point for anyone trying to anticipate what comes next. What Hormuz was worth, and what the crisis destroyed The strait is the world's most critical oil artery after Malacca. In 2025, the IEA measures an average flow of 20 million barrels a day of crude and refined products through it, about 25% of world seaborne oil trade. The EIA puts the first quarter of 2025 at 20.1 million barrels a day, of which 14.2 crude, and estimates that this represents nearly a fifth of world oil consumption. On gas, the IEA recalls that about 19% of world LNG trade depends on Hormuz. The war launched in late February almost shut off the tap. According to Britannica, after the conflict began and then the Iranian threats to attack vessels, more than 95% of traffic was rerouted. Analyses cited by the press speak of a roughly 95% fall in tanker transits and nearly 99% of LNG over the 107 days of blockage. It is, by the IEA's own account, the largest disruption the oil market has ever suffered. The problem is that there is no plan B on the scale needed. The London School of Economics, citing the IEA, recalls that the Saudi and Emirati pipelines that bypass Hormuz can only redirect 3.5 to 5.5 million barrels a day. On a base of 20 million, that leaves a net deficit of 14.5 to 16.5 million barrels a day in the event of a total closure. It is this figure that says how long strategic reserves can hold, and the answer is: not indefinitely. The gap between $83 and $72 Here is the point most commentary misses. Brent at $83 is not Brent at $72. Before the 28 February strikes, the barrel was trading around $70-72. During the crisis, it peaked above $106 in April, on the successive hopes and relapses around the strait. The 14 June deal erased the acute conflict premium, but the market still prices a residual premium on the order of $8 to $15 tied to the logistical aftermath. // Brent 2026: the war premium, then the aftermath (USD/barrel) 110 95 80 65 28 Feb Mar Apr May 15 Jun ~71 ~106 (Apr peak) 83 pre-war floor ~72 This gap is not irrational: it has physical causes. Traffic has not moved since the announcement, shipowners awaiting the 19 June signing and security guarantees, according to AIS data reported by Argus. The Pentagon warns that mine clearance can take up to six months, even if the memorandum sets a 30-day target. Hundreds of vessels are stuck in the Gulf and will have to exit, be inspected and repositioned. And some producers having shut wells for lack of storage, their restart is slow. To follow the barrel live rather than at the moment I write, our guide on reading the oil market sets out Brent and WTI. Three normalisation scenarios From there, one can build three trajectories. They are not predictions, but plausible bounds anchored on the estimates published by market analysts and institutions. Central scenario: orderly normalisation. This is the trajectory that fits Kpler's estimates, for which traffic could recover to nearly 50% of pre-war levels within 30 days of the deal, assuming no major incident. Kpler estimates at 118 the number of stranded tankers that could exit within 15 days. Frontline, which has five vessels stuck in the Gulf, judges that "vessels will move very quickly once the deal is signed". In that case, Brent gradually converges toward the $78-85 zone by the end of summer. This is consistent with Goldman Sachs, which raised its Brent forecast to $85, and with Fitch, which pencils in a 2026 average of $87. High scenario: slow reopening. If mine clearance drags toward the high end of the six months mentioned by the Pentagon, if shipowners demand persistent risk premiums and if the marine insurance premium stays high, logistical congestion keeps a high price floor. This is the scenario the EIA implicitly assumes, whose June Short-Term Energy Outlook projects a 2026 average of $95, on slower-reopening assumptions. In this configuration, the barrel stays stuck above $90 for a good part of the second half, with the known consequences for imported inflation, a subject I developed in the great US inflation comeback. The coming CPI releases and the FOMC meetings that will arbitrate this energy shock are worth watching on the economic calendar. Low scenario: return to the floor. If the 19th signing holds, if mine clearance goes fast and if OPEC+ reopens the valves of shut wells, the residual premium dissipates and Brent returns toward $72-75, its level before the strikes. This is the least likely scenario in the short term, because it assumes all the frictions resolve simultaneously, but it is the long-term anchor once the logistics are purged. // Hormuz traffic recovery (% of pre-war levels) 100% 66% 33% 0% D+0 D+30 D+90 D+180 central slow fast ~50% (Kpler) The variables that can break everything Three threads can unravel the ball. First the mines: as long as they are there, captains bide their time, and Bimco maintains a high-risk advisory on the strait. Next the ambiguity of the deal itself: Iran speaks of a toll-free transit limited to 60 days, after which Tehran and Oman would administer the strait, while Vice President Vance asserts that the US expectation is durable free passage. This divergence of interpretation is exactly the kind of vagueness that derailed the April ceasefire, as I analysed in the Hormuz tolls and the USDT-Tron rail. Finally the Lebanese variable: Israeli operations in Lebanon continue independently of the US-Iran framework, and it was a strike in Lebanon that suspended access to the strait in April. Nothing in the current memorandum resolves this point. What stands out The 14 June deal is good macro news, but the market is right not to cry victory. The war premium is out, the congestion is not. My working scenario remains the central one: convergence toward $78-85 by the end of summer if the signing holds and if mine clearance respects the 30-day window, with an asymmetric risk to the upside as long as the mines and the toll ambiguity persist. For the full geopolitical context of this crisis, my deep piece on the 2026 Iran war keeps all its relevance. The strait reopens. The market, for its part, still takes weeks to clear a four-month jam. --- Sources: EIA (World Oil Transit Chokepoints), IEA (Strait of Hormuz factsheet 2025), London School of Economics Business Review, Kpler, Goldman Sachs, Fitch, EIA Short-Term Energy Outlook of June, Argus Media, CNBC, Reuters, NBC News, Britannica. Price levels as of 15 June 2026. This is not investment advice. ============================================================================ ANALYSIS: The US-Iran MoU: a ceasefire that destabilises more than it soothes URL: https://l0g.fr/en/analysis/us-iran-memorandum-june-2026/ Canonical French source: https://l0g.fr/posts/mou-usa-iran-juin-2026/ Date: 2026-07-14 (reviewed 2026-07-14) Topics: iran, united states, geopolitics, oil, trump, us politics ---------------------------------------------------------------------------- On the announcement of the deal on 15 June, markets applauded: the S&P 500 closed up 1.6%, near its record, and oil fell nearly 5%, to its lowest since early March. That is perhaps the only simple thing in this memorandum. On 17 June, Donald Trump signed with Iran a memorandum of understanding, a 14-point MoU, over a dinner with Emmanuel Macron at the Palace of Versailles, the day after the G7. The agreement took effect at once: the Strait of Hormuz reopens and the US naval blockade is lifted. It extends the ceasefire of the war launched on 28 February, when the United States and Israel had struck Iran and killed its supreme leader. The strait becomes free again, but for 60 days only, the blockade disappears within 30 days, Iran can export its oil from the signing, and sanctions relief remains conditional on Iranian compliance. Everything else, starting with the nuclear file, is deferred to a final agreement to be negotiated in 60 days. This is the direct sequel to the 2026 Iran war and the reopening of Hormuz. What the MoU settles, and what it dodges The market relief rests on one fact: the route that carries about 20% of world energy reopens. But most of the substantive issues are pushed back. The highly enriched uranium remains buried under the bombed sites, with no public Iranian commitment to surrender it. Trump told the New York Times that Iran would be allowed to enrich at a low level, reversing the total-dismantlement demand that had justified the war. The text also provides for a reconstruction plan of at least $300bn, backed by the United States and regional Gulf partners. For the specialists, this is not peace. The Stimson Center and the Atlantic Council underline that the MoU resolves neither the nuclear file, nor the missiles, nor Iranian support for regional militias, that bilateral relations are not restored and that nothing substantial is secured. The most awkward reminder is that Trump claims to have ended a war he himself started. The imbalance of concessions The detail circulating, relayed notably by Washington Post journalist Aaron Blake and cross-checked by Axios and Fortune, shows a very asymmetric exchange. On the US side: ceasefire, lifting of the naval blockade, withdrawal of forces within 30 days, reconstruction funding, a commitment to lift "all types of sanctions", waivers for oil exports, and asset unfreezing. On the Iranian side, two lines: return Hormuz traffic to its pre-war level, under Iranian arrangements, and "reiterate" that it will not seek the weapon. Concrete and irreversible on one side; on the other, the restoration of a strait that Tehran itself had closed, and the repetition of a promise it was already making. On the money, caution with the big numbers. The reconstruction fund is announced at around $300bn, but Vice President Vance asserts it would be funded by a "Gulf coalition", not by Washington, and conditioned on compliance with nuclear commitments. The asset unfreezing, for its part, is counted in tens of billions, not hundreds: the draft reported by the Mehr agency mentions $24bn over the 60 days of negotiation, a figure Washington calls "completely false", drawn from a total stock estimated at around a hundred billion of which only a fraction is liquid, on the order of $30 to $50bn per the JCPOA precedent. The defensible order of magnitude is therefore a potential financial benefit north of $300bn, spread out and conditional, not an immediate transfer of 600. Why it destabilises The strategic problem is simple to state: the deal validates the leverage Iran seized by force. Reopening a strait that was free before the war, lifting its own blockade and unfreezing assets, is to reward the energy hostage-taking. The strait's free passage lasts only 60 days anyway: the chief Iranian negotiator has already warned that Tehran will then collect fees on traffic, and called the MoU a failure for Washington. Tehran, whose leadership has become even more opaque after the death of its chiefs, believes it has won the standoff, and intelligence assessments cited by CNN indicate a resumption of drone production and military reconstruction, while the stock of more than 440 kg of enriched uranium remains beyond any control, deferred to the final agreement. Second shockwave: Israel was kept out of the negotiations, conducted via Qatar. Netanyahu was caught off guard, and Trump criticised him publicly after an Israeli strike on Beirut, while Israel-Hezbollah fighting continues in Lebanon, whose inclusion in the ceasefire is itself contested. A major ally sidelined and displeased is the risk of a unilateral action that would blow up the 60 days. Finally, tolerating Iranian enrichment creates a proliferation precedent that worries the Gulf capitals. The political crisis brewing in Washington This is perhaps where the deal is most explosive. It fractures Trump's own camp. Republican hawks, Ted Cruz, Lindsey Graham, Roger Wicker, Thom Tillis, and former adviser John Bolton, judge the agreement weaker than the 2015 JCPOA, the very one Trump had denounced. A White House confidant speaks of a "low-rent humiliation". Facing them, the wing hostile to endless wars and candidates worried about gasoline prices ahead of the midterms mostly want out. Three fault lines make the situation politically unstable. The narrative first: Trump launched the war to dismantle the Iranian nuclear programme, and he exits by authorising it at a low level, which frontally contradicts Pete Hegseth and the casus belli. The cost next: an unpopular war, at least $29bn and 13 US service members killed, to reopen a route that was already open. The method finally: Congress discovered the text as it was already circulating abroad, and the Senate advanced a war-powers measure. With a falling approval rating and midterms approaching, Trump is in a classic bind: a majority of Americans reject the war, but the price of exiting it, money and enrichment for Tehran, is precisely what his own hawkish base will not forgive. The criticism has moreover spilled beyond his camp: Senate Democratic leader Chuck Schumer judges that Iran won on nearly every one of the 14 points and that the country is worse off than before the war. Trump, for his part, added that he found it acceptable for Iran to keep some ballistic missiles, and warned that absent a final agreement within 60 days, they would go back to bombing. The strategic balance sheet This memorandum stops the fire, and that is real. But it settles none of the causes of the conflict, it enshrines Iran's strategic gain, it cracks the axis with Israel, and it opens trench warfare in a US political class critical on both flanks, six months from the midterms. The destabilisation is therefore not a paradox: a ceasefire that distributes rewards without closing anything leaves all the tensions active, simply displaced from the battlefield to the negotiating table and Congress. Now signed and in force, the agreement opens a 60-day countdown: the delegations' meeting on Friday in Switzerland is to launch the nuclear talks, but a unilateral Israeli strike, the reinstatement of Hormuz fees or a deadlock on uranium would be enough to reignite everything. Oil fell 5%. The risk, for its part, only changed address. Sources - CBS News, Read the 14 points of the agreement between Iran and the U.S. (official text read to the press, 60-day free passage, reconstruction at least $300bn, withdrawal of forces after the final agreement), 17 June 2026, - Times of Israel, Trump signs deal with Iran (signing at Versailles on 17 June; Iranian negotiator speaking of "failure" and of Hormuz fees after 60 days), 18 June 2026, - Axios, U.S., Iran sign deal ahead of Friday meeting (signing at Versailles, agreement in force, Friday meeting in Switzerland), 17 June 2026, - ABC News, Trump signs memorandum while dining at Versailles (Schumer: Iran won on nearly every point; ballistic missiles; threat to resume bombing), 17 June 2026, - Council on Foreign Relations, Trump's Iran Deal Reopens the Strait. Much Remains to Be Done. (Lebanon as a breaking point, Iranian leverage over Hormuz), 17 June 2026, - NBC News, Oil prices fall on Iran deal (15 June close: S&P 500 +1.6%, US crude -4.8% at $80.75, Brent -4.7% at $83.17, lowest since early March), 16 June 2026, - NPR, U.S. and Iran announce an initial deal to end the war and reopen the Strait of Hormuz (deal detail, low-level enrichment), 15 June 2026, - The Hill, Trump's Iran peace deal pits Republican vs. Republican (Cruz, Graham, Bolton, Pletka; $300bn fund), - CNN, Why a possible Iran deal may be almost as divisive as Trump's decision to wage war (Iran rebuilding its drones, validation of Iranian leverage), 25 May 2026, - Al Jazeera, Will a US-Iran deal unlock $300bn in investment fund for Tehran? (Vance on Gulf funding and its conditional nature; $24bn disputed; 440 kg uranium stock), 16 June 2026, - Foreign Policy, A Trump Deal With Iran Could Spell Trouble for Israel's Netanyahu (Israel sidelined, Hormuz leverage handed to Iran), 1 June 2026, ============================================================================ ANALYSIS: Lebanon: the forgotten big loser, and Trump's very dangerous Syrian gamble URL: https://l0g.fr/en/analysis/lebanon-forgotten-loser-syrian-gamble/ Canonical French source: https://l0g.fr/posts/liban-perdant-oublie-pari-syrien-trump/ Date: 2026-07-14 (reviewed 2026-07-14) Topics: lebanon, israel, hezbollah, syria, geopolitics, trump, gas ---------------------------------------------------------------------------- While Iran negotiates $300bn of reconstruction, Lebanon counts its dead and its razed villages. And Washington is preparing a new ordeal for it. Israel today occupies nearly 2,000 square kilometres of southern Lebanon, close to a fifth of the country, its deepest incursion in twenty-five years. More than a million Lebanese have fled, and Israeli strikes have killed nearly 3,800 people since 2 March, according to the Lebanese Health Ministry. And while the memorandum signed on 17 June between Washington and Tehran promises Iran a reconstruction plan of at least $300bn, no one, at this stage, is offering anything comparable to Beirut. This is the direct sequel to the 2026 Iran war and the US-Iran memorandum. A country built on a balance that everything destabilises To grasp what is at stake, you have to keep the history in mind. Lebanon is a mosaic of some fifteen communities, including one of the oldest Christian presences in the region, organised since the 1943 National Pact by a confessional power-sharing arrangement. This balance has already shattered once, in the civil war of 1975 to 1990, closed by the Taif Agreement. It also carries the memory of a Syrian military tutelage, from 1976 until the 2005 withdrawal wrenched after the assassination of Rafik Hariri. Hezbollah, for its part, was born of the 1982 Israeli invasion; the 2006 war closed on UN Security Council resolution 1701, meant to push its fighters north of the Litani river. It is this edifice, already weakened by the 2019 economic collapse, that the current war shakes again. The invasion of the South On 28 February 2026, the United States and Israel strike Iran and kill its supreme leader. Hezbollah resumes fire; Israel responds with a large-scale offensive. From 16 March, five divisions enter southern Lebanon, the Litani bridges are destroyed to cut the region from the rest of the country, and the town of Bint Jbeil becomes a battlefield. In late May, the Israeli army crosses the Litani for the first time since 2006, seizes Beaufort Castle and encircles Nabatieh, the political and economic heart of the Shia community. Resolution 1701 is trampled, and the MoU ceasefire imposes no Israeli withdrawal from Lebanon. The gas prize, one more mirage The Lebanese seabed long nourished hopes of a way out of the crisis. The 2022 maritime agreement, negotiated by Washington, split the disputed zone: Israel kept the Karish field, which it has exploited since, and Lebanon recovered most of the Qana deposit. But the first well drilled on the Lebanese side, Qana 31/1, proved dry in 2023. Official estimates of about 25 trillion cubic feet of reserves remain speculative, and the sector advances only in slow motion, despite the award in January 2026 of a new block to TotalEnergies, Eni and QatarEnergy. While the Jewish state monetises its share, Lebanon still awaits its own, and the war, by freezing all offshore investment and ravaging the agriculture of the South, pushes the horizon back further. Trump's Syrian gamble, a Pandora's box It is on this already-mined ground that the US president launched his idea. On 7 June, then on the sidelines of the G7, Donald Trump suggested that Syria "deal with" Hezbollah in Israel's place, whose campaign he deems too deadly, while praising Syrian president Ahmed al-Sharaa, a former jihadist who passed through al-Qaeda. The idea, which goes back to a March plan to send Syrian troops to disarm Hezbollah, is almost unanimously deemed explosive. Why? Because it would reawaken everything Lebanon is trying to bury. Bringing a Sunni-dominated Syrian army, drawn from Islamist factions some of which have been accused of abuses against minorities, into a country with a large Shia and Christian population, is to relight the confessional fuse. Michael Young, a Lebanon specialist at the Carnegie Middle East Center, calls the idea "completely absurd" and a "Pandora's box" that "would divide Lebanon". The Washington Institute warns that such an intervention would provoke "a sectarian conflict between Syrians and Lebanese", offer Hezbollah a new rallying cry and export the instability of a Syrian power that does not yet fully control its own army. Add to this the memory of the Syrian occupation: even Hezbollah's opponents still prefer the party of God to a return of Damascus's soldiers. Unsurprisingly, Syria refused, Lebanese president Joseph Aoun opposes it, and Saudi Arabia, Qatar, Egypt and Turkey have all discouraged the idea. Still, it was voiced by the world's leading power, and it sketches an outcome in which civil war would be subcontracted. The big loser of a war it did not want Lebanon asked for nothing. Its state denounced the strikes as an attack on its sovereignty, its government argued for the disarmament of Hezbollah, and it was the party's retaliation for the assassination of the Iranian leader that dragged the country into a war whose price it pays dearly. Yet in the great regional bargain, it is Iran that is promised a $300bn reconstruction, while Lebanon faces needs already put by the World Bank at $11bn for the 2023-2024 war alone, of which barely $250m is funded to date. The Lebanese picture is damning: a GDP cut by about 40 percent since 2019, a currency that has lost 98 percent of its value, more than a third of the population below the poverty line, and a 2026 war that adds its own billions of damage. To this already untenable equation, the idea of a Syrian intervention would add the most dangerous risk of all, that of an internal collapse. Lebanon is not the secondary theatre of the regional crisis. It has become its most silent victim, and perhaps the next fault line. Sources - World Bank, Lebanon Recovery and Reconstruction Needs (RDNA, needs at $11bn, cost $14bn, GDP -40% since 2019), March 2025, - World Bank, Lebanon Emergency Assistance Project FAQ ($1bn LEAP programme, $250m funded, $750m gap), February 2026, - SANA, Lebanese Health Ministry toll since 2 March (3,666 killed as of 10 June, rising to 3,798 killed and 11,798 wounded by mid-June), June 2026, - Al Jazeera, Israeli forces push past Lebanon's Litani River (≈ 2,000 km² occupied, Beaufort, Nabatieh; analysis by Imad Salamey, Lebanese American University), 31 May 2026, - Wikipedia, 2026 Lebanon war (invasion chronology, five divisions, Litani bridges, US pressure on Syria), accessed June 2026, - Time, How an Israeli Ground Invasion of Lebanon Could Unfold (course of the ground invasion, mass displacement, Shia population of the South), 16 March 2026, - CNN, Why Trump's proposal for Syria to fight Hezbollah will send shudders across Lebanon (Michael Young, Carnegie; sectarian dimension; Israel does not withdraw), 16 June 2026, - The Washington Institute, The U.S. Should Not Encourage Syria to Enter Lebanon (risk of sectarian conflict, fragility of the Syrian army), June 2026, - i24NEWS, Syria's al-Sharaa rejects military action against Hezbollah despite Trump's push (Damascus's refusal, conditions tied to Israeli withdrawal), 17 June 2026, - Ynet, Could Syria move against Hezbollah? Trump remarks spark regional uncertainty (Trump's G7 quotes, Aoun and Lebanese diplomacy's position), 16 June 2026, - The National, Lebanon signs offshore gas deal with TotalEnergies, Eni and QatarEnergy (Block 8, slow restart after the 2022 maritime agreement), 10 January 2026, - NOW Lebanon, The full story of Block 9 (Qana 31/1 well dry in 2023; speculative estimate of 25 Tcf), November 2023, ============================================================================ ANALYSIS: Aid cuts and a fertiliser shock: the food risk converging on the poorest countries URL: https://l0g.fr/en/analysis/aid-cuts-fertiliser-shock-food-risk/ Canonical French source: https://l0g.fr/posts/trump-famine-arme-destruction-massive/ Date: 2026-07-14 (reviewed 2026-07-14) Topics: geopolitics, macro, energy ---------------------------------------------------------------------------- Two events strike at the same time the countries most dependent on aid and food imports: the dismantling of the US aid agency in 2025, and the disruption of the Strait of Hormuz in 2026. Neither, taken in isolation, is new. Their superposition, on the same countries, is. This article sticks to documented figures and sets aside the death-toll estimates that no one can establish today. US aid, dismantled For two decades, the United States was the world's largest bilateral donor. In July 2025, USAID was dissolved and, according to the figure quoted by secretary of state Marco Rubio himself, about 83% of its programmes were cut. On the consequences, several teams have published convergent projections. The Lancet (Cavalcanti et al., 2025) estimates that USAID-funded programmes were associated with about 91 million deaths averted between 2001 and 2021, and projects, if the cuts continue, more than 14 million additional deaths by 2030 (uncertainty interval: 8.5 to 19.7 million), including 4.5 million children under five. A separate study published in Lancet Global Health (February 2026), which incorporates the cuts of other donors (United Kingdom, Germany, Canada), widens the range to 9.4 million deaths in a moderate scenario and 22.6 million in a halving scenario. The Center for Global Development, for its part, estimates between 500,000 and 1 million lives lost in 2025 alone as a result of the US cuts. These numbers are modelled projections, with uncertainties their authors underline. Their value is not the decimal, but their convergence: different methods and data reach the same order of magnitude. Hunger, already at a record level The context on which these cuts apply is already tight. According to the Global Report on Food Crises 2026, published on 24 April 2026, 266 million people in 47 countries were in acute food insecurity (IPC phase 3 or above) in 2025. The WFP Global Outlook counts a total of about 318 million people in acute hunger, of whom 41 million in phase 4 (emergency) and about 1.4 million in phase 5 (catastrophe). For the first time since monitoring began, two famines were confirmed in the same year, in the Gaza Strip and in areas of Sudan; the famine risk is maintained for Gaza, Sudan and South Sudan in 2026. On nutrition, 35.5 million children were in acute malnutrition, nearly 10 million of them in severe form. The fertiliser channel This is where the second shock comes into play. The Gulf is a major exporter of nitrogen fertilisers (urea, ammonia), whose seaborne transport passes, like part of the hydrocarbons, through the Strait of Hormuz. A disruption there raises the cost of both energy and agricultural inputs. The link between fertiliser prices and food insecurity is documented: the 2022 shock, following the invasion of Ukraine, fed the surge in acute hunger in the following years. The SOFI 2025 report (FAO and UN agencies) recalls that, since 2020, food inflation has outpaced general inflation and hits low-income countries hardest, where it peaked around 30% in May 2023. A new shock to fertilisers and energy therefore transmits first to the poorest importers, those whose yields depend on inputs bought in foreign currency. Why the superposition matters The point is not any single event. It is that the aid cut removes a shock absorber at the precise moment a supply shock raises prices, and that the two hit largely the same countries: Sudan, South Sudan, Yemen, the Democratic Republic of Congo and Somalia are among the largest crises recorded by the GRFC, and were also among the recipients of US aid. No precise quantified toll of this superposition can be established today, and any estimate of a number of "additional" deaths attributable to a month of blockade would be conjecture. What is documented is the amplification of a risk, flagged by the WFP, the FAO and the Lancet authors alike. One aggravating element, often overlooked, deserves noting: funding cuts also reduce data collection. The GRFC 2026 notes the lowest data availability in ten years, which complicates precisely the targeting of aid where it is most lacking. --- Primary sources: Lancet, Cavalcanti et al. (2025) and Lancet Global Health (February 2026); Center for Global Development; WFP Global Outlook and Global Report on Food Crises 2026 (FSIN / GNAFC, 24 April 2026); SOFI 2025 (FAO, IFAD, UNICEF, WFP, WHO); public statements by the US State Department on the scope of the USAID cuts. ============================================================================ ANALYSIS: Copper: the shortage arriving via Hormuz and El Niño URL: https://l0g.fr/en/analysis/copper-shortage-hormuz-el-nino/ Canonical French source: https://l0g.fr/posts/cuivre-shortage-ormuz-el-nino-2026/ Date: 2026-07-14 (reviewed 2026-07-14) Topics: macro, markets, geopolitics, energy, copper ---------------------------------------------------------------------------- Copper is no longer merely China's industrial thermometer. It has become the metal of electrification, data centres, grids and the geopolitics of inputs. As of 26 June, the London Metal Exchange showed three-month copper at $13,649 a tonne, day-delayed closing price, and recalled that its contract is the world benchmark for physical copper. That is very high. But the important point is not only the price: it is the quality of the shock. The market does not yet tell a simple story of a refined-copper shortage. The June ICSG figures even say the opposite on the surface: over January-April 2026, world refined production reaches 9.711 million tonnes, refined consumption 9.471 million, an apparent surplus of 239,000 tonnes. But beneath this surface, the mine is stalling: world mine production falls from 7.551 to 7.446 million tonnes, -1.4% year on year, while the mine capacity utilisation rate slips to 76.6%. In other words, refined output is still holding thanks to inventories, recycling, China and transmission lags. The mine, for its part, is already sending a tighter signal. // Copper: high price, refined balance still positive LME 3-month Mine production Refined balance Mine utilisation $13,649/t -1.4% +239 kt 76.6% Sources: LME, 3-month closing price day-delayed; ICSG Copper Bulletin, June 2026. The weak link: acid, not just ore The Hormuz crisis changes the reading. UNCTAD recalls that the strait concentrates about a quarter of world seaborne oil trade and significant volumes of LNG and fertiliser. For copper, the most insidious transmission channel runs through sulphur and sulphuric acid. The WSJ documented in April the surge in sulphuric acid, from $150 to $800 a tonne in some flows, and the strait's role in sulphur exports. Why does this matter? Because part of copper is extracted by leaching, then electrolysis, a process called SX-EW. It consumes sulphuric acid to dissolve the copper contained in oxide ores. When Gulf sulphur is blocked, it is not only oil that is missing. It is an industrial reagent without which some copper volumes become more expensive, slower, even temporarily unprofitable. El Niño adds weather risk Second layer: climate. On 11 June, NOAA's Climate Prediction Center placed ENSO under an El Niño Advisory. Its summary is clear: El Niño is present, is expected to strengthen through the Northern Hemisphere winter of 2026-2027, and NOAA gives a 63% probability of a very strong episode in November-January. This figure does not prove a production rupture. It says the tail risk is thickening. The link with copper runs through geography. According to the USGS, Chile produced 5.3 million tonnes of mine copper in 2025, Peru 2.7 million, the DRC 3.2 million. Chile and Peru therefore concentrate a major fraction of world mine supply. And El Niño can raise the risk of extreme rainfall, landslides, slowed ports and regional water constraints on the South American Pacific coast. The right phrasing is not "El Niño will close the mines". The right phrasing is: when the mine is already running at 76.6% of capacity, each weather incident matters more. // Two shocks, one critical metal Hormuz sulphur, freight, energy El Nino rain, ports, water, roads Copper tight mine + electric demand Reading: Hormuz acts on inputs and logistics; El Nino acts on mining operational risk. Rigid demand, limited substitution On the demand side, copper remains hard to replace in power grids, transformers, motors, data centres, vehicles and construction. The IEA already places it at the heart of the energy transition and electrical infrastructure. The WSJ noted in late 2025 that electrification, renewables, electric vehicles and data centres were supporting the price even as mining accidents reduced the supply cushion. The WSJ of 26 June also reports that Maybank raised its long-term copper assumption to $9,260 a tonne, a sign that high prices are no longer treated as a mere spike. One must nonetheless stay cool. The ICSG does not yet show a world refined deficit over the first four months of 2026. End-of-period refined copper stocks rise to 2.108 million tonnes in April, against 1.373 million a year earlier. The real stress is therefore less in the instantaneous accounting balance than in the stacking of fragilities: falling mine output, more expensive acid, Hormuz freight, El Niño risk, rigid electrical demand. The working scenario My central scenario: copper stays expensive as long as the market lacks proof that the Hormuz normalisation holds, that sulphuric acid becomes available again and that the El Niño episode does not disrupt Andean production during the austral summer. The level of $13,649 a tonne already prices in a lot of stress. But a real shock to SX-EW or to Chilean and Peruvian ports would force the market to reprice physical scarcity, not only the long electrification story. The signal to watch is therefore not only the LME price. One must look at concentrate treatment charges, LME and COMEX inventories, sulphuric-acid prices, NOAA ENSO bulletins, and the operational announcements of the major producers. For the price dashboard, our guide on reading the copper market remains the entry point. For the geopolitical shock, the logical thread starts from the situation report on Hormuz and the normalisation of the strait. Conclusion: copper is not yet in a documented world shortage. It is in a phase of credible pre-shortage. And in a metal where supply takes ten years to arrive, that nuance is exactly the one the market prices before the official statistics. --- Sources: LME Copper, ICSG Copper Bulletin, June 2026, Table 1, USGS Mineral Commodity Summaries 2026, Copper, NOAA CPC ENSO Diagnostic Discussion, 11 June 2026, UNCTAD, Strait of Hormuz disruptions, IEA Global Critical Minerals Outlook 2025, WSJ, An Acid Test for the Global Economy, WSJ, Copper Price Forecast to Rise as Supply Cushion Dwindles, WSJ Basic Materials Roundup, 26 June 2026. Market data accessed 26 June 2026. This is not investment advice. ============================================================================ ANALYSIS: China: behind the fall in crude imports, a new market power URL: https://l0g.fr/en/analysis/china-crude-imports-fall-market-power/ Canonical French source: https://l0g.fr/posts/chine-importations-petrole-brut-chute-2026/ Date: 2026-07-14 (reviewed 2026-07-14) Topics: china, oil, energy, macro, geopolitics ---------------------------------------------------------------------------- In May 2026, Chinese crude oil purchases fell to their lowest level in eight years. The figure is spectacular, but reading it requires distinguishing three phenomena: the disruption of Gulf flows, the economic arbitrage of refiners and the structural erosion of fuel demand. After a record of 11.55 million barrels a day (m b/d) in 2025, then 11.99m b/d over January-February 2026, Chinese imports fell back to 11.77m b/d in March, 9.36m b/d in April and only 7.79m b/d in May, according to Chinese customs and Reuters. This last volume, 33.08 million tonnes, represents a decline of about 29% year on year. Taken in isolation, the figure might suggest that the world's largest importer has just passed its oil peak. That would be too quick: the decline is first the product of a supply shock and a refusal to pay top price for the marginal barrel. China crude oil imports Averages in million barrels a day: 11.55 in 2025, 11.99 in January-February 2026, 11.77 in March, 9.36 in April and 7.79 in May. // China crude oil imports million barrels a day · annual or monthly average 0 4 8 12 11.55 11.99 11.77 9.36 7.79 2025 Jan-Feb 2026 Mar Apr May Sources: Chinese customs, Reuters · tonnes/barrels conversion per the cited publications After the massive stockpiling of early in the year, the decline concentrates on April and May 2026. Hormuz broke the supply chain The first factor is physical. The near-paralysis of the Strait of Hormuz after the conflict with Iran began dried up part of the Gulf arrivals. In April, China is estimated to have received only 648,000 b/d of crude that had transited the strait, against 4.07m b/d on average between January and March, according to Kpler. Refiners dependent on Middle Eastern grades could not immediately replace these barrels: Russian, African or American grades offer neither the same characteristics, nor the same transport times, nor always the same payment terms. The shock described in the analysis of the Asian bill of the Hormuz crisis therefore transmitted directly to Chinese terminals and refineries. Beijing did not "destroy" four million barrels a day of consumption in a few weeks. It absorbed a supply shock by cutting purchases, refining runs and fuel exports. This distinction is essential: a decline caused by a maritime bottleneck can reverse faster than a durable contraction in final demand. Refiners refused to pay for the marginal barrel The Chinese reaction also stems from a price arbitrage. In May, the premium of Saudi Arab Light reached as much as $19.50 a barrel over regional references. Refiners preferred to cut their runs rather than turn overpriced crude into gasoline or diesel sold on a sluggish domestic market. In April, crude throughput fell 5.8% year on year, to 13.3m b/d, its lowest level since August 2022. The utilisation rate fell to 63.6% and the average margin to -649 yuan a tonne. In May, runs fell further to 12.66m b/d, while inventories were drawn down only moderately. This behaviour shows that Chinese imports have become an optimisation variable more than a simple mirror of GDP. Having accumulated cheap crude in 2025 and early 2026, China can defer its purchases when physical differentials spike. It fills its tanks when sanctioned or surplus barrels are discounted, then disappears from the spot market when security of supply becomes too costly. A cyclical weakness against a structural mutation The geopolitical shock does not, however, explain everything. Chinese road-fuel demand is already losing steam. The International Energy Agency estimates that the Chinese electric fleet avoided about 1m b/d of oil consumption in 2025, nearly 15% of what Chinese road transport would have consumed with a purely combustion fleet. The rise of electric or LNG heavy trucks, high-speed rail and the persistent weakness of real estate also reduce diesel and gasoline growth. As early as 2024, crude imports had fallen 1.9%, the first annual decline outside Covid in two decades, before rebounding in 2025 on the back of refining and storage. The right reading is therefore that of a cyclical shock amplified by a structural trend. The reopening of Hormuz and the gradual return of Gulf cargoes can trigger a rebound in arrivals. Asian flows were already picking up in June, but Chinese buyers stayed cautious in the face of high prices. To track the normalisation of the market rather than a monthly snapshot, our guide on reading the oil market sets out the main energy prices. The balance of power has changed The May fall does not mean China is giving up on oil. Petrochemicals, aviation and reserve-building will keep supporting purchases. But the country no longer plays the role of an automatic engine of world demand: each additional barrel depends more on refining margins, inventory levels and the rebate obtained than on GDP growth alone. For OPEC+, the consequence is uncomfortable. The world's main buyer now has enough inventory, diversification and marginal restraint to bide its time, cut its runs and protect its domestic market by limiting fuel exports. China does not yet set the world oil price alone, but it can refuse the price offered. This is therefore not only a story of falling volumes. It is the sign that, on the physical market, part of the bargaining power has shifted from producers to the Chinese buyer. --- Primary sources: - General Administration of Customs of China, "China's Major Imports by Quantity and Value, Mar 2026", 8 April 2026. - General Administration of Customs of China, "China's Major Imports by Quantity and Value, Apr 2026", 8 May 2026. - Reuters, "China's 2025 oil imports, December inflows both hit record highs", 13 January 2026. - Reuters, "China January-February crude imports surge on higher refinery throughput", 10 March 2026. - Reuters, "China's commodity imports show Hormuz impact as oil slides, metals rise", 12 May 2026. - Reuters, "China's April oil throughput hits lowest since August 2022, inventories rise", 18 May 2026. - Reuters, "China's imports of major commodities show price remains key driver", 10 June 2026. - Reuters, "China did use crude stockpiles to ease Iran shock, but not that much", 17 June 2026. - Reuters, "Asia has plenty of crude oil, but refined fuels remain tight", 22 June 2026. - International Energy Agency, Global EV Outlook 2026, Outlook for electric mobility, 2026. - U.S. Energy Information Administration, "China, International energy analysis", accessed 26 June 2026. - Center on Global Energy Policy, Columbia University, "China's Slowing Oil Demand Growth Is Likely to Persist and Could Impact Markets", 13 November 2024. - Reuters, "China's crude oil imports fall in 2024, first time in two decades outside of COVID", 13 January 2025. ============================================================================ ANALYSIS: Oil: the Chinese inventory that caps prices URL: https://l0g.fr/en/analysis/oil-the-chinese-inventory-capping-prices/ Canonical French source: https://l0g.fr/posts/petrole-le-stock-chinois-qui-plafonne-les-prix/ Date: 2026-07-14 (reviewed 2026-07-14) Topics: oil, china, commodities, macro, geopolitics ---------------------------------------------------------------------------- As of 3 July 2026, Brent trades around $71 a barrel, down from $72.68 on 1 July. It is a counter-intuitive fact: a few weeks earlier, the partial closure of the Strait of Hormuz and the war around Iran had pushed the barrel toward $80, with a scenario at $105 if the strait stayed blocked. Yet the price came back down. Part of the explanation lies with an actor that produces almost nothing: China, the world's largest crude importer. For the fundamentals, see our guide on reading the oil market. The buyer that sets a floor, then a ceiling Beijing's role is double-acting, and that is what makes it poorly understood. In 2025, China played the role of a floor. Between January and August, it added about 900,000 barrels a day to its stocks, while Brent held around $68. Over that period, world inventories were swelling by 1.4 to 1.8 million barrels a day: China absorbed most of it. Without that appetite, according to the US Energy Information Administration (EIA), downward pressure on prices would have been far stronger. In other words, by buying cheap oil, China kept it from becoming even cheaper. In 2026, the same mechanism flipped to set a ceiling. When the barrel jumped with the Iran crisis, China did not bid up. Its imports fell to 9.25 million barrels a day in April 2026, the lowest level since July 2022, a drop of about 2.4 million barrels a day year on year, nearly 20%. The utilisation rate of its refineries fell to its lowest since August 2022. Beijing stopped buying at high prices and preferred to live off its reserves. By refusing to fight other importers over scarce, expensive oil, China withdrew demand from the market at the worst moment, which mechanically capped the surge. The war chest: 1.24 billion barrels This discipline is only possible because China holds a considerable cushion. Its onshore crude stocks are estimated at about 1.24 billion barrels in April 2026, which would make it the largest national reserve on the planet. The figure remains an estimate: Beijing does not publish the detail of its strategic reserves, and analysts reconstruct it from import flows, satellite data and tanker tracking. The order of magnitude, however, is a consensus. Notably, China kept filling its tanks even during the collapse in imports: between 430,000 and 580,000 barrels a day went into storage in April 2026, according to Reuters and Vortexa estimates. The fuel for this filling is not Gulf crude at market price, but discounted, sanctioned barrels bought at a rebate from Russia, Iran and Venezuela. About 166 million barrels of Iranian crude are said to be floating in Asian waters, positioned outside the Strait of Hormuz and closer to Chinese ports than to Middle Eastern terminals. Beijing can thus help itself from a stock already on the water, without fuelling the bidding war on the open market. The limits of Chinese leverage One must avoid attributing everything to Beijing. The retreat in Brent also owes, and perhaps above all, to two independent factors. First, the de-escalation around Iran lifted the geopolitical risk premium that was inflating the barrel. Second, supply is rising: seven OPEC+ countries are increasing production by 188,000 barrels a day from July 2026, after already raising their quotas by nearly 600,000 barrels a day between April and June. The market is slowly tilting from fear of shortage toward fear of surplus. Chinese demand is therefore not the only brake, but it is the silent one. J.P. Morgan sees Brent at about $60 on average over 2026; the EIA even anticipated a trough near $52 in the first quarter. In a structurally well-supplied market, an importer able to cut its purchases by 2 million barrels a day without suffering becomes a de facto stabiliser. China did not decide to keep prices low out of benevolence: it buys when it is cheap and withdraws when it is expensive, in service of its own energy security alone. The ceiling on prices is a side effect of this methodical opportunism. One unknown remains: the day Beijing stops building reserves and starts selling them, the same leverage will work in the other direction. Sources 1. Fortune, oil prices on 1 and 2 July 2026, Brent at $72.68 then $71.53: https://fortune.com/article/price-of-oil-07-01-2026/ 2. Trading Economics, Brent at $72.10 on 3 July 2026: https://tradingeconomics.com/commodity/brent-crude-oil 3. U.S. Energy Information Administration, Chinese strategic stockpiling supports prices: ~900,000 b/d added from January to August 2025, Brent stable around $68, trough anticipated at $52 in Q1 2026: https://www.eia.gov/todayinenergy/detail.php?id=66319 4. OilPrice, China boosts stocks despite the plunge in imports: 9.25m b/d in April 2026 (lowest since July 2022, -20% year on year), 430,000 to 580,000 b/d into storage, record onshore stock of 1.24bn barrels: https://oilprice.com/Latest-Energy-News/World-News/China-Boosts-Oil-Stockpiles-Despite-Import-Plunge.html 5. OilPrice, Chinese stocks as strategic leverage: ~1m b/d stored in 2025 around $60, ~166m barrels of Iranian crude in floating storage in Asian waters: https://oilprice.com/Energy/Crude-Oil/As-Oil-Surges-To-80-Chinas-Stockpiles-Become-Strategic-Leverage.html 6. Axios, how China kept a lid on world oil prices: https://www.axios.com/2026/05/29/china-oil-iran-war 7. CNBC, China cushions oil prices below $100 during the Iran war: https://www.cnbc.com/2026/06/08/china-oil-iran-war-us-israel-energy-prices-strait-hormuz.html 8. Sarkaritel, OPEC+ production increase of 188,000 b/d from July 2026, ~600,000 b/d raised between April and June: https://www.sarkaritel.com/opec-production-increase-july-2026/ 9. J.P. Morgan Global Research, forecast of Brent at ~$60 on average over 2026: https://www.jpmorgan.com/insights/global-research/commodities/oil-prices ============================================================================ ANALYSIS: "The American people as shareholders of AI": anatomy of a presidential scam URL: https://l0g.fr/en/analysis/american-people-shareholders-of-ai-presidential-scam/ Canonical French source: https://l0g.fr/posts/le-peuple-americain-actionnaire-de-lia-decryptage-dun-scam-presidentiel/ Date: 2026-07-14 (reviewed 2026-07-14) Topics: finance, ai, us politics ---------------------------------------------------------------------------- So here we are. On Friday, aboard Air Force One, Donald Trump told reporters that his team is "looking at" the idea of AI labs giving the American public a stake in their companies, that the federal government would become "essentially a partner" of the firms, and that there is "something very interesting there". The full quote of the programmatic apparatus: "We're going to look at it." Monday morning, in pre-market, Nvidia +1%, Marvell and Micron between +4% and +7%, AMD +1%. An offhand phrase on a plane moved tens of billions of market cap. That is the only mechanism that actually worked here. And that is the whole point. The precedent that says it all: the "strategic Bitcoin reserve" Remember the media buzz of March 2025. An executive order, "Strategic Bitcoin Reserve", America becomes a "Bitcoin superpower", and so on. Fifteen months later, where do things concretely stand? - The government holds ~328,000 BTC (~$25bn). Zero bought on the market: it is entirely bitcoin seized in court cases. - The order did three things: consolidate the coins scattered across agencies, ban selling them, and ask for "budget-neutral" acquisition strategies without a single taxpayer dollar. - In August 2025, Treasury secretary Scott Bessent confirmed in black and white that the United States "will not buy" additional bitcoin. - The laws meant to turn this into a real purchase programme (Lummis's BITCOIN Act, Begich's ARMA) are still stuck in Congress. The "moderate" version even quietly dropped the million-BTC target. Net result: a reserve that is nothing but a sticker slapped on a pile of seizures, plus a ban on reselling. The narrative promised a sovereign buyer; the operational reality is a vault that does nothing. Maximal promise-to-delivery gap. That is the template. And "the people's stake in AI" ticks the same boxes. Why "the people as shareholders of AI" will not happen (in this form) 1. There is no legal vehicle for it. "A partnership with the American people" means nothing in corporate law. The people do not hold shares. At best, a sovereign wealth fund (US SWF) carries the stake, that is, the state, not you. The phrase "a stake for the people" is rhetorical packaging, not a capital structure. 2. The timing gives away the intent. The idea "rises" right after Bernie Sanders announced a one-off 50% tax on AI-lab shares, earmarked for a sovereign fund. Trump himself admits it "resonates" with part of his base and could "ease the anxiety" of a public facing AI. Translation: you appropriate the opponent's populist framing to neutralise it, without committing to any mechanism whatsoever. Political defusing at zero cost. 3. This administration cannot even sign an AI executive order. A useful reminder, in the same article: the signing ceremony for the 21 May AI order was cancelled at the last minute because the tech industry opposed it. The revised version merely asks for voluntary cybersecurity tests. When you cannot impose a voluntary audit, you do not nationalise a slice of OpenAI the following Monday. 4. The targeted companies are going public. Anthropic, OpenAI, SpaceX are aiming for IPOs at trillion-dollar valuations. Diluting these deals by bolting the state onto the cap table right before the listing? Trump's VC advisers, David Sacks first among them, who has already trashed the Sanders proposal, will not let it pass. The balance of power is known. The honest objection: "but Intel, MP Materials..." And here I will not feed you a line, because this is the angle from which people will try to contradict me. Yes, this administration takes direct stakes, and it is not hot air: - Intel, 22 August 2025: conversion of ~$8.9bn of CHIPS aid into ~10% of the equity. - MP Materials (rare earths), July 2025: the DoD puts in $400m of convertible preferred shares plus a warrant on ~11.2 million shares at $30.03. - Plus the Nippon Steel / US Steel golden share, the TikTok file, and others. So the state knows how to take stakes. But look at the nature of these operations: targeted, negotiated, national security, precise financial terms (preferred, warrants, strike price). This is niche industrial policy, contractualised, on assets deemed strategic. It has nothing to do with "the American people receive a share of AI". The Intel precedent makes plausible one narrow stake in one lab, negotiated behind closed doors. It does not remotely make credible the version sold on the tarmac, universal popular participation. What could happen is not what is being announced. That is the whole sleight of hand. The real product sold: an option on a narrative That is all there is: the announcement is the deliverable. It costs nothing, commits to nothing, and immediately produces two effects: it lifts chip stocks and it steals Sanders's "AI for all" flag. The political cost of not following through? Nil... as with the Bitcoin reserve, in twelve months no one will demand an accounting of "the partnership with the people". This is a communication operation monetised by the market, not a policy. The tell is always the same: no mechanism, no legal basis, no timeline, just a conditional spoken in front of a camera. Exactly as the United States never actually started buying Bitcoin for its reserves, it rebranded a pile of seizures and banned selling them. My bet: a White House meeting next week, a wishy-washy statement, maybe eventually a surgical stake in a single lab wrapped as a "win for Americans". The people's participation in AI, however, will stay what it is today: an article headline, and a green candle on Nvidia. ============================================================================ ANALYSIS: Strait of Hormuz, April 2026: the state of a blocked chokepoint URL: https://l0g.fr/en/analysis/strait-of-hormuz-situation-april-2026/ Canonical French source: https://l0g.fr/posts/point-situation-detroit-ormuz-avril-2026/ Date: 2026-07-14 (reviewed 2026-07-14) Topics: geopolitics, energy, macro ---------------------------------------------------------------------------- The situation as of 10 April On 28 February 2026, the United States and Israel launched an air campaign against Iran, during which the supreme leader Ali Khamenei was killed. In retaliation, Iran targeted vessels, laid mines and banned passage through the strait through the voice of the Islamic Revolutionary Guard Corps. In normal times, about 20% of the world's oil and a comparable share of liquefied natural gas transit through Hormuz, most of it bound for Asia (EIA, IEA). The effect was immediate. Traffic first fell by about 70%, and more than 150 vessels dropped anchor on either side of the strait; before the war, about 138 vessels transited it each day. Iran opened its own route north of Larak island and made passage subject to authorisation, charging tolls of up to $2m per vessel, settled in yuan according to Lloyd's List. As of 10 April, a ceasefire was announced but remained conditional, and US-Iran talks were expected in Islamabad. The oil shock, in numbers The Brent price illustrates the scale of the rupture. After an average of about $71 a barrel in February, it rose to $103 on average in March, $32 more, and the daily price reached nearly $128 on 2 April (EIA). In its Short-Term Energy Outlook of 7 April, the EIA raised its 2026 Brent forecast to $96 a barrel, against $78.84 a month earlier, and its WTI forecast to $87.41 against $73.61, anticipating a peak at $115 in the second quarter. The production outages in the Gulf, estimated at 7.5 million barrels a day in March, were expected to peak around 9.1 million in April according to the same source. The International Energy Agency called the event the largest supply disruption in the history of the oil market, with Gulf countries cutting their production by at least 10 million barrels a day. The shock absorbers Faced with the rupture, strategic stocks were mobilised. On 11 March, the IEA member countries decided to release 400 million barrels from their emergency reserves, an unprecedented scale. The OECD's emergency stocks count about 1.25 billion barrels held by governments and a further 600 million of industry obligation. On the US side, a sale from the strategic reserve was announced on 11 March, together with a 60-day Jones Act waiver to allow foreign vessels to carry products between US ports. The EIA then forecast a retail gasoline price peak around $4.30 a gallon in April, and diesel above $5.80. The IEA warned, however, that these reserves remain a stopgap, whose effect depends on how long the blockade lasts. The transmission to inflation This energy shock feeds directly into US consumer prices, of which energy is a volatile component. That is the subject of a separate piece on the return of US inflation. Update, since April The April ceasefire did not hold. The Islamabad talks collapsed on 12 April, and on 13 April the United States imposed a naval blockade of Iranian ports, maintained until 29 May. Brent peaked at $117 on average in April, its highest monthly level since June 2022, before easing to $107 in May (EIA). The war ended in mid-June: a memorandum of understanding between Washington and Tehran was signed on 17 June, opening the reopening of the strait and a sixty-day toll-free passage period. For the detail of the way out of the crisis, see the normalisation scenarios; for the big picture, the anatomy of the economic earthquake. --- Primary sources: EIA, Short-Term Energy Outlook (April, May and June 2026) and 7 April 2026 release; IEA, Oil Market Report (March 2026); Congressional Research Service, "Iran Conflict and the Strait of Hormuz"; Lloyd's List; Kpler; CNBC. ============================================================================ ANALYSIS: Iran and its Hormuz tolls in USDT on Tron: a masterclass in OFAC evasion URL: https://l0g.fr/en/analysis/iran-hormuz-tolls-usdt-tron-ofac/ Canonical French source: https://l0g.fr/posts/iran-peages-ormuz-usdt-tron-ofac/ Date: 2026-07-14 (reviewed 2026-07-14) Topics: geopolitics, crypto, sanctions ---------------------------------------------------------------------------- While Trump tweets and the phony Hormuz ceasefire holds together with Iranian duct tape, the IRGC (Islamic Revolutionary Guard Corps) has turned the strait into a high-tech cash machine. According to Blockonomi, tankers that want to pass through must now cough up between $1 per barrel and as much as $2m per supertanker... in USDT on the Tron blockchain. Payment in three seconds flat, no SWIFT, no Fed, no possible freeze by OFAC. It is not a blockade, it is a sovereign toll, crypto edition. And it is brilliant. For Tehran. The mechanism is diabolically simple. The vessel's operator emails the ship's details (owner, flag, cargo, crew) to an IRGC intermediary. The Hormozgan provincial command rates the vessel on a "friendliness" scale toward the United States and Israel (1 to 5). If the score passes, the fee is negotiated. Once the USDT transfer is confirmed on Tron, the IRGC sends a VHF code and an escort boat guides you through the Larak corridor. Alternative payment: Chinese yuan via CIPS (China's answer to SWIFT) through the Bank of Kunlun. Two options, zero risk of a US freeze. TRM Labs confirms it: the system has been operational since mid-March 2026, and at least two vessels have already paid in yuan. Why Tron and USDT? Because it is OFAC's kryptonite. The Tron blockchain (registered in the British Virgin Islands) settles transactions in under three seconds, with laughable fees and insane liquidity. USDT is pegged to the dollar... but lives outside the US banking system. Impossible for the Federal Reserve or OFAC to block it in real time. Chainalysis: the IRGC already moved $3bn in crypto in 2025, more than 50% of all Iranian crypto activity in Q4. TRM Labs traced $1bn through the offshore exchanges Zedcex and Zedxion (OFAC-sanctioned on 30 January 2026), almost entirely in USDT on Tron. Iran's central bank itself held $507m in USDT even before the conflict escalated (Elliptic). This is not improvisation, it is a financial-warfare infrastructure that has been ready for months. But Iran is not betting everything on the American stablecoin. Its evasion arsenal is a genuine sanctions-evasion 2.0 toolbox: - Yuan via CIPS: Beijing quietly funds the whole thing. At least two tankers have already paid in RMB. No dollar, no problem. - Bitcoin: Less liquid than USDT but impossible to freeze (no central issuer). Some reports explicitly mention BTC for the tolls, because Tether can in theory blacklist addresses... but not fast enough for a tanker waiting for its VHF code. - Hawala and informal networks: The good old traceless money-transfer system, juiced with crypto. The IRGC has used it for years to fund its proxies (Hezbollah, Houthis, and others). - Gold, precious metals and barter: Direct oil-for-gold or oil-for-goods swaps. Iran's "shadow fleet" (hundreds of ghost tankers) keeps selling crude through shell companies in the Emirates, Malaysia or China. - Offshore exchanges and mixers: Even after the sanctions on Zedcex/Zedxion, other unregulated (or "semi-regulated") platforms take over. Iran's central bank even set up a crypto exchange window on Qeshm island to convert USDT into rials or route it to foreign accounts. The result? The IRGC funds its war (missiles, drones, proxies) with the dollars of Western or Asian shipowners, while Trump keeps issuing Treasury bonds to pay for his aircraft carriers and his 2,400 air sorties. Poetic: the enemy uses the "United States Dollar Tether" to finance a war against the United States. As TRM Labs puts it, this is "the first conflict where the enemy's currency funds both sides". The ultimate irony? OFAC spent years sanctioning wallet addresses one by one. Iran industrialised the thing: a state-scale exchange infrastructure, rails already in place since 2025, and now a sovereign toll potentially generating hundreds of millions a month if traffic resumes. Even sanctioned, Zedcex and Zedxion processed $94bn in cumulative transactions. The Treasury's new GENIUS Act rules arrive too late: the crypto horse has already bolted the stable. While Trump screams, the IRGC cashes in silently via a blockchain no one really controls. Shipowners pay, tankers pass, and the revenue lands directly in the coffers of the Iranian war machine. This is not resistance: it is war fintech. And at this particular game, Tehran is winning the financial round hands down. ============================================================================ ANALYSIS: Semiconductors: the stack of constraints behind AI URL: https://l0g.fr/en/analysis/semiconductors-a-stack-of-constraints/ Canonical French source: https://l0g.fr/posts/semi-conducteurs-pile-contraintes/ Date: 2026-07-14 (reviewed 2026-07-14) Topics: markets, semiconductors, ai, geopolitics, valuation ---------------------------------------------------------------------------- import InfographicBars from '../../components/infographic/InfographicBars.astro'; The semiconductor is not a sector. It is a stack of constraints. At the top, the market sees AI, GPUs and the market cap of the winners. Below, there are wafers, lithography, memory, packaging, machines, energy, permits, talent, export controls and industrial timelines that do not compress with an earnings release. The right question is therefore not: "are chips expensive?" The useful question is: which part of the chain is already sold out, which part can still raise prices, and which part risks becoming the cycle's breaking point? The observable fact The cycle has restarted very strongly. The Semiconductor Industry Association, from WSTS statistics, announces on 6 July 2026 global semiconductor sales of $120.6bn in May 2026. That is 9.2% more than April 2026 and 104.1% more than May 2025. The same release calls May 2026 the highest monthly total ever recorded and flags a fifteenth consecutive month of monthly rise. This aggregate figure masks the composition. The SIA recalls in its introduction to semiconductors that components split into logic, memory, analog, sensors, optoelectronics and discretes. A car, an inverter, a modem, a GPU and a memory chip do not tell the same cycle. AI mainly pulls the top part: advanced logic, high-bandwidth memory, interconnect, advanced packaging and the associated production tools. The stack, not the ticker An AI accelerator sold by NVIDIA is not only an NVIDIA product. It is a supply promise across a whole stack: 1. chip design, software and design libraries; 2. wafers produced by an advanced foundry; 3. EUV and DUV lithography; 4. fast memory, often close to the processor; 5. advanced packaging to make compute, memory and interconnect communicate; 6. network, optics, power, cooling and the data center's electrical capacity. The top of the stack shows up immediately in the accounts. NVIDIA reported on 20 May 2026 a record quarterly revenue of $81.6bn for its Q1 fiscal 2027, ended 26 April 2026. Data Center weighs $75.2bn, up 92% year on year. Under the old reporting split, Data Center compute reaches $60.4bn and Data Center networking $14.8bn. This signal is a demand fact. It is not proof that all the sector's valuations are justified. Demand turns into margin only if each layer of the stack receives its share of capacity at the right moment. The foundry sets the pace The market can order GPUs faster than a foundry can build, qualify and fill an advanced line. That is the first reading point. As of 8 July 2026, TSMC's investor page for Q2 2026 indicates an earnings call scheduled for 16 July 2026 and a quiet period from 6 to 15 July. The data published on that date is therefore not yet an observed Q2 result: it is guidance. TSMC shows $35.90bn of realised revenue in Q1 2026 and guidance of $39.0 to $40.2bn for Q2 2026, with a gross margin expected between 65.5% and 67.5%. It is a load figure, not an oracle. It says the order book and prices still support the top of the stack. It does not say where the margins will be if the hyperscale customers slow, if geopolitical restrictions cut certain outlets, or if a chip generation arrives with a packaging or memory lag. Lithography is the cleanest bottleneck The purest part of the industrial monopoly reads at ASML. ASML describes its EUV systems as the technology that makes possible the mass production of the most advanced microprocessors. The company indicates that EUV uses 13.5 nm light, that this technology is unique to ASML, and that the NXE systems serve to produce complex layers for the 7 nm, 5 nm and 3 nm nodes. The technical detail matters because it makes substitution hard. To generate EUV light, ASML explains that a CO2 laser strikes tin droplets up to 50,000 times per second. EUV light is absorbed by almost everything, even air, so the optical path must be under vacuum. The wafer stage positions the plate with a precision on the order of a quarter of a nanometer and adjusts its position 20,000 times per second. ASML also indicates having invested more than €6bn in EUV R&D over 17 years. This is not a component you double by decree. It is an accumulated industrial system: optics, mechanics, vacuum, metrology, software, suppliers and know-how. ASML reports for 2025 €32.7bn of net sales, 52.8% gross margin and €4.7bn of R&D. The valuation of a supplier like ASML is debatable, but its role in the chain is simple: without a tooling layer, no advanced capacity. Geography changes slowly Reshoring is real, but it does not remove the concentration risk. CHIPS for America, on the NIST side, recalls that the CHIPS and Science Act entrusted $50bn to the Department of Commerce: $11bn for the R&D ecosystem and $39bn for incentives to facilities and equipment in the United States. The SIA / Boston Consulting Group report on supply-chain resilience projects a 203% rise in US manufacturing capacity by 2032, a US share of global capacity going from 10% to 14%, and a US share of advanced logic going from 0% in 2022 to 28% in 2032. The same document insists on the vulnerabilities that remain: advanced logic, legacy chips, memory, advanced packaging and key materials. Geopolitics adds another filter. In its Q2 fiscal 2027 guidance, NVIDIA specifies not including any Data Center compute revenue from China. That sentence alone recalls that the AI cycle is not only a cycle of private demand. It is also a cycle of permission: who has the right to buy, who has the right to sell, who has the right to produce, and under what constraints. Market reading The market prices three things at once. First, it prices demand. The SIA/WSTS sales and the NVIDIA figures say the observed demand is already massive. It is not a distant promise: the revenue exists. Next, it prices scarcity. EUV, advanced foundry, packaging and certain memories cannot be added instantly. When scarcity is real, margins rise and pricing power sets in. Finally, it prices duration. That is the fragile point. A semiconductor chain invests in years. AI demand sometimes reads in quarters. If the cloud investment cycle stays vertical, scarce capacity becomes an annuity. If customers slow before the new lines mature, the same capacity becomes a cycle risk. I therefore sort the theme into three floors: - industrial quality: ASML, TSMC, memory suppliers and advanced packaging have real barriers; - demand leverage: NVIDIA and the accelerators capture the visible part of the AI shock; - second-round risk: equipment, materials, secondary foundries and more commoditised chips may receive the orders after the peak of pricing power. Scenarios Central scenario. AI demand stays strong enough to absorb the available advanced capacity. Margins stay high among the players that hold a non-substitutable constraint: accelerator design, advanced foundry, lithography, fast memory, packaging. Stress scenario. The hyperscalers slow their orders or spread out their data-center plans. The visible part of the theme corrects first, then the market tests the upstream suppliers. The risk is not only the fall in volumes: it is the gap between a rising cost base and prices that stop rising. Geopolitical scenario. Export restrictions, industrial policies and tensions around Taiwan do not mechanically destroy demand. They fragment the chain. In that case, the duplication of capacity can support equipment orders, but reduce overall efficiency: more capex to produce a partial resilience. The l0g point Semiconductors have become a macro risk class because they mix four incompatible timeframes: the earnings quarter, the capex cycle, the long time of technology and the political time of export controls. The conclusion is not "buy the whole sector". The conclusion is narrower: semiconductors must be read as a map of bottlenecks. The figure that matters is not only market growth. It is the distance between final demand and the least stretchable layer of the chain. For now, that layer is not a single one. It is shared between advanced foundry, EUV, packaging, memory and data-center energy. It is precisely for that reason that the theme stays powerful. And precisely for that reason that it can break brutally if the market confuses AI demand, industrial capacity and durable annuity. ============================================================================ ANALYSIS: MiCA, Binance and Lagarde's shadow: what is established, what is narrated URL: https://l0g.fr/en/analysis/binance-mica-and-the-ecb/ Canonical French source: https://l0g.fr/posts/binance-mica-lagarde/ Date: 2026-07-14 (reviewed 2026-07-14) Topics: crypto, regulation, mica, binance, ecb, lagarde, stablecoins ---------------------------------------------------------------------------- On 1 July, MiCA's transitional regime expires. On that date, the largest crypto exchange in the world still has no European licence, and the specialist press pins this blockage on a personal intervention by the ECB president. Neither the central bank nor the Greek government confirms it. This article separates what is documented from what is, for now, narrative. MiCA, where the framework stands MiCA, Regulation (EU) 2023/1114 adopted in June 2023, is the Union's unified framework for crypto assets not covered by existing financial regulation. It applies in stages: the rules on stablecoins (ART and EMT) since 30 June 2024, the service-provider regime (CASP) since 30 December 2024, with a transitional window provided in Article 143 that closes on 1 July 2026. The central principle comes down to a mechanism: a licence granted by a national competent authority (NCA) opens a "passport" to the 27 member states. ESMA coordinates enforcement, the EBA supervises stablecoins. The state of play, in mid-2026, is one of slow conversion. Per the counts drawn from the ESMA register and sector tracking, about 210 of the more than 1,200 entities registered under the old national regimes have obtained the full CASP licence, on the order of 17%. The large licensed platforms count Coinbase, Kraken, Bitstamp, OKX, Crypto.com and Bitpanda. On the stablecoin side, only Circle's USDC and EURC figure among the large compliant tokens; Tether's USDT stays excluded from EU regulated markets, the issuer having refused to submit to them. Ten jurisdictions have still granted no licence, and Poland has still not adopted its implementing law, its president having vetoed it several times. Binance: the Greek path closes Binance filed its application in January 2026 via a Greek subsidiary, Binary Greece, with the Hellenic Capital Market Commission (HCMC). Athens had been chosen as European base; co-CEO Richard Teng praised its workforce and security profile in February. On 16 June 2026, Reuters, citing two people familiar with the file, reports that the HCMC is about to reject the application. The regulator does not comment, invoking confidentiality. Binance, for its part, says it worked eighteen months with regulators, judges its file MiCA-compliant, indicates the HCMC found it so and that it was reviewed at ESMA level, and maintains it received no formal notification of refusal. The stake is mechanical. Without a licence in a single member state, no passport; and without a passport, the halt of regulated services to EU residents on 1 July, or the presentation of an orderly wind-down plan. The platform now falls back on France and the AMF, presented by the reports as the only path still viable within the deadline; no application, however, has reportedly been formally filed there at this stage. The Lagarde hypothesis, and what can be said of it The angle comes from two concordant but unofficial sources. The Big Whale reported on 17 June, citing people familiar with the file, that Christine Lagarde reportedly played a decisive role in derailing the Greek application, after a message to Prime Minister Kyriakos Mitsotakis in May indicating that Binance was not welcome in Europe, an instruction then relayed to the HCMC via the finance ministry, which nonetheless supported the licence. Journalist Gareth Jenkinson (Cointelegraph) claimed on X to have been "informed by a reliable source" that she directly ordered the rejection. Neither the ECB nor the Greek government confirmed it, and no written decision establishes it. At this stage, it is a coherent narrative, not a proven fact. The motives ascribed to her read in the light of positions that are, themselves, public and verifiable. Lagarde is openly reserved about stablecoins: at a Banco de España forum in May, she judged the argument in favour of euro-denominated stablecoins more fragile than it appears, warning against the erosion of banks' ability to lend and of control over monetary policy. On 1 June, executive-board member Isabel Schnabel stressed the dominance of dollar stablecoins and the risk of anchoring American monetary influence. In the background, the ECB pushes the digital euro. A platform the size of Binance, a vector of dollar stablecoins, ticks precisely the boxes of what the president says she wants to contain. The motive is plausible; it does not amount to a demonstration. There remains an institutional objection that deserves better than a shrug. MiCA entrusts the licence to national authorities, not the ECB. If political pressure from Frankfurt overturned a national technical decision, the question would go beyond Binance: it would touch the independence of national regulators and the very promise of a harmonised single window. That is why the distinction between facts and narrative is not a comfort of prudence, but the heart of the matter. What it changes In the short term, if the rejection is formalised without a French relay in time, Binance will have to stop its regulated services to EU residents on 1 July, or present an orderly wind-down; the already-licensed rivals would mechanically capture euro volume. More broadly, two things are at stake: a possible precedent on a central bank's ability to bend a national licensing decision, and one more signal that MiCA, sold as a harmonised framework, stays politically disputed at its margins. Three elements will decide: a written HCMC decision, reasoned or not; a possible referral to ESMA, absent an extension of the deadline; and the AMF's timeline. Until then, honesty commands holding the two narratives at a distance from each other. --- Primary sources: Regulation (EU) 2023/1114 (MiCA), art. 143; interim ESMA register; Reuters (16 June 2026); The Big Whale and Cointelegraph / G. Jenkinson (17 June 2026); public statements by C. Lagarde (Banco de España forum, May 2026) and I. Schnabel (1 June 2026); Binance communication (June 2026). ============================================================================ ANALYSIS: Agent payments: how an AI settles the bill, and who holds the rails URL: https://l0g.fr/en/analysis/how-an-ai-agent-pays-the-bill/ Canonical French source: https://l0g.fr/posts/paiements-d-agents-comment-une-ia-regle-la-facture/ Date: 2026-07-14 (reviewed 2026-07-14) Topics: crypto, ai, stablecoins, payments, tech ---------------------------------------------------------------------------- In our overview of the convergence between crypto and artificial intelligence, one vector stood out as the most accomplished: agent payments. An AI program that buys data or compute on its own now settles its bill in stablecoins, in a few cents, with no human intervention. This piece goes one level deeper into the machine. How does an agent pay, concretely? Which standards clash, and is it really a clash? Who controls the rails, and how secure is all of this? Because behind the smoothness of the demonstration plays out a question of power, that of knowing who will hold the tap when billions of micropayments circulate between machines. The problem cards cannot solve The starting point is not ideological, it is economic. An autonomous agent that consumes a service, an API call, a data request, a second of compute, must pay very small amounts, very often, and machine to machine. Yet card networks are built for the opposite: a fixed fee of about $0.30 per transaction makes settling a one-cent payment absurd. Per the Keyrock report, 76% of agent payments fall below this threshold, most between one and ten cents. The card is not expensive for these flows, it is structurally impossible. It is this void the stablecoin fills. Programmable, continuously available, transferable without opening an account or signing a subscription, it allows payments the banking infrastructure refuses by construction. The money framed by the GENIUS Act thus becomes the natural fuel of agentic commerce. What remains is how an agent uses it without a human validating each step. The answer lies in a forgotten web code. x402, the resurrection of the 402 code When the HTTP protocol was designed, a status code, 402, was reserved for future use, labelled "Payment Required". It stayed dormant for nearly thirty years, for lack of an internet-native means of payment. In late 2025, Coinbase and Cloudflare woke it up with the x402 protocol, and the simplicity of the idea is its strength. Coinbase provided the Base chain and the USDC rails, Cloudflare the software component that lets any interface hosted with it accept these payments. The protocol is free and with no protocol fee, in stablecoins only since its second version of December 2025. What this unlocks was unthinkable with the card: charging an AI interface per request, web content paid per article, service markets between agents, autonomous purchasing of supplies. An agent can spend a budget over thousands of micro-transactions without a human authorising each one. It is powerful, and it is precisely there that the risk sets in, more on that later. Not a standards war, a stack The press loves the "protocol war" narrative, x402 against Google's standard against Visa's. The reality is more interesting: these bricks do not fight so much as they stack, each reigning over a different layer of the same payment. Google's standard, AP2, announced in September 2025 with about sixty partners including Mastercard, PayPal and Coinbase, handles authorisation: it encodes in signed mandates, carried as verifiable attestations, the fact that a given agent can spend a given amount, under given conditions, on behalf of a given user. Google donated it to the FIDO alliance in April 2026 for open governance. Visa's protocol, for its part, handles identity: the agent signs its requests with a private key the merchant verifies in Visa's registry, a digital passport. Mastercard rolls out its own version on its rails. x402 and stablecoins occupy the bottom layer, the one where value really moves. Seen this way, the convergence does not pit crypto against established finance, it assigns them floors. The players, and the dependence on USDC This stack has heavyweights on each floor. Amazon Web Services launched an infrastructure, Bedrock AgentCore Payments, which lets agents pay in stablecoins relying on x402 and on Stripe's wallets. Stripe, precisely, built an agent money stack from its acquisition of the Bridge infrastructure and its Privy wallets. Stablecoin issuers, Circle in the lead, provide the fuel. Most of these giants are listed or backed by listed companies, which gives the whole an adult infrastructure rather than a lab patchwork. This maturity has a flip side, concentration. Almost all agent payments settle today in USDC, Circle's stablecoin. It is a point of dependence on a single issuer, with the risk that implies: an incident on USDC, like its temporary depeg of March 2023 when part of its reserves was frozen at Silicon Valley Bank, would propagate instantly to the whole agentic economy built on it. The convenience of a de facto standard is paid in concentrated systemic fragility. Diversifying issuers would be safer, but would fragment liquidity, a classic trade-off we find everywhere in financial plumbing. Security, a gaping blind spot Here is the least mature side, and the most worrying. Giving an autonomous piece of software the ability to spend money, without human validation at each transaction, opens an attack surface research is only beginning to map. Several academic works published in 2026 inventory the flaws of the x402 protocol alone, with unambiguous titles, from a "systematic security analysis" to "five attacks" documented. The most characteristic flaw is prompt injection: a malicious instruction slipped into a page or an interface response diverts the agent and pushes it to pay where it should not. Add free-riding attacks to get the resource without paying, personal-data leaks in the negotiation phase, and a fundamental problem: since no human validates each payment, an erroneous or trapped instruction repeats at machine speed, thousands of times before it is noticed. The defences exist, strict spending caps, data filtering before execution, cryptographic verification of the agent's identity, the idea of a "Know Your Agent" modelled on "Know Your Customer". But they are young, and the gap between deployment speed and security maturity is the real Achilles' heel of this vector. Who holds the tap There remains the political question, the most interesting for anyone who follows power in finance. Is agentic commerce decentralising payment, or recentralising it differently? The execution layer, x402 and stablecoins, is open and permissionless, faithful to the crypto promise: any agent can pay any server without going through a gatekeeper. But the layers above, identity and authorisation, are being captured by the established players, Visa for the agent's passport, Google for the spending mandate. Yet whoever controls identity and authorisation controls the tap, even if the value flows on open rails. It is the telling sign we already noted about traditional finance: it adopts the stablecoin's function without ceding its toll-booth role. By seizing the high layers of the stack, card networks and cloud giants ensure that, even in a world of stablecoin payments, it is still their registry that will say which agent is legitimate and how much it can spend. The disintermediation promised by crypto moves one notch, it does not disappear. Payment becomes open, authorisation stays guarded. The points to watch To follow this vector without being carried away, a few markers. The first is real adoption, measured in settled volume, not to be confused with the trillion-dollar projections we put into perspective in the crypto-AI overview: the use case is proven, its scale stays modest. The second is the concentration on USDC, any incident on which would become systemic for the agentic economy. The third is security, where each protocol will have to prove itself against the attacks research already documents. The fourth is the sharing of layers: watching who wins identity and authorisation will say who really holds the power, whatever the settlement rail. Payment changes hands Agent payments are the most tangible part of the crypto-AI convergence because they solve a concrete problem nothing else solves: paying a machine, continuously, for almost nothing. The plumbing works, the giants are laying it, and the stablecoin finally finds a massive use outside speculation. But the smoothness of the demonstration must not mask the three reservations that will decide its trajectory: a still anecdotal scale, a dangerous dependence on a single issuer, and a security lagging the deployment. Payment is changing hands, from humans to agents. The real question is not whether this will happen, but who, of open crypto or the established gatekeepers, will hold the keys to the new tap. Sources 1. Amazon Web Services, "Agents that transact: Introducing Amazon Bedrock AgentCore payments, built with Coinbase and Stripe": agent payments in USDC via x402: https://aws.amazon.com/blogs/machine-learning/agents-that-transact-introducing-amazon-bedrock-agentcore-payments-built-with-coinbase-and-stripe/ 2. AWS Industries, "x402 and Agentic Commerce: Redefining Autonomous Payments in Financial Services": mechanics of the protocol and the facilitator contract on Base: https://aws.amazon.com/blogs/industries/x402-and-agentic-commerce-redefining-autonomous-payments-in-financial-services/ 3. CoinDesk / Keyrock report: ~$73M across 176M transactions, 76% of agent payments below the cards' $0.30 threshold: https://www.coindesk.com/business/2026/05/21/crypto-rails-are-becoming-the-default-payment-layer-for-ai-agents-report-says 4. Google Cloud, "Announcing Agent Payments Protocol (AP2)": signed mandates (intent, cart, payment) and launch partners: https://cloud.google.com/blog/products/ai-machine-learning/announcing-agents-to-payments-ap2-protocol 5. Visa, "New AI, Stablecoin and Token Innovations to Power Intelligent, Programmable Commerce": Trusted Agent Protocol and identity registry: https://investor.visa.com/news/news-details/2026/Visa-Announces-New-AI-Stablecoin-and-Token-Innovations-to-Power-Intelligent-Programmable-Commerce-at-Visa-Payments-Forum/default.aspx 6. "Free-Riding the Agentic Web: A Systematic Security Analysis of x402 Payments", arXiv:2605.30998: https://arxiv.org/abs/2605.30998 7. "Five Attacks on x402 Agentic Payment Protocol", arXiv:2605.11781: https://arxiv.org/abs/2605.11781 8. "SoK: Security of Autonomous LLM Agents in Agentic Commerce", arXiv:2604.15367: https://arxiv.org/abs/2604.15367 9. l0g, Crypto and AI: anatomy of a convergence, The GENIUS Act, the 18 July deadline and the guide Stablecoins and the GENIUS Act. ============================================================================ ANALYSIS: From attention to intention: the agentic economy and the resurrection of the 402 code URL: https://l0g.fr/en/analysis/the-intention-economy/ Canonical French source: https://l0g.fr/posts/economie-des-intentions/ Date: 2026-07-14 (reviewed 2026-07-14) Topics: crypto, stablecoins, ai, agents, advertising, macro, ecb ---------------------------------------------------------------------------- When an artificial intelligence navigates in your place and delivers you an answer, the ad network loses its point of insertion. Behind AI agents, another economic architecture is being built: pay-per-request, stablecoins as monetary vehicle, and an old HTTP code no one had finished wiring. This article reconstructs that stack and confronts it with the figures, academic research and the law. First clarification: the attention economy does not disappear, it moves, and the ability to predict our intentions remains, at this stage, more a commercial promise than a demonstrated fact. For twenty years, the web ran on a single market: that of your attention. A user searches, scrolls, clicks, and along the way crosses ads. Their time on site and their click rate are the raw material. This market has a size, and it is colossal. But it rests on a fiction installed as self-evident: the idea that information would be free. It never was. It was paid by a third party, the advertiser, in exchange for access to your behaviour. The arrival of AI agents, those software programs that act for you rather than displaying results, moves the point of insertion. When the agent answers, there is no longer a page to load, no space to sell in the same place. The question becomes economic: if part of the web ceases to be funded by pay-per-click advertising, by what will it be? An answer is assembling, made of open protocols, a pay-per-request layer, stablecoins, and a debate on monetary sovereignty. The rest of this article describes this construction without overestimating it. Advertising: a cycle, not an eternal annuity, and above all not a corpse Let us start with the order of magnitude, because the figures that circulate often understate the market. Per WPP Media's (formerly GroupM) late-2025 forecast, global advertising revenue reached $1.14 trillion in 2025, excluding political advertising, up 8.8% year on year, with growth expected at 7.1% in 2026. The firm Dentsu, which measures differently, places the crossing of the trillion in 2026. The two houses diverge on the timing, which recalls a method rule: each figure depends on the definition used. We reason in attributed ranges, never in a single truth. On the digital share, the convergence is clear: about 73% of global ad spend. GroupM's late-2024 forecast put "pure-play" digital at $813 billion for 2025; Statista's estimate for the same year is $799 billion, with search as the top segment at $334 billion. Google, Meta and Amazon capture on their own nearly three-quarters of global digital revenue excluding China. For Google's parent company alone, advertising represents more than $200 billion of annual revenue, nearly 80% of Alphabet's turnover. One point must be made at once, because the analytical error is tempting: advertising does not die. It migrates. The clearest signal comes from the players reputed to be "killers" of the model. OpenAI, whose ChatGPT app passed 800 million weekly users at the end of 2025 then about 900 million in early 2026 (figures released by the company and relayed by Reuters), launched in early 2026 an advertising test inside ChatGPT, targeting free and "Go" users. The logic is exactly that of the old web: fewer than 10% of users pay, so the rest must be funded otherwise. Google, for its part, told advertisers in late 2025 it wanted to introduce advertising into its Gemini assistant in 2026. The attention model does not die, it settles into the new interfaces. This dominance has a documented qualitative cost. The search algorithm does not adjudicate truth: it reads mathematical signals (publication frequency, engagement, inbound links) and rewards optimisation. Anxiety-inducing content triggers more clicks than a nuanced analysis, so optimising for engagement structurally favours the sensational, and expertise that refuses the rules of SEO ends up made invisible. The author Cory Doctorow named this degradation: "enshittification", named word of the year 2023 by the American Dialect Society and then by the Macquarie Dictionary in 2024. Its three-step mechanics are precise: a platform first makes itself useful to its users, then exploits them for the benefit of its business customers, then captures the value for itself alone. The nuance imposes itself, because the diagnosis is often caricatured. It is less an intention than a predictable result of a business model backed by attention: the engines amplify a pre-existing cognitive laziness as much as they create it. And the platforms' power, real, is not absolute. The 2015 slide from the motto "Don't be evil" to the more malleable "Do the right thing", at the moment of Alphabet's creation, records a model in tension with its original promise: the strict separation between organic results and advertisers. The click against the figures The effect of AI-generated answers on traffic is no longer an intuition, it is measured. The Pew Research Center analysed in March 2025 the behaviour of 900 American adults who agreed to share their browsing, nearly 69,000 Google searches. Result: 18% of searches triggered an AI summary ("AI Overview"), and 58% of participants encountered at least one in the month. Above all, in the presence of an AI summary, the user clicked toward an external link only in 8% of cases, against 15% in its absence, nearly twice less. The click toward a source cited inside the summary fell to 1%. And the user ended their browsing session in 26% of cases after a summary page, against 16% without. There is the real pressure on publishers, quantified and dated: not a disappearance of traffic, but an erosion of the outbound click where AI answers directly. This observation underpins what follows, but it immediately calls for a counterpoint the alarmist commentary forgets. Because commercial behaviour tells a more nuanced and far more instructive story. Adobe Analytics data, based on more than a trillion visits to US retail sites, show that the traffic arriving from a generative AI was initially of poor commercial quality, then reversed. The conversion of AI traffic was about 49% lower than that of classic traffic in January 2025, a gap narrowed to 23% in July 2025. Then the flip: this same traffic converted about 31% better than non-AI traffic during the 2025 holidays, and up to 42% better in March 2026, a record per Adobe. Visitors coming from an AI spend 45 to 48% more time on the site and view about 13% more pages. The volume, meanwhile, exploded: traffic to retail from generative-AI tools jumped about 693% year on year during the 2025 holiday season. Two lessons, and they run against the triumphalist narrative. First, AI serves mostly the research and comparison phase: the user informs themselves via the agent, then converts, which explains longer sessions and a better-prepared basket. Second, the conversion reversal is very recent, specific to retail, and says nothing of a generalised collapse of advertising. Prudence forbids extrapolating from online sales to the whole information economy. The intention economy: a promise, and a warning The term deserves dating, because it carries two opposing readings. "The intention economy" was first a pro-consumer concept, coined by Doc Searls in a Linux Journal column in March 2006, then developed in his 2012 book and his ProjectVRM at Harvard's Berkman Klein Center: the idea that the customer, not the platform, would eventually control the data of their own purchase intentions. The contemporary version is markedly darker, and it is academic. In "Beware the Intention Economy: Collection and Commodification of Intent via Large Language Models", published on 30 December 2024 in the Harvard Data Science Review, researchers Yaqub Chaudhary and Jonnie Penn, of Cambridge's Leverhulme Centre for the Future of Intelligence, describe an emerging market where language models capture, manipulate and resell no longer attention, but motivation. Conversational agents can subtly influence intentions, for example by imitating the user's writing style to seem familiar, or by guessing their phrasing in advance. This point reframes any honest analysis, and it answers a legitimate objection: no, near-total prediction of behaviour is not settled. Chaudhary and Penn present the intention economy as a "concerning if unchecked" prospect, not as an established fact. They note that the formalisation of "intent" by the researchers themselves stays crude: a Microsoft team, in 2024, sorts user intentions into boxes such as "information seeking", "problem solving" or "leisure", which underlines how poorly the object is defined. Humans remain largely unpredictable, and the promise of reading intent is, at this stage, a commercial argument as much as a technical capability. The whole architecture described below is built on this promise, without having demonstrated it. 402 Payment Required: the web's forgotten code The HTTP protocol has contained since its first specifications a status code that stayed a dead letter for decades: 402 Payment Required. Reserved "for future use", it described a web where you would natively pay for access to a resource. For lack of a request-scale payment rail, that future never came. Two players that everything opposes have just resurrected it for the same reason: AI agents need to pay. On 1 July 2025, Cloudflare became the first major infrastructure provider to block AI bots by default on new domains, switching from an "opt-out" model to an "opt-in" one. The same announcement launched Pay Per Crawl, a marketplace where a publisher can demand payment each time an AI scrapes a page: allow for free, charge, or block. Mechanically, the bot presents a payment intent in the request header and gets a 200, or is returned a 402 with a price. Cloudflare acts as merchant of record and handles the settlement. The company serves about one-fifth of global web traffic, which gives the setup a real potential reach, but in beta the balance of power stays overwhelming: a small independent publisher has no leverage against the large models, and the immediate revenue is modest. On the crypto side, Coinbase published in May 2025 the x402 white paper, an open standard that reactivates this same 402 code to embed a stablecoin payment directly into the HTTP exchange. The principle holds in a round trip: a client requests a resource, the server responds with a price, the client signs a stablecoin payment, the resource is delivered, with no account or API key. Settlements happen mainly in USDC, on several chains. In September 2025, Coinbase and Cloudflare founded the x402 Foundation, since moved under the governance of the Linux Foundation, with about twenty members including Google, Stripe and Visa. The traction figures must be handled with caution, because they come from the standard's promoters and mix technical transactions with real payments: more than 35 million transactions on the Solana chain alone by March 2026, an integration into Stripe's PaymentIntents API, and a payment layer estimated at around $600 million on an annualised basis. Legacy web infrastructure and crypto aim at the same primitive: make payment native to the protocol, at request granularity. It is the material condition of the micropayment (paying a few cents for a precise answer) and of the streamed payment (paying continuously, as consumption flows), impossible with legacy rails, designed for human transactions, slow and costly in fixed fees. The agentic stack: MCP, A2A and an IETF draft For an agent to pay, it must first know how to talk to tools and to other agents. Three protocols structure this stack. Two are already de facto standards, the third is a draft whose status must be described honestly. MCP, the Model Context Protocol, introduced by Anthropic on 25 November 2024, is a framework based on JSON-RPC 2.0 that standardises how an AI reads files, executes functions and retrieves context from external sources. Before it, each integration was bespoke. Adoption was fast and cross-partisan: OpenAI adopted it in March 2025, Google DeepMind in April 2025, Microsoft integrated it into Windows and Copilot Studio. In December 2025, Anthropic transferred MCP to the Agentic AI Foundation, under the aegis of the Linux Foundation, co-founded with Block and OpenAI. There were then more than 16,000 MCP servers in circulation. Where MCP links an agent to tools, the A2A protocol, Agent2Agent, links agents to one another. Announced by Google on 9 April 2025, it lets agents from different providers discover each other, authenticate and delegate tasks. Google transferred it to the Linux Foundation as early as 23 June 2025, with the support of Amazon, Microsoft, Salesforce, Cisco, SAP and ServiceNow. Objectivity requires it: despite a solid architecture, A2A's momentum seemed to slow against MCP in the autumn of 2025. In matters of protocols, adoption often beats technical elegance. The third document is often over-interpreted. It is an individual Internet-Draft (draft-zeng-mcp-network-mgmt-01), published on 16 October 2025 by Zeng Guanming, an engineer at Huawei, valid until 19 April 2026. Neither an RFC, nor an adopted working-group document: an individual proposal, "work in progress", whose current version is moreover expired as of this article's date. On substance, the idea illuminates a trajectory: extend MCP so that network equipment (routers, switches) behaves as MCP servers. The draft defines seven tools, addressable resources and dedicated error codes. Today, a network controller must speak CLI, NETCONF, SNMP, gNMI and proprietary APIs; tomorrow, an agent would diagnose a fault via the same MCP channel it uses for everything else. The real reach will depend on adoption by an IETF working group, which is not settled. The money layer: who builds the agents' rail The most disputed layer is not technical, it is monetary. An agent that decides needs a programmable, instant, borderless, low-unit-cost monetary vehicle. Stablecoins tick these boxes, and they already serve as a concrete rail outside the lab, as shown by the settlement in USDT of the Strait of Hormuz tolls. Per a Federal Reserve staff note (April 2026), the aggregate market cap of stablecoins reached about $317 billion on 6 April 2026, up more than 50% since the start of 2025. The sector is highly concentrated: USDT (about $184 to $187 billion) and USDC (about $77 billion) represent the overwhelming majority. The US framework clarified with the GENIUS Act, enacted on 18 July 2025, which mandates full reserve backing; it is part of the regulatory movement detailed in the CLARITY Act and US crypto framing. One fact, finally, is decisive: per the ECB, about 99% of the stablecoin supply in circulation is denominated in dollars. On this base, two families of players clash. On one side, the crypto-native standards: x402 and its integration into Google's AP2 protocol, for which x402 is the stablecoin facilitator. On the other, the card networks, which do not intend to be bypassed. Visa launched Visa Intelligent Commerce on 30 April 2025 with nine founding partners, including OpenAI; Mastercard launched Agent Pay in April 2025, extended in June 2026 into "Agent Pay for Machines" for micropayments on the order of a fraction of a cent. OpenAI launched Instant Checkout on 29 September 2025 with Stripe, via the ACP, Agentic Commerce Protocol, first for Etsy, Walmart and Shopify. The most accurate reading grid comes from sector analysts: trust and authorisation at the top (AP2, Visa, Mastercard), execution and settlement at the bottom (x402, on-chain stablecoins). Visa plays complementarity, aligning its Trusted Agent Protocol with OpenAI's ACP and the x402 standard. The reality check tempers the enthusiasm. Per a sector estimate, only about 4% of consumers today let an AI finalise a purchase autonomously: the infrastructure arrives well ahead of the trust. And the volumes must be read rigorously. An aggregate transfer volume on the order of $33 trillion in 2025 (reported by Artemis and Bloomberg) is not to be confused with a real payment volume, far more modest. Who has already switched, and who resists The architecture ceases to be theoretical when you look at the companies operating it. Three concrete, quantified cases show the diversity of models, and the fact that none has renounced monetisation. Perplexity, the answer engine valued around $20 billion per the press, launched in August 2025 a revenue-sharing programme with publishers, Comet Plus. The mechanism breaks with pay-per-click: a $42.5 million endowment, an 80/20 split in favour of publishers, funded by a subscription at $5 a month, and remuneration triggered by direct visits, citations and agent usage. The first partners cited include Fortune, Time, Der Spiegel, Gannett and The Independent. The Comet browser, launched in July 2025 then made free in October, is its entry point. It is a model where the source is paid because it is cited, not because it attracts an advertising click. OpenAI illustrates the coexistence of models rather than the rupture. On one side, Instant Checkout turns ChatGPT into an agentic buying surface; on the other, the company introduced in early 2026 advertising into its free and "Go" offers, exactly the model said to be threatened. With about 900 million weekly users and annualised revenue above $20 billion in 2025, OpenAI does not choose between attention and intention, it stacks the two. Amazon, finally, shows there is no mandatory convergence toward open protocols. Its Rufus assistant, become "Alexa for Shopping" in May 2026, was used by more than 300 million customers in 2025 and generated nearly $12 billion of annualised incremental sales, per the Q4 2025 results published in February 2026. Its "Buy for Me" function executes a purchase on external stores on the user's behalf. But Amazon adopted neither MCP, nor x402, nor AP2: it keeps a walled garden, backed by its catalogue, its reviews, its logistics and its payments. Standards fragmentation is as probable an outcome as their unification. The return of regulation: GDPR, DMA, antitrust The most neglected angle of the agentic debate is legal, and it weighs heavily. An agent that decides and pays on its own runs first into the GDPR. Its Article 22 confers on every person the right not to be subject to a decision based solely on automated processing producing legal effects or significantly affecting them. An autonomous purchase, a service refusal, a financial arbitrage executed without human intervention fall within this perimeter. The agentic layer will not deploy in a normative vacuum. The European Digital Markets Act adds a structural constraint. Seven companies are designated "gatekeepers" (Alphabet, Amazon, Apple, Booking, ByteDance, Meta, Microsoft), precisely those building the agents and their interfaces. The Commission imposed its first fines on 22 and 23 April 2025: €500 million to Apple for practices restricting the referral of users out of its App Store, and €200 million to Meta for its "consent or pay" model, judged contrary to the obligation to offer a less data-hungry alternative. These decisions show that an agent's ability to steer, compare and conclude a transaction will be read against the rules on lock-in and consent. On antitrust, the case United States v. Google sets the bounds. After ruling in August 2024 that Google held an illegal monopoly on search and associated text advertising, Judge Amit Mehta handed down on 2 September 2025 behavioural, not structural, remedies: no divestment of Chrome, but a ban on default exclusivity contracts for search, Chrome, Assistant and Gemini, and an obligation to share certain index and usage data with qualified competitors. The final judgment came in December 2025, followed by cross-appeals in early 2026. The lesson is double-edged, and it answers directly the temptation to overestimate platforms' omnipotence: regulation bites, but Google keeps about 90% of the search market and can keep paying to remain the default engine. Platform power is contested, constrained, but resilient. The digital euro, or sovereignty against the programmable dollar The thesis that "everything will happen in crypto" runs into a blind spot: almost all this crypto is in reality dollars. When 99% of stablecoins are denominated in dollars, the agents' money layer is not neutral, it is dollarised. The European Central Bank seeks to counter this, and this is what gives the digital-euro project its full meaning. The project's state is documented and dated. On 30 October 2025, the ECB closed its preparatory phase. The official timeline is explicit: if the co-legislators adopt the regulation during 2026, a pilot could start in mid-2027, for a potential first issuance in 2029. The build cost is estimated at around €1.3 billion, plus about €320 million a year thereafter. In December 2025, Christine Lagarde summed up the situation: the technical work is done, the ball is in the political camp. She presents the digital euro, a central-bank digital currency (CBDC), as a sovereignty tool reducing dependence on Visa, Mastercard or dollar stablecoins. The private sector is not waiting: a consortium of about ten banks (including BNP Paribas, ING and UniCredit), Qivalis, is preparing a euro stablecoin. These worries surface even in the licensing files, as illustrated by the tug-of-war over Binance's MiCA registration. The objection to formulate is not ideological, it is factual. A private dollar stablecoin and a euro CBDC answer to different functions and balances of power. The real question: who controls the unit of account in which agents will settle their transactions? As long as the answer stays the dollar, via private American issuers, the neutrality promise of the agentic layer remains partial. An infrastructure can be technically decentralised and monetarily very centralised. Trust as an asset: alignment becomes the product In a world where AIs decide for us, the first criterion is no longer relevance, it is alignment. An agent that recommends a choice because a company paid behind the scenes is no longer a tool, it is a Trojan horse. For a player like Google, switching to agents that answer directly would amount to eating into the advertising machine that generates almost all its revenue, hence the temptation of an in-between, discreetly biased results. The tool's credibility collapses precisely there, and that is why Chaudhary and Penn's warning is worth as much as the promoters' promises. The intention economy diverges here from the attention economy without replacing it. In the latter, friction is profitable; in the former, value resides in the absence of friction and in trust. If the agent refers nowhere, the only thing you really buy is its neutrality. But two safeguards impose themselves. An open standard is not a neutral standard: MCP, A2A and x402 are governed by foundations co-founded by the same giants they are supposed to discipline, and open code moves the ground of capture toward governance. And methodological transparency has a discoverability cost: a site that refuses trackers and aggressive SEO sends fewer signals to the engines. In the attention web, this cleanliness is a penalty; in the intention web, it becomes an asset only if agents really value verifiability, with no guarantee that they do. The innovator's dilemma, formalised by Clayton Christensen, illuminates the final bet. A dominant company rarely dies from poor management, but because it clings to the profitable model that made its success and misses the next turn. If the giants refuse to sacrifice their advertising annuity to guarantee agentic neutrality, they will be challenged by players with a different native model. But challenged does not mean replaced: Visa, Mastercard, Google and OpenAI are moving into the agentic payment layer precisely so as not to be disintermediated. The most probable scena [...] ============================================================================ ANALYSIS: Persistent inflation risk in 2026: the Iranian energy shock, upward revisions and challenges for central banks URL: https://l0g.fr/en/analysis/us-inflation-risk-2026/ Canonical French source: https://l0g.fr/posts/risque-inflationniste-2026/ Date: 2026-07-14 (reviewed 2026-07-14) Topics: inflation, macroeconomics, fed, ecb, geopolitics, energy, tariffs ---------------------------------------------------------------------------- Persistent inflation risk in 2026: the energy shock of the Middle East conflict and the upward revisions of forecasts Inflation, thought to be on a durable disinflation trajectory toward central-bank targets (2% for the Fed and the ECB), saw a marked rebound in spring 2026. Driven by a major geopolitical shock, the Iran/Israel conflict and the disruptions of the Strait of Hormuz, the May 2026 data confirm an acceleration of consumer prices, both in the United States and in the euro zone. Forecasts are systematically revised upward. This phenomenon, called a "silent contagion" of energy pressures toward the core of inflation, poses a tangible risk for the rest of 2026 and beyond. This article examines the latest quantified data (BLS, Eurostat, ECB, Fed, forecaster surveys), identifies the transmission channels and sketches realistic scenarios for the second half of 2026, without excessive extrapolation. 1. United States: a clear inflationary rebound in May 2026 The Bureau of Labor Statistics (BLS) data published on 10 June 2026 for the month of May are unambiguous: - Headline CPI (all items): +0.5% month on month (adjusted) and +4.2% year on year (against +3.8% in April). It is the highest level since April 2023. - Core CPI (excluding food and energy): +0.2% MoM and +2.9% YoY (against +2.8% in April), highest since September 2025. - Energy: +3.9% MoM and +23.5% YoY. The energy index contributed more than 60% of the monthly rise in the overall CPI. - Gasoline: +7.0% MoM and +40.5% YoY. - Food: +0.2% MoM and +3.1% YoY. - Shelter: +0.3% MoM and +3.4% YoY. Source: BLS CPI News Release, 10 June 2026 and detailed report. Core PCE (the Fed's preferred measure), available through April, stood at +3.3% YoY (against +3.2% in March). The Cleveland Fed's nowcasts for June 2026 anticipate a headline CPI around 4.05% YoY. Macro context and labour market The unemployment rate stood at 4.3% in May 2026. The labour market stays relatively tight, limiting the disinflation of services components. Consumers' short-term (1-year) inflation expectations rose, while long-term expectations stay broadly anchored near 2% per the New York Fed and University of Michigan surveys. 2. Euro zone: inflation at 3.2% in May, highest since September 2023 Per Eurostat's preliminary data (May 2026): - Headline HICP: 3.2% YoY (against 3.0% in April and 2.2% a year earlier). Highest level since September 2023. - Core HICP (excluding energy and food): about 2.5% (up from 2.2% in April). - Energy: +10.9% YoY, the strongest rise since February 2023. Germany (+2.7%), France (+2.8%), Spain (+3.6%) and Italy (+3.3%) all show an acceleration. Eurosystem staff projections (June 2026) · published very recently: | Indicator | 2026 | 2027 | 2028 | Notes | |-----------------------------|----------|----------|----------|-----------| | Headline HICP (average) | 3.0% | 2.3% | 2.0% | +0.4 pp vs March 2026 | | Quarterly peak | 3.4% (Q3-Q4) | - | - | Energy | | HICP excl. energy & food (HICPX) | 2.5% | 2.5% (peak 2.7% early 2027) | 2.2% | Services ~3.3% at peak | | Energy | 8.4% | -1.3% | -0.1% | Peak 12.5% Q3 2026 | Assumptions: average oil price at $96.9/barrel in 2026 (then falling). The energy shock of the Middle East conflict is the main driver of the upward revision. The indirect effects on non-energy components stay contained thanks to weaker demand and the penetration of Chinese imports. Source: Eurosystem staff macroeconomic projections, June 2026 The ECB signalled a probable rate hike (the first in three years) at its June 2026 meeting to anchor expectations. 3. Exogenous shocks: the Iran/Israel conflict and US tariffs The geopolitical energy shock (main 2026 driver) - Start of the conflict: late February / early March 2026. - Closure / major disruptions of the Strait of Hormuz (20% of global oil trade). - Brent price: from ~$72/bbl in late February to a peak near $120/bbl, then ebbing toward $92-98/bbl in early June (extreme volatility depending on ceasefire hopes). - Direct impact: sharp rise in fuel, transport and energy-input prices. The IMF (World Economic Outlook April 2026) incorporates this shock in its baseline scenario (limited conflict +19% of energy prices in 2026): - Global growth: 3.1% in 2026 (downward revision). - Global headline inflation: 4.4% in 2026 (then 3.7% in 2027). - Adverse scenario (prolonged conflict + financial tightening): growth 2.5%, inflation 5.4%. Source: IMF WEO April 2026 The World Bank anticipates a 24% rise in energy prices in 2026 and inflation of 5.1% in developing economies. Tariffs and trade fragmentation US tariffs (existing or reinforced policy) add upward pressure on imported goods (clothing, electronics, certain consumer goods). Estimates from the San Francisco Fed and other institutions show a gradual pass-through over 6-12 months toward goods inflation and, indirectly, toward services. These two shocks (energy + tariffs) overlay already-"sticky" components: shelter in the US and services in the euro zone. 4. Forecasts revised upward for 2026 United States (recent surveys): - Survey of Professional Forecasters (Philadelphia Fed, Q2 2026): headline CPI Q4/Q4 2026 at 3.5%, core 2.9% (significant upward revisions versus previous surveys). - March 2026 FOMC SEP (since revised): 2026 core PCE around 2.7% median (upside risks acknowledged by many participants). Euro zone: see the ECB table above (3.0% headline 2026). Global: 4.4% per the IMF (baseline scenario). 5. Scenarios for the second half of 2026 and risks Central scenario (likely if gradual de-escalation) - Headline inflation peak in Q3 2026 (US ~4.0-4.3%, EA ~3.4%). - Gradual ebb in H2 thanks to base effects on energy (if Brent falls back toward $80-90/bbl). - Core stays high: US ~2.8-3.0%, EA ~2.5-2.7% through 2027. - Fed: hold or slight hike of the fed funds rate (currently 3.50-3.75%); possible first hike late 2026 if the data persist. - ECB: one or two 25-bps hikes in 2026. Adverse scenario (prolonged conflict or escalation) - Oil price $110/bbl durably. - US headline inflation 4.5% on average in H2, core 3.2%. - EA: HICP 3.5% on average in 2026, second-round effects on wages and services prices. - Global growth < 2.5% (IMF adverse). - Risk of de-anchoring of medium-term inflation expectations → more aggressive monetary tightening → risk of recession or light stagflation. Identified upside risks 1. Energy → core pass-through stronger than expected (production, transport, food costs). 2. Tariffs: cumulative effect on global supply chains. 3. Expectations: rise in market measures and short-term surveys. 4. Supply: persistent supply-chain disruptions (Hormuz + other geopolitical friction points). 5. Demand: US consumer resilience despite inflation. Downside risks: rapid resolution of the conflict + sharp oil fall + marked demand slowdown (rising US unemployment). 6. Implications for markets and monetary policy - Policy rates: "higher for longer" or "higher and hiking" becomes the base scenario. The Fed's new chair, Kevin Warsh, faces his first FOMC (16-17 June 2026) in a hot-data context. - Bond markets: US 10-year and Bund yields under upward pressure (possible steepening if growth resists). - Equities: defensive sectors (energy, utilities, consumer staples) favoured; growth and tech under pressure if rates rise. - Currencies: dollar supported by the expected rate differential. - Crypto / stablecoins: indirect correlation via macro risk and liquidity (risk of negative correlation in case of tightening). Conclusion: vigilance required through 2027 The May-June 2026 data mark a turning point: inflation risk is no longer residual but has become central again for the rest of the year and 2027. The geopolitical energy shock acts as an "accelerator" on economies already facing structural rigidities (shelter, services, tariffs). Central banks (the Fed under Warsh, the ECB) have no choice but to stay data-dependent and ready to tighten if necessary to avoid a de-anchoring of expectations. Investors must factor into their scenarios an average 2026 inflation significantly above the forecasts of early in the year (US headline probably between 3.7 and 4.2% on an annual average depending on the conflict's outcome; EA around 3.0%). "Disinflation" is not dead, but it is postponed and made more costly. The second half of 2026 will be decisive: everything will depend on the evolution of the Middle East conflict and the economies' capacity to absorb the shock without a surge in underlying prices. --- Sources and links (all verified June 2026) 1. US CPI May 2026 · Bureau of Labor Statistics: https://www.bls.gov/news.release/cpi.nr0.htm and detailed PDF. 2. Eurosystem Staff Projections June 2026 · European Central Bank 3. World Economic Outlook April 2026 · International Monetary Fund: https://www.imf.org/en/publications/weo/issues/2026/04/14/world-economic-outlook-april-2026 4. Trading Economics / aggregated data · US Inflation Rate, Euro Area Inflation (June 2026 updates). 5. Reuters, CNBC, Bloomberg · Real-time coverage of the Iran conflict, oil prices and market reactions (May-June 2026). 6. Survey of Professional Forecasters Q2 2026 · Federal Reserve Bank of Philadelphia. 7. FOMC Minutes & SEP March 2026 · Federal Reserve. 8. World Bank Commodity Markets Outlook April 2026 · for energy and emerging-market inflation forecasts. Article written on the basis of official public sources and market consensus as of 12 June 2026. The data are accurate as of the publication date of the official releases cited. The scenarios remain conditional on geopolitical developments and on Donald Trump's moods. ============================================================================ ANALYSIS: Warsh's first FOMC: the easy status quo, the rest much less so URL: https://l0g.fr/en/analysis/warsh-first-fomc/ Canonical French source: https://l0g.fr/posts/warsh-premier-fomc-juin-2026/ Date: 2026-07-14 (reviewed 2026-07-14) Topics: fed, warsh, fomc, inflation, rates, bonds, qt ---------------------------------------------------------------------------- The decision drops this 17 June at 2 p.m. Washington time. The status quo is settled in advance. Everything else, much less so. Kevin Warsh chairs this 16-17 June his first FOMC since his swearing-in on 22 May. The market expects no rate move: futures gave, on 13 June, nearly 97% odds of holding the range at 3.50-3.75%, unchanged since December 2025. The suspense is therefore not the decision, it is what surrounds it: the new dot plot, the chair's tone, and the fundamental question he has dragged since his nomination, the balance sheet. This checkpoint extends our coverage of the Fed's balance sheet under Warsh. Angle: the risk to the economy and employment, without losing sight of the bond market, which already commands. The setting: inflation has picked up, employment holds on the surface Warsh inherits a dual mandate under tension. Inflation has taken off again. May's CPI, published on 10 June by the BLS, comes out at 4.2% year on year, its highest since April 2023 and a third consecutive monthly acceleration, after 3.3% in March and 3.8% in April. Energy explains more than 60% of the month's rise, a direct consequence of the oil shock tied to the war in Iran and the Strait of Hormuz, a subject we documented in the US inflation comeback and persistent inflation risk. Relative good news for the Fed: the core, excluding energy and food, rose only 0.2% on the month, at 2.9% year on year. The shock stays for now confined to energy, without massive diffusion to the rest of prices. Employment, for its part, holds. May saw 172,000 job creations, well above the 85,000 expected, and unemployment stayed at 4.3%. It is the best three-month sequence in over two years. But under the surface, two signals weigh the other way. Hourly wages rise only 3.4% year on year, below inflation at 4.2%: purchasing power falls about 0.7%, for the second consecutive month. And long-term unemployment now represents 27.5% of the jobless, up 524,000 year on year. The labour market is solid on the surface, but the real economy of households is tightening. That is the whole trap: inflation argues for not cutting, eroding real income argues for not over-tightening. The dot plot, the real arbiter Like all the March, June, September and December meetings, this one comes with the SEP, the members' quarterly projections, including the famous dot plot. It is that which must be read, not the decision. In March, the median of the 19 participants placed the policy rate at 3.4% at the end of 2026, a single cut of 25 basis points, then another in 2027 to finish at 3.125%, a level also used for the longer run, the highest since 2016. The same grid saw PCE inflation at 2.7% at the end of 2026, growth of 2.4% and unemployment at 4.4%. A detail that matters: 16 of the 19 members already saw upside risks to core inflation. Three months later, inflation has accelerated and markets have broken from that grid: after May's CPI, futures now incorporate that the Fed's next move will be a hike, expected in December. Hence the day's stake. If June's median keeps its cut, Warsh maintains on paper a dovish bias against inflation at 4.2%, at the risk of credibility. If the cut disappears and the median rises toward 3.6% or more, the Fed validates what the market already prices, and signs a turn. The number to watch holds in one figure: the 2026 median. The bond market already commands While the Fed temporises, the bond market has ruled. The 10-year, the benchmark of the federal state's borrowing cost, trades around 4.45% on 15 June, after rising to nearly 4.7% at the height of the war, when it traded below 4% before the conflict. The 30-year touched 5.2% in mid-May, its highest since 2007. The 2-year, more sensitive to Fed policy, hovers around 4.05%. This curve tells a story the Fed does not control. The term premium, the extra yield demanded to lend long, is inflated by three forces: the Iranian energy shock, the deficits and the rise in defence spending. Along the curve, the market is already doing part of the tightening the Fed does not own on its short rates. The easing of the last few days, against a preliminary peace deal between Washington and Tehran on 15 June, shows how much these levels depend on a geopolitical variable, not a central-bank decision. The balance sheet, the weapon Warsh wants to draw This is where the new chair becomes unpredictable. Warsh is a long-standing critic of the Fed's balance sheet, swollen to about $6.6 trillion, nearly 25% of GDP against 6% before 2008. He denounces a "monetary dominance" where quantitative easing benefits financial-asset holders first, distorts markets and eases public borrowing. His stated intent: bring the balance sheet back toward $4 trillion, in coordination with the Treasury, the most aggressive balance-sheet normalisation the Fed has undertaken. His thesis, counter-intuitive: a smaller balance sheet could justify lower policy rates, the two tools ceasing to work at cross purposes. The problem, for the bond market, is exactly there. Accelerating quantitative tightening would return duration for the market to absorb, which mechanically pushes long rates and mortgage rates higher, at the very moment the term premium is already stretched. And the operational warning is recent: in late 2025, the combination of QT and state borrowing had drained the money markets, forcing the Fed to stop dead and buy short securities. Warsh's instinct therefore runs into two walls, the fragility of the long end and banks' demand for reserves. To watch: any mention today of a balance-sheet timeline would be a heavier signal than the dot plot. In the background, Jerome Powell, who stayed on the board to ensure continuity, publicly warned against political pressures on the institution's independence. Warsh's options The rate status quo is locked. The real choice is on the trajectory, and no option is comfortable. Keeping the cut in the dots means holding a dovish bias that inflation at 4.2% makes hard to defend, and that the market no longer believes. Removing it, or even leaning toward a hike, means aligning with the data, but tightening on an economy where households' real income is already falling and long-term unemployment is rising. Communicating less, as Warsh claims with his "messier meetings" and his retreat on forward guidance, means making each word heavier, therefore the market more nervous. And opening the balance-sheet project means pulling the riskiest lever for long rates already under strain. The bind is real. Warsh cannot cut without capitulating on inflation, cannot hike without hitting an already-weakened demand, and cannot shrink the balance sheet without lifting the long end he does not control. His first FOMC is less a decision than a positioning. To watch tonight: the 2026 median, the dispersion of the dots and any dissents, in both directions, and the slightest word on the balance sheet. A Duke University survey conducted in early June among former Fed officials gave half of them for a probable hike in 2026. The wildcard is called Hormuz: if the peace deal holds and oil recedes, the energy shock empties, the heart of Warsh's problem loosens, and the token cut becomes sustainable again. Otherwise, today's status quo will only have been the easy part. Sources - Federal Reserve, FOMC calendar and Summary of Economic Projections of 18 March 2026, - FXStreet, Warsh opens first Fed meeting June 16 with rate hold expected, 15 June 2026, - J.P. Morgan Chase, What To Expect at Kevin Warsh's First Federal Reserve Meeting, June 2026, - REX Shares, FOMC June 2026 Preview: The Decision Is Settled, the Dot Plot Isn't, 16 June 2026, - J.P. Morgan Asset Management, FOMC Statement March 2026 (dot-plot medians, longer-run 3.125%), - CNBC, CPI inflation report May 2026 (CPI 4.2% year on year, core 2.9%), 10 June 2026, - CNBC, Jobs report May 2026 (172,000 created, unemployment 4.3%), 5 June 2026, - BLS, Employment Situation, May 2026, - CNBC, Treasury yields slide as Iran deal drives rethink on Fed (10-year 4.45%, 2-year 4.05%, 30-year 4.96%), 15 June 2026, - CNN, 30-year US Treasury yield hits highest level in 19 years (5.2%), May 2026, - Axios, Battles to shrink the Federal Reserve's balance sheet begin (Warsh's balance-sheet doctrine), 20 May 2026, - Bloomberg, Warsh's Return Revives Tensions Over the Fed's $6.6 Trillion QE Hangover, January 2026, - Kiplinger, June Fed Meeting live updates (Duke survey of former officials), 17 June 2026, ============================================================================ ANALYSIS: The basis trade: at its highest per the Fed, moribund per the market URL: https://l0g.fr/en/analysis/the-fed-on-the-basis-trade/ Canonical French source: https://l0g.fr/posts/basis-trade-fed-radiographie-pari-record/ Date: 2026-07-14 (reviewed 2026-07-14) Topics: macro, markets, central banks, regulation, liquidity ---------------------------------------------------------------------------- On 22 June 2026, a Federal Reserve note X-rayed the hedge funds' Treasury book: $4 trillion of gross exposure, including a basis trade estimated at $830 billion, nearly double its 2020 peak. Three weeks earlier, the specialist press was announcing the slow death of that same arbitrage, rendered barely profitable. A trade at its peak and out of breath at the same time: the paradox deserves decoding. The basis trade is one of those invisible cogs we only talk about when it breaks. It contributed to the Treasury-market panic of March 2020, when the Fed had to buy hundreds of billions of dollars of debt to restore order. Since then, regulators and central bankers watch it. We described its mechanics in our reference piece, the basis trade at the heart of US debt. What is new is the quantified snapshot the Fed has just delivered, and the tension between what its data say and what the market says. The bet, in brief Let us recall the idea in one sentence. The price of a cash Treasury bond and that of its futures contract diverge slightly. A fund buys the security in cash, sells the future, and pockets the gap, the basis, at expiry. The gap is tiny, a few basis points, so it is amplified by leverage of 15 to 20 times, obtained by financing the purchase on the repo market, where cash is borrowed overnight against the security as collateral. The trade earns a steady income as long as funding stays cheap and volatility contained. It becomes dangerous when these two conditions reverse at the same time. The Fed's X-ray The note, signed by economist Phillip Monin, breaks down for the first time with this precision the $2.4 trillion of hedge funds' long Treasury positions, as of September 2025. The basis trade is its first brick: about $830 billion, or 35% of these long positions. Then come matched-maturity positions ($395 billion), curve-steepness bets ($375 billion) and swap-spread arbitrage ($305 billion). Two figures give the measure of the phenomenon. The basis trade is today nearly double its early-2020 peak, the one that preceded the liquidity crisis. And hedge funds now hold about 8.5% of all US debt in private hands, against 4.5% in early 2023. A player that was a twentieth of the market now weighs nearly an eleventh. The moribund-basis paradox Here is the tension. At the very moment the Fed documents this record, practitioners bury the trade. Per Risk.net, the basis has lost its allure: spreads become too tight, under the inflow of capital that chases them, and a repo funding cost that has risen. As a result, the trade's net income, the gap between what it earns and what its funding costs, has thinned to the point that many managers consider the operation finished. The two observations contradict each other only in appearance. Size measures a stock of positions accumulated over years; profitability measures the flow they generate today. A trade can be both enormous and barely remunerative: it is even the most uncomfortable situation, the one where massive positions earn almost nothing, and where the slightest setback tips the balance toward the exit. An arbitrage that no longer pays is an arbitrage one is tempted to unwind, and an unwind at this scale is never done quietly. Why the concentration worries Because the real fragility is not size alone, it is concentration. The Fed note says it plainly: the 50 largest funds carry about 90% of these exposures, and the combination of a large scale, a strong concentration and high leverage creates a potential for systemic stress. When a few players hold the same positions with the same leverage, they tend to sell at the same time. This is not a theoretical worry. In April 2025, swap-spread arbitrage, the basis trade's cousin, saw about $60 billion of positions unwind abruptly in a few sessions, under a volatility spike. Apollo's chief economist, Torsten Slok, warned as early as April 2026: this level of leverage exposes global bond markets to a shock if the positions were forced to unwind. Apollo says it is, moreover, reducing its own risk and building cash. Who will buy the debt if the bet retreats A more discreet angle deserves attention. By financing the purchase of cash Treasuries, the basis trade makes hedge funds an important marginal buyer of US debt, at the very moment the Treasury issues record amounts of it. If the arbitrage retreats, for lack of profitability or after a shock, a source of demand disappears from the auctions. The relay is not obvious, and its absence would be paid in higher yields. It is one of the reasons for the awakening of the term premium, that extra yield demanded to hold long debt, which we analysed in our piece on the awakening term premium. The basis trade is not only a stability risk; it is also, by implication, a pillar of financing the federal state. The wildcard of mandatory clearing A regulatory deadline can reshuffle the cards. The SEC mandates central clearing of Treasury transactions: cash from end December 2026, repo from end June 2027, after a one-year delay granted in 2025. By going through a clearing house like FICC, recently joined by CME, positions gain in transparency and flow netting, but face more systematic margins. For the basis trade, the effect is ambiguous: clearing can make it safer by reducing counterparty risk, or amputate it by raising its funding cost. In both cases, it will not leave it unchanged. How the bet can unwind Three outcomes emerge, to be held as analyst hypotheses and not forecasts. The first is a gentle deflation. The trade no longer earning, funds lighten it gradually, the hedge-fund share of Treasuries recedes, and the arbitrage goes out of fashion without causing a tremor. It is the outcome Risk.net's observation implies: a basis that dies slowly is a basis that does not break. The second is a reshaping by regulation. The mandatory-clearing timeline transforms the trade's conditions before a market shock does. Depending on the margin setting, the arbitrage contracts, shifts to other players or changes form. The transition itself carries a risk, if it forces position adjustments within a narrow window. The third is the forced unwind, the scenario everyone dreads. A volatility spike, a margin call, a brutal move in yields, and the most leveraged positions unwind in disaster. To meet the margins, funds sell their Treasuries, which pushes yields up, which triggers new margin calls: the March 2020 spiral, bigger. It is not the most probable scenario, but it is the one whose cost would be heaviest, and it is the reason the Fed, the OFR and the FSB keep an eye on it. The serious objections are not lacking One must guard against catastrophism, and several arguments plead for calm. The basis trade renders a real service: by linking the cash price and the futures price, it maintains the coherence of a $31 trillion market and provides liquidity to Treasury auctions. Without it, the US state would finance itself at a slightly higher cost. The Fed also has a net it did not have in 2020, the Standing Repo Facility, a permanent window where primary dealers can obtain cash against Treasuries, precisely to prevent a repo drought from degenerating. And a basis trade that deflates on its own, for lack of profitability, reduces the risk instead of increasing it: it is an orderly exit, not a panic. This reassuring reading nonetheless has its blind spots. The Standing Repo Facility has never been tested in a real storm, and nothing guarantees the deflation is slow: a barely profitable trade is a trade hanging by a thread, not a safe trade. In sum The Fed note and the market's verdict do not contradict each other: they illuminate two faces of the same object. The basis trade is simultaneously bigger than ever and less remunerative than ever, a massive stock backed by a drying flow. It is an unstable configuration by nature, without being an immediate alarm. The most probable remains a gradual retreat, aided by the clearing timeline. The most costly would be a disorderly unwind, in a debt market already heavy with issuance. Between the two, the deciding variable is not the displayed size, it is the speed at which fifty funds will decide, or be forced, to exit at the same time. Sources 1. Federal Reserve, FEDS Notes, Phillip J. Monin, "Decomposing Hedge Funds' U.S. Treasury Exposures", 22 June 2026: gross exposure of $4trn ($2.4trn long, $1.6trn short), basis trade ~$830bn (35% of longs, nearly double the 2020 peak), matched maturity $395bn, steepness $375bn, swap spread $305bn, 50 funds = ~90%, hedge funds at ~8.5% of private Treasuries: https://www.federalreserve.gov/econres/notes/feds-notes/decomposing-hedge-funds-u-s-treasury-exposures-20260622.html 2. Bloomberg, "Fed Says Basis Trade Key Driver of Hedge Fund Treasury Exposure", 24 June 2026: https://www.bloomberg.com/news/articles/2026-06-24/fed-says-basis-trade-key-driver-of-hedge-fund-treasury-exposure 3. Risk.net, "Treasury basis trade loses its allure as returns shrink", June 2026: spreads too tight and rising funding cost: https://www.risk.net/markets/7963653/treasury-basis-trade-loses-its-allure-as-returns-shrink 4. Bloomberg, "Apollo's Slok Warns Hedge Fund Treasury Bets Risk Market Shock", 17 April 2026: hedge funds at ~8% of the Treasury market ($31trn), against 3% five years ago: https://www.bloomberg.com/news/articles/2026-04-17/apollo-s-slok-warns-hedge-fund-treasury-bets-risk-market-shock 5. SEC, extension of the Treasury mandatory-clearing timeline (cash on 31 December 2026, repo on 30 June 2027) and approval of CME as a clearing house: https://www.sec.gov/newsroom/speeches-statements/uyeda-statement-update-continuing-work-toward-treasury-clearing-implementation-122325 6. Financial Stability Board, "Vulnerabilities in Government Bond-backed Repo Markets", 4 February 2026: https://www.fsb.org/uploads/P040226.pdf 7. l0g, The Treasury basis trade: the leveraged arbitrage at the heart of US debt. 8. l0g, US debt and the awakening of the term premium. 9. l0g, Repo and SOFR market guide. ============================================================================ ANALYSIS: Record auctions: the weekly referendum on US debt URL: https://l0g.fr/en/analysis/record-treasury-auctions-debt-referendum/ Canonical French source: https://l0g.fr/posts/adjudications-record-referendum-dette-americaine/ Date: 2026-07-14 (reviewed 2026-07-14) Topics: macro, debt, rates, us treasury, central banks ---------------------------------------------------------------------------- On 9 July 2026, the US Treasury sold its 30-year bond at the highest yield since 2007, a little above 5%. The auction again tailed, for the second time in a row, but foreign buyers turned up. Every week, these auctions are a discreet referendum: the market sets there, live, the price at which it agrees to finance a heavily indebted state. Here is how to read it. We watch the 10-year yield like a thermometer, without always seeing where it comes from. It comes, for a good part, from a trading room where, several times a month, the Treasury auctions its debt. These auctions are not an accounting formality: they are the only moments when real demand for US debt is observed, at a negotiated price rather than a supposed one. When this demand weakens, it shows there first, in the detail of the results, before it reads in the broad indices. July 2026 offers a clear illustration. Reading a Treasury auction Three figures are enough to take an auction's pulse. The first is the bid-to-cover, the ratio of bids received to the amount sold: above 2, appetite is judged decent. The second is the tail, the gap between the final yield and the one the market anticipated just before, on the When Issued segment. A positive tail means the Treasury had to pay more than expected to place its paper, a sign of softer demand than hoped. The third is the share of indirect bidders, the usual approximation of foreign demand, central banks and sovereign funds included. Our Treasury-market guide details this grammar; the essential is to read it together, because a single isolated figure misleads. July's verdict July's refunding week delivered two contrasting signals. On 8 July, the 10-year note went off without a hitch, at 4.58%, with a bid-to-cover of 2.59, clearly solid demand. The next day, the 30-year told a tenser story: awarded a little above 5%, its highest level since 2007, with a bid-to-cover of 2.30 and a tail of about half a basis point, the second in a row. A decisive nuance: indirect bidders took nearly $16.6 billion, a foreign participation that stays robust. The balanced reading is this: the Treasury still places its debt, but at a rising price. Two consecutive tails on the 30-year do not make a crisis, they signal demand that requires being better paid. The awarded yield, up from 4.876% in April to above 5% in July, measures exactly that: the cost of long borrowing rises, auction after auction. The supply wall This hardening is nothing mysterious. It answers a supply-and-demand equation whose two terms play against the Treasury. On the supply side, the structural federal deficit runs around 6% of GDP over 2025-2030, which mechanically swells the volume of debt to place, and the Fed's quantitative tightening returned more than $2 trillion of duration to the market since 2022. On the demand side, the big historical buyers are keeping a lower profile. The Fed, once the top buyer via quantitative easing, is today a net seller. Foreign demand, for its part, is not collapsing but stagnating: non-residents' holdings rise much slower than the debt stock over a decade, so their relative share shrinks. And a new marginal buyer, the hedge-fund basis trade, is retreating in turn, as we analysed in our piece on the $830 billion bet the market judges moribund. When supply rises and three categories of buyers retreat together, the adjustment happens through the price, that is, through the yield. The term premium is negotiated here This price has a name: the term premium, the extra yield demanded to hold a long bond rather than roll short placements. Per the New York Fed's ACM model, it stood around 0.73% in spring 2026, back clearly positive after a decade of zero or negative values, but still well below its historical median of 1.41%. In other words, normalisation is under way without being complete. We set this diagnosis in our piece on the awakening of the term premium; the auctions are its concrete stage. Every tail on the 30-year, every yield coming above the When Issued, is a small increment of term premium wrested by the market. The aggregate statistic the Fed publishes is only the sum of these weekly negotiations. Where the tug-of-war leads Three trajectories emerge, to be treated as hypotheses and not certainties. The first trajectory, the most probable, is demand holding. Auctions keep covering, carried by a now more attractive term premium and by the dollar's status. Financing costs more, the interest bill grows heavier, but without rupture. That is what July's week suggests, where even the tensest auction found takers. The second is a creeping buyers' strike. Not a crash, but an erosion: tails that widen, a bid-to-cover that erodes from one auction to the next, long yields that step up. The Treasury would respond by shifting its issuance toward short maturities, the T-bills, less sensitive to the term premium, a real flexibility but one that defers the problem and shortens the debt's maturity. The third is the Fed's forced return. If a link seizes, like the repo market in September 2019 or Treasuries in March 2020, the central bank buys back to restore order. This would be, de facto, a form of fiscal dominance: monetary policy put at the service of financing the state, at the cost of its credibility in fighting inflation. It is the least probable and most consequential scenario. The reasons not to panic Prudence commands not over-interpreting two tails. Several elements invite calm. July's week precisely showed solid foreign demand, with $16.6 billion of indirect bids on the 30-year alone: the thesis of a global disaffection with US debt is not borne out in the day's figures. A tail of half a basis point is tiny against history, and no US auction has ever failed for lack of buyers. The dollar remains the reserve currency, which guarantees structural demand for its safe assets, and the Treasury keeps the flexibility to arbitrate between maturities to smooth the pressure. This reading nonetheless has its limits. A normalising term premium is still a rising premium, therefore an interest bill that swells and eats into the federal budget. And the history of bond markets teaches that demand looks infinite until the day it is no longer, often without warning. The comfort of reserve-currency status is not an acquired right, it is a privilege earned auction after auction. At bottom We should neither dramatise a somewhat tense auction, nor trivialise a 30-year yield at its highest in almost twenty years. The truth of July 2026 holds in one sentence: America still finances its debt, but it finances it more and more expensively, and a growing share of the bill falls on private investors as the Fed and foreigners step back. The auctions are where this shift reads first, figure after figure. To follow them is to take seriously the only question worth asking on sovereign debt: not how much is owed, but who still agrees to lend, and at what price. Sources 1. US Treasury, TreasuryDirect, official auction results (10-year note of 8 July, 30-year bond of 9 July 2026): https://www.treasurydirect.gov/auctions/announcements-data-results/ 2. Bloomberg, "US 30-Year Bond Auction Set to Draw Highest Yield in 20 Years", 9 July 2026: 30-year yield at highest since 2007, bid-to-cover 2.30, second consecutive tail, indirect bidders ~$16.6bn: https://www.bloomberg.com/news/articles/2026-07-09/us-30-year-bond-auction-set-to-draw-highest-yield-in-20-years 3. Result of the 10-year auction of 8 July 2026 (yield 4.58%, bid-to-cover 2.59): https://www.kucoin.com/news/flash/us-treasury-10-year-note-auction-clears-at-4-58-yield-with-strong-demand 4. Federal Reserve Bank of New York, term-premium estimates (ACM model), ~0.73% in spring 2026, below the historical median of 1.41%: https://www.newyorkfed.org/research/dataindicators/term-premia-tabs 5. FRED, 10-year term premium (series THREEFYTP10): https://fred.stlouisfed.org/series/THREEFYTP10 6. Federal Reserve, Phillip Monin note on hedge funds' Treasury exposure and the retreat of the basis trade, 22 June 2026: https://www.federalreserve.gov/econres/notes/feds-notes/decomposing-hedge-funds-u-s-treasury-exposures-20260622.html 7. Yahoo Finance, Apollo warning (Torsten Slok) on the debt refinancing wave: https://finance.yahoo.com/economy/policy/articles/brace-14-trillion-debt-wave-185523964.html 8. l0g, US debt: the awakening of the term premium. 9. l0g, The basis trade: at its highest per the Fed, moribund per the market. 10. l0g, Treasury-market guide. ============================================================================ ANALYSIS: Semi-liquid private-credit funds and gating: the HLEND (BlackRock) case and the lessons for retail investors in 2026 URL: https://l0g.fr/en/analysis/semi-liquid-private-credit-gating/ Canonical French source: https://l0g.fr/posts/private-credit-semi-liquide-gating-hlend-2026/ Date: 2026-07-14 (reviewed 2026-07-14) Topics: finance, private-credit, alternative-investment, gating, blackrock, hlend, semi-liquid ---------------------------------------------------------------------------- Semi-liquid private-credit funds and gating: the HLEND (BlackRock) case and the lessons for retail investors in 2026 A Reuters article of 12 June 2026 reports that a BlackRock/HPS private-credit fund of about $25 billion (the HPS Corporate Lending Fund, or HLEND) received in the first quarter redemption requests representing 13.3% of its shares in circulation. The fund decided to honour only 5% of these shares, about $620 million. A smaller vehicle, the BlackRock Private Credit Fund (BDEBT, $2.7 billion), saw requests of 5.3% and redeemed 5% (about $83 million). These decisions are part of a broader wave of withdrawals seen in 2025-2026 on private-credit funds open to wealthy investors. This phenomenon highlights the concrete workings of semi-liquid private-credit funds and, for many retail investors or family offices, the operational discovery of gating mechanisms. What is a semi-liquid private-credit fund? Private credit mainly designates senior secured direct loans granted to mid-market companies (often with an EBITDA of several tens to hundreds of millions of dollars). These loans are generally floating-rate, with credit spreads, and held to maturity (typical duration 5-7 years). They offer attractive yields relative to listed public credit, with historically lower volatility and low correlation to equity/bond markets. To widen access beyond traditional institutional investors, managers developed vehicles registered with the SEC: non-traded Business Development Companies (BDCs) or tender offer funds / interval funds. These structures, marketed via wealth-management platforms, target qualified or wealthy investors (accredited investors, qualified purchasers per the US thresholds). Unlike classic private-equity funds (7-10-year or longer lock-ups with sporadic distributions), these vehicles promise periodic liquidity: redemption windows (tender offers) generally quarterly. This is what makes them "semi-liquid". In exchange, they keep the advantage of private-credit yields (distributions often monthly or quarterly from the interest received). The liquidity mechanism and gating: how it works concretely The fund periodically organises an offer to redeem a certain percentage of its shares in circulation. The very widespread practice on these private-credit vehicles is a 5% cap per quarter (sometimes up to 25% for some interval funds, but 5% is the common standard for large tender-offer BDCs like HLEND or the Blackstone Private Credit Fund, BCRED). If redemption requests stay below or equal to the cap, the fund redeems everything (or pro rata if a slight excess). If requests largely exceed the cap, this is gating, the fund limits redemptions to the amount authorised by its liquidity policy (generally pro rata to the requests). Excess requests are deferred to the following windows or handled per the prospectus rules. Stated objective: align the liquidity offered to investors with the illiquid nature of the underlying assets. Direct loans do not trade easily on a deep secondary market; a mass forced sale could entail significant discounts and hurt the net asset value (NAV) of all remaining shareholders. Gating therefore protects the long-term investment strategy and avoids a "bank run" on illiquid assets. In the HLEND case, the letter to shareholders confirms: requests of 13.3% of shares as of 31 March 2026, redemption limited to 5% (about $620 million), in line with the fund's usual liquidity parameters. The fund also stresses a conservative portfolio ( 95% first-lien senior secured), low leverage (1.0x, the low end of the target range), liquidity estimated at $7.2 billion (borrowing capacity, cash and liquid assets) and subscriptions plus distribution reinvestment expected to more than offset redemptions in the first half of 2026. Similar phenomena hit other large vehicles: Blackstone capped BCRED (about $79 billion) at 5% in the second quarter of 2026 after requests of 10%; Apollo anticipates persistent withdrawals on its retail/wealthy funds. Why this wave of redemptions and this "discovery" of gating in 2025-2026? Several converging factors, documented by managers and observers: - Prior massive inflows: during the low-rate period, retail and wealthy investors were massively steered toward these products via wealth-management advice, attracted by high distributed yields (often 8-12% annualised) and a "quarterly" liquidity presented as progress over classic private equity. - Slowing inflows and accelerating outflows: per RA Stanger, sales of non-traded BDCs aimed at wealthy investors fell 45% in the first quarter of 2026 versus the first quarter of 2025 ($8.9 billion against $16.3 billion). Kevin Gannon (Stanger) speaks of a "rotation of capital out of private credit" now well under way. - Cited concerns: doubts about credit quality and the transparency of private-loan valuations; fears tied to artificial intelligence's potential impact on some borrowers (notably in the tech or software sectors financed during the low-rate period); an observed rotation toward strategies backed by tangible assets (real estate and infrastructure, which saw their inflows rise). Apollo notes that the funds' underlying performance stays "solid", but that "we are not yet out of the turbulence" and that managers are learning to distinguish "long-term" investors from more flow-sensitive "tourists". In parallel, billions of dollars of redemption requests were "trapped" behind the caps industry-wide in early 2026. Factual analysis: the model's strengths and limits for the retail investor Structural positives: - Gating is contractually provided for and disclosed in the prospectuses and letters to shareholders. It is not a legal surprise. - It fulfils its protective role: it avoids forced sales that could degrade the NAV for all. HLEND highlights solid portfolio metrics (borrower EBITDA growth, interest coverage, moderate fund leverage) and notes that an environment of persistent or rising rates, with widening credit spreads, could offer better opportunities. - Realised yields have often been at a premium to listed public credit (HLEND claims an annualised excess of about 3.8% since its inception in 2022). Factual limits and points of vigilance: - Liquidity is conditional and capped. The term "semi-liquid" masks a reality: in case of concentrated requests, a significant fraction of the capital may not be immediately available. This is the concrete discovery many retail investors or family offices are currently making. - Duration mismatch: the assets (5-7-year loans) have a duration well above the quarterly redemption windows. The model rests on the assumption that requests will stay moderate and spread out over time. When this assumption is tested at scale, gating kicks in. - Valuations: private loans are valued at fair value by the manager (often with the support of a committee or a third party), and not marked to market daily like listed bonds. This can generate debates on transparency and the possible smoothing of NAVs, a subject regularly raised by the market. - Product scaling: democratisation has widened the investor base to profiles more sensitive to news and portfolio rotations. Managers must now manage a higher liability volatility than before. This is not, per the available data, a systemic liquidity crisis (the funds cited show solid balance sheets, contained leverage and performing portfolios), but rather the "growing pains" of an asset class that has developed strongly among a non-institutional clientele. In conclusion The HLEND episode, like those seen at Blackstone, Apollo or others, factually illustrates the mechanisms inherent to semi-liquid private-credit funds. Gating is not a malfunction but a feature designed to protect long-term value. For the retail investor, it nonetheless represents an important realisation: these products suit a long investment horizon, a measured allocation within a diversified portfolio, and require an attentive reading of the redemption rules, caps and liquidity risks described in the documentation. The attractive performance of private credit (yields, diversification) stays documented, as do the operational challenges tied to its democratisation. Investors and their advisers will gain from fully incorporating these gating parameters into their risk analysis and allocation construction. Sources and references 1. Reuters, "Investors asked to pull 13.3% from BlackRock private credit fund in first quarter", 12 June 2026: https://www.reuters.com/legal/transactional/investors-asked-pull-133-blackrock-private-credit-fund-first-quarter-2026-06-12/ 2. Reuters, "Apollo's president sees continued withdrawals from US private credit funds for the wealthy", 28 May 2026: https://www.reuters.com/legal/transactional/apollos-president-sees-continued-withdrawals-us-private-credit-funds-wealthy-2026-05-28/ 3. Reuters, "Private credit funds for wealthy individuals raise 45% less new money in Q1, RA Stanger says", 22 April 2026: https://www.reuters.com/legal/transactional/private-credit-funds-wealthy-individuals-raise-45-less-new-money-q1-ra-stanger-2026-04-22/ 4. HPS Corporate Lending Fund (HLEND), Q1 2026 Client Repurchase Letter to shareholders, details on the 13.3% of requests, 5% honoured (~$620M), performance, portfolio and liquidity (via hlend.com and associated filings). 5. Reuters, "Blackstone caps withdrawals from flagship private credit fund", 4 June 2026: https://www.reuters.com/business/blackstone-caps-withdrawals-flagship-private-credit-fund-2026-06-04/ 6. AdvisorHub / Bloomberg, "Trapped in Private Credit, Investors Wait to Pull Out $5 Billion", March 2026. 7. HLEND.com (performance page and SEC 10-Q / 8-K filings) for historical net-asset data and the fund's official communications. ============================================================================ ANALYSIS: Q2 bank earnings: reading the risk behind the profit URL: https://l0g.fr/en/analysis/q2-2026-bank-earnings-reading-the-risk/ Canonical French source: https://l0g.fr/posts/resultats-bancaires-t2-2026-lire-le-risque/ Date: 2026-07-14 (reviewed 2026-07-14) Topics: banks, earnings, risk, private credit, fed, markets ---------------------------------------------------------------------------- On Tuesday 14 July, before the open, five behemoths of US finance report their second-quarter accounts: JPMorgan, Bank of America, Citigroup, Wells Fargo and Goldman Sachs. The consensus expects good numbers, and it will probably be right. But beating the consensus is precisely what matters least, because it is already in the prices. For anyone who reads risk rather than the headline, the interest is elsewhere: in the interest margin, in the direction of provisions, and above all in a discreet thread linking these very profitable banks to the most fragile link of the system, private credit. Here is Tuesday's reading grid, the one that looks under the profit. The consensus, and why beating it matters little Let us start by setting the expectations, keeping in mind that these are estimates, not results. The consensus targets for JPMorgan a profit of about $5.49 per share on $48.7 billion of revenue, growth on the order of 10% year on year, and for Bank of America about $1.12 per share on $30.7 billion, up about 25%. Recent revisions are slightly higher for JPMorgan, Bank of America and Citigroup, a little less for Wells Fargo. The method lesson is simple. When the market already expects a 25% rise, beating it slightly surprises no one, and a merely in-line report can send the stock back. Earnings per share is a communication figure; it tells past performance, not the risk trajectory. For the latter, you have to open the hood. The interest margin, the real arbiter A retail bank's queen metric is not the profit, it is the net interest margin, the gap between what the bank earns on its loans and what it pays on its deposits. It is what says whether the model's heart is breathing. Two forces clash this quarter. On one side, banks' funding cost eased, from about 2.61% in 2024 to 2.26% in 2025, relieving the margin. On the other, the yield curve lost some slope in the second quarter, compressing the gain banks draw from maturity transformation. The context of durably high rates, under a Fed still leaning toward tightening, is double-edged for banks. It swells interest income on new loans, but it raises competition for deposits and, as we will see, deepens unrealised losses on bond portfolios. Listening, on Tuesday, to what the chief financial officers say about the margin trajectory for the rest of the year will be worth more than the profit of the past quarter. Credit, calm on the surface The second dial is the cost of risk, that is, the provisions banks set aside to cover doubtful loans. The snapshot is reassuring today: household and corporate defaults stay contained, US consumption holds up, and in the first quarter JPMorgan even released reserves, a sign of confidence in the quality of its credit. A provision release mechanically inflates the profit, inviting a distinction between real performance and the accounting effect. This is where one must beware the calm. A reserve release can reflect genuine solidity, or end-of-cycle complacency. The message bank executives have started to convey is nuanced: credit is fine today, but the foundations of the next stress are forming quietly, in commercial real estate, in private credit and in the layered structures linking banks to private equity. Banks' real risk, in 2026, does not read in their current losses; it reads in their exposures. The real signal: private-credit exposure Here is the most important thread, and the least commented in the headlines. The big banks are, for the most part, not directly exposed to the private credit that competes with them on corporate lending. They are exposed to it indirectly, through the massive loans they extend to private-credit funds, private-equity firms and real-estate platforms, those non-bank financial institutions that depend on bank funding to run. Goldman Sachs's risk, in particular, is increasingly defined by this exposure. The mechanism is exactly that of the silent contagion of private credit we described, and it reconnects the banking system to the shadow credit the regulator thought it had moved off balance sheets. Banks externalised direct credit risk toward private funds, but they re-internalised it through the door of financing those funds. On Tuesday, the line to track is not Goldman's profit, it is the size and growth of its exposure to non-bank actors, an item the most attentive investors now watch more than the trading result. It is the point where a good earnings season can mask a systemic risk piling up, a recurring theme of our coverage of shadow banking. The losses the balance sheet does not show, and commercial real estate Two pockets of risk complete the picture, both invisible in the profit. The first is unrealised losses. With rates staying high, the bond portfolios bought when rates were low are worth less than their cost, and this markdown weighs on capital only if it is recognised. It is the AOCI and held-to-maturity trap we dissected in our bank-health guide, and which was fatal to Silicon Valley Bank. The second is commercial real estate: the Fed's latest stress-test scenario incorporated a 39% fall in prices and about $75 billion of losses for the sector. These two pockets will not appear in Tuesday's headlines, but they condition the true solidity of balance sheets. A record profit backed by unrealised, unmaterialised losses and a fragile real-estate exposure is not the same thing as a record profit on a sound balance sheet. The distinction is the whole point of a risk-prism reading. Capital and the stress-test paradox On 24 June, the Fed published the results of its annual stress test, and they are reassuring on paper: the thirty-two largest banks would pass a severe-recession scenario while absorbing more than $708 billion of losses, their hard capital ratio giving up only 1.6 point, well above the minimums. The paradox is there. This pat on the back comes as the so-called "Basel III endgame" reform, re-proposed in 2026, lightens capital requirements, and as banks push to return capital to their shareholders through buybacks. In other words, the prudential corset is loosened at the very moment the most discreet risks, private credit and commercial real estate, are piling up. Several voices, from the Bank Policy Institute to sector analysts, warn that passing the test is not a green light to lower one's guard. The capital question, on Tuesday, will therefore be less "how much banks have" than "how much they intend to return", and at what risk. Tuesday's reading grid From all this emerges a grid, that of the bank-health guide applied live. Seven dials deserve attention, well beyond the displayed profit. At bottom Tuesday's results will probably be good, and this good news is the least interesting of all, because it is expected. The analyst's work begins where the press release ends: in the interest-margin trajectory, in the direction of provisions, in the size of the private-credit exposure, in the losses the balance sheet does not display. The season now opening will say whether US banks are solid or only profitable, two things the profit conflates and the risk distinguishes. The headline will announce the performance; we will read the plumbing. Sources 1. Consensus and calendar for Q2 2026 earnings (JPMorgan, Bank of America, Citigroup, Wells Fargo, Goldman Sachs, 14 July before the open): https://stocktwits.com/news-articles/markets/equity/top-wall-street-banks-kick-off-q2-earnings-next-week-here-s-what-analysts-expect/cZmrm9pR78o 2. IG, "US bank earnings preview: Q2 2026 in focus": profit expectations and the importance of net interest margin: https://www.ig.com/uk/news-and-trade-ideas/us-bank-earnings--what-to-expect-from-q2-2026-260707 3. Forbes (M. Rodriguez Valladares), "Wall Street's Big Banks Signal The Next Credit Risks": private credit, commercial real estate and exposure to non-bank actors: https://www.forbes.com/sites/mayrarodriguezvalladares/2026/04/15/wall-streets-big-banks-signal-the-next-credit-risks/ 4. Federal Reserve, 2026 stress-test results (24 June): $708 billion of absorbable losses, CET1 down 1.6 point: https://www.federalreserve.gov/publications/2026-stress-test-scenarios.htm 5. Bank Policy Institute, "The 2026 Federal Reserve Stress Test Results: A Framework in Transition": https://bpi.com/the-2026-federal-reserve-stress-test-results-a-framework-in-transition/ 6. FDIC, "Risk Review 2026": overview of banking-sector risks: https://www.fdic.gov/analysis/2026-risk-review-full.pdf 7. l0g, Bank-health guide, The silent contagion of private credit and CLOs and leveraged loans guide. ============================================================================ ANALYSIS: Sintra 2026: ECB hike, Warsh panel, tokenisation session URL: https://l0g.fr/en/analysis/sintra-2026-ecb-hike-warsh-tokenisation/ Canonical French source: https://l0g.fr/posts/sintra-2026-bce-tokenisation/ Date: 2026-07-14 (reviewed 2026-07-14) Topics: ecb, sintra, warsh, rates, inflation, tokenisation, cbdc, stablecoins ---------------------------------------------------------------------------- ECB Forum on Central Banking, Sintra, 29 June to 1 July 2026. Official theme: "Shaping Europe's future: innovation, growth and stability". This account sticks to facts and attributed remarks, sourced at the end of the article. The Sintra forum, hosted each year by the ECB since 2014, gathers central-bank governors, academics and market participants. Three strands of the 2026 edition are covered here: the rate trajectory, Kevin Warsh's first international panel as Fed chair, and the session on tokenisation. This piece connects our macro and central-banks coverage with our tracking of crypto and monetary infrastructure. The 11 June hike and the Eurosystem projections On 11 June, the ECB raised its three key rates by 25 basis points: deposit facility to 2.25%, main refinancing to 2.40%, marginal lending to 2.65%, effective 17 June. It is the first hike since September 2023; it closes a cycle of eight cuts begun in June 2024. The ECB attributes the decision to the inflationary pressures tied to the war in the Middle East. Headline HICP for May came in at 3.2% year on year, the highest since September 2023, the core at 2.5%. The Eurosystem projections published with the decision assume average inflation of 3.0% in 2026, 2.3% in 2027 and 2.0% in 2028, and growth of 0.8%, 1.2% then 1.5%. Euro-area GDP contracted 0.2% in the first quarter of 2026. Economists cited by Euronews raise a risk of stagflation. The debate over a second hike On the sidelines of the forum, Alexander Demarco, acting governor of the Central Bank of Malta and a member of the Governing Council, told Reuters that the ECB should not rush a further hike, given the retreat of oil toward its pre-conflict levels. In his view, the second-round effects that might justify tightening (drift in expectations, wage demands, indirect transmission) are not materialising, which argues for waiting for the next set of projections. He adds that the mildest scenario of the June projections already embedded a further turn of the screw, which would then remain possible. According to Reuters, markets see about a one-in-three chance of a hike in July and a fully priced move by October. The Warsh, Lagarde, Bailey, Macklem panel On 1 July, Kevin Warsh took part in his first international panel as Fed chair, Jerome Powell's term having ended this year, alongside Christine Lagarde, Andrew Bailey (Bank of England) and Tiff Macklem (Bank of Canada). According to CNBC's live coverage, Warsh said prices remain too high and that price stability remains the primary objective, while saying he is open to the possible disinflationary effects of AI. He asserted that the Fed's independence would not change, whatever the pressure exerted by Donald Trump. Warsh embraces a communication style at odds with the Powell era: fewer public remarks, no forward guidance. CNBC reports, citing Bank of America, twelve interventions by Fed officials since the June meeting, against an average of about twenty-three over the same window since 2022. The frame of his doctrine had been set at his first FOMC and his battle over the balance sheet; on the US side, inflation is tracked in the return of US inflation. The tokenisation session: Project Hangang The session on tokenisation, chaired by Piero Cipollone (ECB), rested on a paper signed by a Bank of Korea team (Jaemin Ryu, Hyun Song Shin, Joonyi Sung, Sung Guan Yun) devoted to Project Hangang. According to this paper, Hangang is an implementation of the unified ledger formalised by the BIS in 2023: a single programmable platform where tokenised central-bank money (wCBDC), tokenised commercial-bank deposits and tokenised assets, government bonds for example, coexist. The test gathered about 80,000 users and 12,000 selected merchants, from April to June 2025. The figure of "7 banks representing 80% of banking assets", present in some summaries, does not appear in the paper and is not repeated here. The paper describes two design choices. First, consensus: Hangang runs on proof of authority, the central bank validating transactions, and not proof of work or proof of stake. The paper explains that a decentralised ledger with strict consensus must compensate its validators for the risk of non-coordination, a cost funded by fees and therefore by congestion, which pushes toward fragmentation into competing chains; the central bank, for its part, does not bear that cost. Second, the "burn-and-issue" mechanism: in an interbank transfer, a smart contract executes in a single transaction the destruction of the tokenised deposits at the issuing bank, the transfer of wCBDC between the two banks, then their re-issuance at the receiving bank; if a step fails, everything is cancelled. According to the paper, this mechanism preserves the singleness of money: a deposit remains the liability of a single bank and the payment goes at par. The paper contrasts this model with the stablecoin, where the token is transferred as is and the holder bears a claim on the issuer, whose value can deviate from par depending on the issuer's solvency and risk appetite. The case of dollar stablecoins is treated in our work on the sovereign tolling of the Strait of Hormuz in USDT on Tron. On the separation between the monetary layer, which carries value, and the programming layer, which carries the conditions of use, the paper describes the principle but does not specify, in the sections consulted, the token standards (ERC-20, ERC-1155, ERC-3525) sometimes attributed to the project; they are therefore not asserted here. The European initiatives: Pontes and Appia The paper situates Hangang against two ECB work streams: Pontes, which links the tokenised to the existing real-time gross settlement (RTGS) infrastructure, and Appia, which aims to deliver by 2028 the blueprint for an integrated wholesale-settlement ecosystem, anchored in tokenised central-bank money. In its 11 June statement, the Governing Council indicates that the digital euro and tokenised wholesale central-bank money will strengthen Europe's "strategic autonomy". On the mechanics of settlement and collateral, see the liquidity factory of repo; on central-bank reserves, de-dollarisation through gold. Other sessions: AI, productivity, migration The session on financial stability and AI was chaired by Isabel Schnabel, with the participation of Sarah Breeden (Bank of England) and Torsten Slok (Apollo). According to session dispatches, Breeden identified a jump in the cyber capabilities of agentic AI as her main financial-stability concern. The academic work included Bart Van Ark's paper on the alignment of investment, innovation and diffusion, and Giovanni Peri's on immigration as a productivity factor in OECD countries. On the closing panel, according to CNBC, Lagarde said Europe is behind on AI and on frontier-technology firms, and that Europe and the United States are "hostage to each other", Europe representing, in her view, about 25% of many hyperscalers' revenue. Sources - European Central Bank, monetary policy decision of 11 June 2026 (rates, Eurosystem projections), - European Central Bank, monetary policy statement and press conference, 11 June 2026 ("strategic autonomy", digital euro, wCBDC), - European Central Bank, ECB Forum on Central Banking 2026, programme and speakers, - Jaemin Ryu, Hyun Song Shin, Joonyi Sung, Sung Guan Yun, "A unified ledger in practice: lessons from Project Hangang", Bank of Korea, ECB Forum 2026 (80,000 users, 12,000 merchants, burn-and-issue, proof of authority, Pontes and Appia), - Reuters, "ECB should not rush any further rate hike, Demarco says", Sintra, 1 July 2026, - CNBC, ECB Forum live updates: Fed Chair Warsh speaks (prices too high, Fed independence, Bank of America comparison, Lagarde on AI and hyperscalers), 1 July 2026, - Euronews, "Christine Lagarde's Sintra speech signals a new ECB playbook", 30 June 2026, - Newsquawk, programme and schedule of the Sintra 2026 forum (sessions, chairs, speakers), - ECB Forum on Central Banking 2026, Day 1 (Sarah Breeden's statement on agentic AI), - Bart Van Ark, "Rewiring Europe's productivity framework", ECB Forum 2026, - Giovanni Peri, "The immigration impact on population, labor productivity and growth", ECB Forum 2026, ============================================================================ ANALYSIS: Ethereum and TradFi: the real-world test of programmable finance URL: https://l0g.fr/en/analysis/ethereum-tradfi-infrastructure/ Canonical French source: https://l0g.fr/posts/ethereum-tradfi-infrastructure-finance/ Date: 2026-07-14 (reviewed 2026-07-14) Topics: crypto, ethereum, tradfi, tokenisation, stablecoins, rwa, infrastructure ---------------------------------------------------------------------------- import InfographicBars from '../../components/infographic/InfographicBars.astro'; We must start by discarding two too-convenient ideas. Ethereum is not going to replace, on its own, banks, clearing houses, SWIFT, central securities depositories or central banks. But Ethereum is also not a mere on-chain casino separate from traditional finance. Since 2024, the border has moved: spot ether ETFs have installed the asset in regulated portfolios, stablecoins give a cash leg to on-chain activity, tokenised money market funds like BUIDL put public securities into smart contracts, and the big institutions look closely at the same technical promise that the BIS formalises in its work on tokenisation: bringing the asset, the settlement money and the execution rules together in a single programmable layer. This article's thesis is measured: Ethereum matters because it serves as a public test bench for programmable finance. It shows, in production, how assets can be issued, transferred, composed, verified and settled around the clock on a common infrastructure. It does not yet resolve the hardest institutional questions: confidentiality, compliance, legal recourse, governance, smart-contract risk, rollup sequencers, bridges, quality of stablecoin reserves. But it makes visible mechanisms that often stayed buried in private databases and reconciliation chains. To understand Ethereum, one must separate five layers. Ethereum is the network and the settlement layer. ETH is the native asset that pays for gas and secures consensus. The EVM is the virtual machine that executes contracts. Rollups move execution to layer 2s by publishing proofs or commitments to Ethereum. Financial applications use this stack for stablecoins, DEXs, lending, tokenised funds, collateral and payments. Confusing these layers leads to bad conclusions: a rise in stablecoins is not proof of monetary superiority, a fall in L2 fees is not a guarantee of decentralisation, and a tokenised fund is not yet a complete institutional market. The useful technical entries are added to the l0g glossary: Ethereum, EVM, rollup, L2, blob, EIP-4844, data availability, sequencer, account abstraction, PBS and MEV. 1. Why Ethereum interests TradFi Traditional finance already has ledgers. It knows how to settle securities, manage collateral, centralise risk, run clearing houses and apply compliance rules. The problem is not the absence of infrastructure. The problem is fragmentation. Payments, securities, messages, confirmations, margin calls and ownership registers still circulate between silos. The BIS describes this architecture as a series of separate databases, messaging and reconciliation before final settlement. Ethereum offers a public version of a different intuition: an asset can carry its own transfer rules and interact with other assets in the same execution environment. In a traditional operation, the message and the settlement are not the same thing. An instruction can be sent, confirmed, cleared, then settled later in another infrastructure. In Ethereum, the model is more brutal: a valid transaction modifies the ledger's state. This simplicity hides much technical complexity, but it gives a central idea: the ledger, execution and proof can be brought together. It is exactly the direction explored by institutional tokenisation projects, even when they do not choose a public chain. The BIS Innovation Hub does not build Project Agorá on Ethereum. It tests a programmable platform for wholesale cross-border payments with tokenised central-bank reserves and tokenised bank deposits. But the vocabulary is the same: atomic settlement, around-the-clock availability, embedded compliance logic, reduced reconciliation. Ethereum matters because it has exposed these properties in a public, adversarial, liquid environment, observable by all. TradFi can reject total permissionlessness and still take up part of the technical grammar. The second reason is liquidity. A financial infrastructure becomes useful when assets, counterparties, developers and tools gather there. On 8 July 2026, DefiLlama shows about $309.1 billion of stablecoins, of which about $151.9 billion on Ethereum, or 49.15% of the total. Circle reports $73.0 billion of USDC in circulation on 6 July 2026 and native issuance on 35 networks. RWA.xyz shows $30.87 billion of tokenised assets distributed and $398.59 billion of assets represented, with Ethereum at $16.4 billion of tokenised value and 52.80% market share in its by-chain table. These figures do not say that global finance has migrated on-chain. They say that the layer has enough liquidity to be tested by non-crypto-native actors. BlackRock's tokenised fund BUIDL appears at $2.4 billion on RWA.xyz at the time of consultation. Reuters recalled that in March 2024 BlackRock had launched BUIDL on Ethereum, invested in cash, US Treasury bills and repo operations, with a market cap of $530 million in October 2024 in the article cited below. The trajectory is more interesting than the isolated figure: institutional tokenisation begins with the simplest, most liquid assets, the closest to cash. The third reason is access. Spot ether ETFs brought ETH into the world of listed products. The Financial Times reported that the first US spot ether ETFs, including BlackRock and Fidelity products and the conversion of the Grayscale trust, began trading in July 2024. This does not turn Ethereum into banking infrastructure. It creates a regulated interface between traditional markets and the network's native asset. For an institution, it becomes possible to distinguish three exposures: the price of ETH, the flows of on-chain applications, and the infrastructure value of the layer. The fourth reason is more discreet: operational experimentation. Visa told Reuters in January 2026 that its annualised stablecoin settlement volume reached $4.5 billion, a tiny fraction of its $14.2 trillion of annual volume, but growing according to its crypto chief. The important sentence is not that Visa would be replaced. It says the opposite: existing networks seek to connect stablecoins to their own merchant network. TradFi does not adopt Ethereum as a religion, it evaluates it as a new settlement and distribution layer. 2. The minimal pedagogy: Ethereum in ten blocks 1. Ethereum is a shared computer, but the image is imperfect. Each node verifies the network's state. Smart contracts are programs deployed on the chain. The useful metaphor is rather that of a programmable ledger: the balance, the rule and the proof live in the same logical space. That is why a stablecoin, a fund token, a lending protocol and a DEX can interact without classic bilateral integration. 2. ETH is not just a ticker. ETH pays for gas, remunerates validators, can be staked and serves as an economic-security asset. The official staking page shows about 40.4 million ETH staked, or about 32% of the supply, with an APR displayed at 2.6% at the time of consultation. This yield is not a risk-free coupon: it depends on the protocol, slashing, the operator, liquidity and the price of ETH. 3. Ethereum moved to proof of stake. The Merge of 15 September 2022 replaced proof of work with proof of stake. Ethereum.org estimates this transition cut energy consumption by about 99.95%. Its energy page gives a CCRI estimate of 0.0026 TWh per year for post-Merge Ethereum, against 21 TWh for proof-of-work Ethereum before the transition. The environmental criticism has not disappeared for all of crypto, but it no longer reads the same way for Ethereum. 4. The L1 is expensive by design. Ethereum L1 maximises security and verifiability rather than raw user throughput. That is the reason for the rollup-centric strategy: everyday execution migrates to L2s, while Ethereum keeps the role of settlement, data availability and finality. 5. Rollups are not ordinary sidechains. A rollup executes transactions off L1, groups the results, then publishes to Ethereum the necessary data and commitments. Ethereum.org describes rollups as the central scaling method: they group transactions and send the results to Ethereum. The stake is that user cost falls without breaking the security link with the L1. 6. Blobs are a data corridor for rollups. EIP-4844 introduced blob transactions, packets of data not accessible to EVM execution but available to verify rollups. The EIP specifies a target of about 0.375 MB per block and a limit of about 0.75 MB in this stopgap toward full sharding. Blobs cost less than calldata because they are not kept indefinitely by execution. 7. Composability is the real conceptual leap. A tokenised bond, a stablecoin, a liquidity pool and a lending contract can call one another. In traditional finance, this interoperability requires conventions, contracts, files, counterparties and integrations. On Ethereum, it can be native, but at the cost of a more direct software and financial contagion risk. 8. Accounts remain difficult. A bank does not want to lose an asset because a private key was copied. Neither does an individual. ERC-4337 and EIP-7702 seek to improve the account experience: programmable validation, bundlers, paymasters, temporary code delegation for external accounts. This layer is indispensable if Ethereum wants to move beyond the audience already familiar with wallets. 9. L2 decentralisation remains unfinished. Ethereum.org notes that current rollups still use centralised components, notably sequencers and provers. L2BEAT ranks several large L2s with maturity stages and mentions of additional trust assumptions. The message is simple: L2s are useful, but you must read their own risks, not just their marketing link to Ethereum. 10. Ethereum is not central-bank money. The BIS insists on the singleness of money: forms of money must exchange at par, anchored by central-bank money. Private stablecoins can serve as an on-chain cash leg, but they are not central-bank reserves. This difference is decisive for wholesale markets, banks and systemic infrastructure. 3. The numbers that really matter The trap with Ethereum is to choose the figure that confirms your side. Optimists cite TVL, DEX volumes, stablecoins and ETFs. Sceptics cite hacks, speculation, historical fees, sequencer concentration and dependence on stablecoins. Both readings must be kept. On 8 July 2026, DefiLlama Chains shows Ethereum with about $38.9 billion of DeFi TVL, 1,909 protocols, about $152.5 billion of stablecoin market cap and about $967.9 million of 24-hour DEX volume. The same page places Solana, BSC, Tron and Base far behind Ethereum in DeFi TVL, but with different profiles: Solana has high DEX volume, Tron concentrates stablecoins, Base plays the role of a mainstream rollup. Ethereum is not alone, but it remains the centre of gravity for complex programmable finance. TVL must also be read with caution. A 2025 academic article on TVL verifiability recalls that aggregators rely on heterogeneous methodologies and sometimes community contributions. Another paper on TVL and double-counting stresses that wrapping, leverage and derivatives can inflate the apparent amounts. On l0g, TVL is therefore a thermometer, not an audited balance sheet. On rollups, L2BEAT shows 22 rollups, 6 validiums and optimiums, and a long tail of other projects. Arbitrum One appears at $17.39 billion of secured value, Base at $11.25 billion, OP Mainnet at $1.41 billion at the time of consultation. The lesson is not only scale. L2BEAT also flags stages and trust assumptions: the L2 ecosystem is already financial, but still in a hardening phase. On Ethereum itself, the Ethereum for Institutions site presents the network as an institutional liquidity layer and shows, at the time of consultation, about 10.9 years of uptime, 40.6 million ETH securing the network, $70.5 billion of economic value associated with staking, $161 billion of stablecoin TVL, $39 billion of DeFi TVL and $2.22 billion of average 24-hour DEX volume. This source is an ecosystem showcase, so read it as such. It remains useful to understand the institutional argument carried by the Ethereum Foundation. The most underestimated figure is perhaps not TVL, but the stablecoin share. Stablecoins are the cash leg of the on-chain economy. Without them, tokenised RWAs remain hard to settle, DEXs are limited to crypto against crypto, and international payments lose their reference asset. With them, Ethereum hosts a private quasi-dollar, programmable, mobile and liquid. That is useful, but also fragile: the trust depends on the issuer, the reserve, the legal framework, redemptions and compliance. 4. The bridge with TradFi: stablecoins, RWAs, ETFs, tokenised deposits The first bridge is the stablecoin. A well-designed stablecoin turns a claim on an issuer into a transferable token. Circle presents USDC as redeemable 1:1 and backed by cash or cash equivalents, with monthly attestations. For a bank or a manager, it is not final money. It is a very liquid private claim, usable as a settlement instrument in on-chain applications. That is the whole difference: the stablecoin gives an operational cash leg, but not the ultimate quality of central-bank money. The second bridge is the tokenisation of simple financial assets. Tokenised money market funds are the cleanest case: short duration, liquid assets, simple accounting, collateral demand. Reuters wrote that tokenised Treasuries formed a growing segment, mostly on Ethereum, and recalled that BUIDL invested 100% in cash, Treasury bills and repo. WSJ reported in March 2024 that BlackRock had launched its first tokenised fund on a public blockchain with Securitize. The third bridge is the ETF. A spot ether ETF does not put market finance on Ethereum; it puts ETH in a market wrapper. This difference is fundamental. The ETF makes the asset accessible to allocators who do not want to manage keys, wallets or smart contracts. It does not circulate securities on Ethereum. It brings ETH's price risk into securities accounts. To understand the infrastructural impact, one must therefore separate ETFs on ETH from asset tokenisation on Ethereum. The fourth bridge is the tokenised deposit. This is where the debate becomes systemic. The BIS clearly distinguishes private stablecoins from tokenised deposits and wholesale CBDCs. In a tokenised-deposit model, the client holds a programmable bank claim, not a stablecoin issued outside the banking system. Project Agorá tests precisely this kind of logic with central banks and private institutions, but in a controlled wholesale environment. Ethereum has demonstrated public programmability. Wholesale finance wants programmability with safe settlement money, confidentiality and native compliance. The fifth bridge is collateral management. A tokenised asset can circulate faster, be used in lending contracts, be immobilised as collateral, be settled against stablecoin or against tokenised deposit. It is also the most dangerous zone. If the same asset is reused, framed by several contracts, valued by oracle, locked by a bridge then financed elsewhere, the risk becomes legible but also faster. Composability reduces friction and accelerates cascades. Pedagogical example: a tokenised money-market-fund share Take a deliberately simple case. A manager issues a tokenised money-market-fund share. The fund's portfolio holds cash, Treasury bills and repo. The token represents an economic share of the fund. A qualified investor buys this share against traditional dollars or against stablecoin, depending on the issuer's set-up. Once the token is received in their institutional wallet, they can keep it, transfer it to an authorised counterparty, use it as collateral in a compatible contract, or redeem it from the issuer per the fund's rules. In a classic infrastructure, several ledgers intervene: the fund's transfer agent, the custodian bank, the cash account, the messaging, the investor controls, the confirmations, then the internal entries. In a tokenised version, part of this logic can condense: the token carries the ownership identifier, the contract encodes certain transfer restrictions, the on-chain history gives an audit trail, and the cash leg can be a stablecoin if the framework allows. The potential gain comes from this condensation, not from the word blockchain in itself. The important point is synchronisation. If a fund share and a stablecoin live in the same execution environment, a transaction can swap the two legs together. It is the intuition of atomic settlement: everything happens or nothing happens. In traditional finance, delivery versus payment also exists, but it depends on specific market infrastructures. On Ethereum, a contract can reproduce a form of programmable DvP, provided both assets are valid, liquid and legally recognised. The limits appear immediately. The token only has value if the issuer honours the redemption, if the underlying asset exists, if custody is correct, if the law recognises the ledger, if the list of authorised investors is maintained, if the eventual oracle is reliable, and if the contract cannot be arbitrarily updated. The blockchain can make the transfer clean. It does not replace due diligence on the issuer. This example explains why tokenised money market funds are more credible than overly broad promises of real-estate or private-credit tokenisation. A Treasury bill is liquid, standardised, short, closely followed and valuable. A building, a private loan or a disputed claim requires law, expertise, recovery and event management. The simpler the asset, the more the on-chain part can create value. The more complex the asset, the more the off-chain returns to the centre. 5. Why Ethereum, and not just a bank database A bank database can be faster, cheaper and more compliant than Ethereum. A private blockchain can be simpler to govern. A central-bank unified ledger can better preserve settlement money. The question is therefore not: why not rebuild Ethereum in-house? The question is: what properties only a public, open and liquid infrastructure brings? First property: common access. An application can deploy a contract, issue an asset, read a stablecoin, integrate a DEX, plug in an oracle, use a wallet and access existing liquidity without signing a bilateral integration with each actor. In traditional finance, integration is a fixed cost. On Ethereum, integration is often a public software interface. This does not remove the law, but it changes the speed of innovation. Second property: auditability. Balances, contracts and transactions are observable. The limits are real: the address is not always the actor, flows can be internalised on exchanges, and institutional confidentiality is insufficient. But for a risk analyst, Ethereum produces raw data that private infrastructures do not always publish. That is the interest of a public ledger: it also reveals the fragilities. Third property: relative neutrality. Ethereum does not require a single actor to authorise the execution of each application. This point attracts developers and frightens regulators. It explains the fast innovation and the difficulty of control. For TradFi, public neutrality is useful as a liquidity layer and a market standard, but problematic when sanctions, KYC, confidentiality and judicial freezes must be applied. Fourth property: programmable settlement. A contract can impose a transfer rule, verify a condition, distribute a yield, lock collateral, liquidate a position or synchronise several legs of a transaction. The BIS speaks of programmability as a way to create arrangements not practicable with current rails. Ethereum is one of the production proofs of this idea, even if the BIS prefers to anchor it in central-bank money for wholesale finance. Fifth property: the tooling network. Developers, audits, indexers, wallets, custodians, rollups, ERC standards, stablecoins, DEXs, oracles, block explorers and analytics constitute an invisible capital. A bank can create a private blockchain, but it does not instantly recreate this ecosystem. It is the same logic as for the Internet: value comes not only from the protocol, but from the accumulated layer of tools and standards. 6. The roadmap: read Ethereum as a succession of constraints The Ethereum roadmap is not a classic product plan. Ethereum.org specifies that it is built by a development community, that elements may change, and that long horizons are intent rather than hard commitment. It must therefore be read as a map of constraints: security, scalability, decentralisation, user experience, censorship resistance, node lightness. The Merge resolved the energy constraint and changed the security model. Shapella made staking withdrawals possible, turning staking into a more institutional activity. Dencun, with EIP-4844, created blobs and cut the cost of data for rollups. Pectra brought EIP-7702 to give external accounts capabilities close to smart wallets, raised certain validator-related parameters and increased blob throughput. Fusaka, activated on 3 December 2025 per the roadmap, introduces PeerDAS, blob adjustments and safer gas limits. Glamsterdam, expected in the second half of 2026 in the roadmap, notably aims at enshrined proposer-builder separation and block-level access lists. The logic is not that of a blockchain simply increasing block size. Ethereum seeks to preserve the ability to verify the network while moving execution to L2s. This strategy creates a trade-off: the L1 stays robust and relatively expensive; the L2s become the user interface; the data needed for the L2s must stay available long enough to verify their state; nodes must not become out-of-reach data-center machines. The key word is data availability. A rollup can publish a result, but participants must be able to reconstruct or contest the state if needed. If the data is not available, security becomes theoretical. EIP-4844 created a cheaper path for temporary data. PeerDAS goes further: instead of each node downloading everything, validators can sample the data. The goal is to increase blob throughput without exploding hardware requirements. The second key word is PBS. In a network with MEV, specialised actors know how to build more profitable blocks than ordinary validators. Proposer-builder separation acknowledges this reality: builders build, proposers choose. Ethereum.org presents PBS as a censorship-resistance improvement, especially with mechanisms like inclusion lists or encrypted transactions. The institutional stake is clear: a market cannot depend on an opaque and concentrated block construction without safeguards. The third key word is finality. Ethereum already has economic finality, but Ethereum.org explains that single slot finality would reduce the current window for chain reorganisation, around 15 minutes, toward finality in a single slot. For TradFi, finality is a familiar language: when a settlement is final, it must no longer be cancelled by a simple state change. SSF is therefore more than a technical optimisation; it is a translation of the settlement vocabulary into a public protocol. The fourth key word is user experience. Institutions do not want their traders handling seed phrases like personal passwords. Neither do individuals. ERC-4337 introduces a form of account abstraction through UserOperations and bundlers, without a consensus change. EIP-7702 lets an external account temporarily delegate code, to improve wallets. In plain terms: transaction batching, recovery, permissions, gas payment by a third party, signature policies and automation become easier. 7. Effects of the roadmap for finance For a DeFi protocol, a fall in L2 cost means more users and more operations. For a TradFi actor, it means something else: the possibility of moving low-margin flows to programmable rails. A stablecoin transfer costing several dollars in fees stays a niche product. A transfer [...] ============================================================================ ANALYSIS: Crypto and AI: anatomy of a convergence, between real plumbing and narrative casino URL: https://l0g.fr/en/analysis/crypto-and-ai-convergence/ Canonical French source: https://l0g.fr/posts/crypto-et-ia-anatomie-d-une-convergence/ Date: 2026-07-14 (reviewed 2026-07-14) Topics: crypto, ai, stablecoins, regulation, markets, tech ---------------------------------------------------------------------------- Crypto and artificial intelligence are the two narratives that have magnetised capital and attention for five years. Seeing them merge was inevitable, and the buzzword "crypto-AI" has become a magnet for funding as for scams. For the analyst, the difficulty is not to note the convergence, it is to sort it. Because under the same label coexist a plumbing that already works, software agents settling invoices in stablecoins, and a casino of substance-less tokens that wiped out billions in 2025. This piece maps the intersection, vector by vector, separates what rests on a real solved problem from what is only a ticker, examines the players and the law, and poses the question that will decide everything: does traditional finance follow, and how far. Two worlds that everything opposes, and that attract each other At first sight, nothing brings the two technologies together. Artificial intelligence is centralising: it demands colossal capital, giant data centers and rare chips, favouring a handful of players and concentrating power. Crypto was born of the opposite ambition, to decentralise trust and remove intermediaries' toll-booth position. One builds black boxes; the other builds transparent, verifiable ledgers. Their meeting is therefore less about kinship than about the complementarity of two lacks. Three needs of AI find a crypto answer. Autonomous software agents need a programmable, permissionless means of payment, which banking rails cannot offer. Generative AI creates a crisis of trust, deepfakes and synthetic content, to which cryptographic verification brings the beginning of an answer. And AI's hunger for compute seeks sources of supply alternative to the hyperscalers. In reverse, crypto, long in search of uses beyond speculation, finds in AI a real demand. It is this double gap that makes the attraction. But one must still distinguish the vectors where it produces something from those where it produces only a price. The most lucid framework remains the one laid down as early as 2024 by Vitalik Buterin, co-founder of Ethereum, who distinguishes four uses: AI as a player in a crypto mechanism (bots that arbitrage or feed prediction markets), AI as an interface (an assistant that protects the user from scams), AI as a rule (an AI arbitrating inside a smart contract) and crypto to build a more trustworthy AI. His warning deserves to open any serious analysis: entrusting an AI with the role of rule, for example to back a stablecoin, is the most dangerous, because an exposed model gets attacked through adversarial learning. The loudest convergence is often the most fragile. The most solid vector: agents that pay The first use that left the PowerPoint slide for production is the payment of autonomous agents. An AI agent that buys data, compute or an interface call on its own needs to pay continuously, in very small amounts, without opening an account or signing a subscription. Yet card rails cannot do this: per the Keyrock report, 76% of agent payments fall below the cards' $0.30 fixed-fee threshold, most between one and ten cents. This is the concrete economic problem crypto solves: programmable, permanent, machine-to-machine micropayments, settled in stablecoins. The infrastructure came together at a notable speed. Amazon Web Services launched Bedrock AgentCore Payments, which lets agents pay in USDC, built on Coinbase's x402 protocol and Stripe's Privy wallets. Coinbase published an MCP connector for its Base chain, Visa and Mastercard are preparing their agentic-commerce platforms. The dominant stablecoin there is Circle's USDC, refocusing the dependence on a single issuer. This mechanism directly links AI to the money framed by the GENIUS Act. The reading guardrail is essential here. The $73 million settled by agents in one year is a fact, but a tiny fact against the $14.5 trillion Visa processes each year. Gartner's projections, $15 trillion intermediated by agents by 2028, or McKinsey's, $3 to $5 trillion of agentic commerce by 2030, are adoption bets, not observations. The use case is genuine; its future scale remains to be demonstrated. Decentralised compute: arbitraging a commodity The second vector attacks the sinews of the AI war, compute. The thesis of DePIN networks is simple: Nvidia graphics cards are a commodity, the hyperscalers' margins are therefore vulnerable, and a network that pays owners of idle GPUs in tokens can rent compute far cheaper. The figures give body to the argument. Per Q1 2026 market data, io.net aggregates more than 100,000 GPUs and Akash offers H100 cards around $1.20 to $1.80 an hour, against $4.50 to $5.50 at AWS, a 60 to 70% discount. The sector, around $180 to $220 million of annualised revenue, stays modest but is growing. The weakness is structural and comes down to the token's reflexivity. GPU providers are paid partly in the network's native token; when this token falls below their profitability threshold, they unplug, and the compute supply evaporates at the worst moment. Akash's available capacity thus contracted by more than half from one quarter to the next in early 2026. Add service guarantees inferior to the giants' and a hard technical limit: these networks excel at massively parallel tasks like rendering or inference, but remain unable to train large models, which demand the tight interconnections only integrated data centers provide. Decentralised compute nibbles a commodity through price; it does not replace the factory. Verifying the machine: zkML, provenance, proof of humanity The third vector is the most intellectually promising, and the earliest. It answers the question every "black box" AI poses: how to prove a model did what it claims, without revealing either the model or the data. zkML, zero-knowledge machine learning, uses cryptographic proofs to certify an inference. An academic survey published in 2025 catalogues the work since 2017 and confirms the potential as much as the obstacles: high proof cost, limited circuit expressiveness, deployment complexity. Tools like EZKL make the thing possible on real models, but industrial scale is not there. The authenticity side is more advanced, provided one is precise. The C2PA standard, carried since 2021 by Adobe, the BBC, Microsoft and others, affixes a cryptographic signature on content to trace its origin and edits. It is cryptographic, but not necessarily backed by a blockchain, and one must be careful not to file it automatically under the crypto label. The properly crypto part is elsewhere: in attestation on a neutral ledger and above all in proof of humanity, which aims to distinguish a real human from a bot. The stake swells with deepfakes, whose recorded number reportedly went from about 500,000 cases in 2023 to more than 8 million in 2025. Devices like World ID combine biometrics and selective disclosure to prove one is a unique human without revealing one's identity. It is the vector that best solves, in theory, the problem Vitalik judged the thorniest, that of trust in an opaque system; it is also the one whose realisations remain the most embryonic. Why it can work Three fundamental reasons support the bullish thesis, once the speculative froth is set aside. First, the identified complementarities answer real problems unsolved elsewhere: paying an agent below the cost of a card transaction, buying commoditised compute cheaper, verifying an AI output without trusting its author. None of these problems has an obvious better non-crypto solution. Second, the US regulatory wind has turned: the GENIUS Act legalised payment stablecoins, the base of agent payments, and the CLARITY Act began to clarify the status of digital assets, subjects we treated in our pieces on the CLARITY Act and the GENIUS Act. Finally, and this is decisive, traditional finance has begun to lay the rails, giving the convergence an adult infrastructure rather than a start-up patchwork. When an agent settles in stablecoin via an AWS-Coinbase-Stripe stack, it does not use a marginal gadget, it borrows a plumbing that listed companies operate at scale. The convergence then ceases to be a technological bet to become a question of adoption, a wholly different risk, and a more bankable one. Why it can fail The opposite reading is at least as well supplied, and recent history gives it ammunition. The first ground for failure is that most of the "crypto-AI" of 2024-2025 was not technology but narrative sold as such. The AI-agent-token bubble is the proof by absurdity. The ai16z token, which fraudulently borrowed the name of the a16z fund, peaked around $2.6 billion of market cap in early 2025 before collapsing nearly 99.9%, and its creators face a class action for fraud, accused of having promoted a non-existent AI. The Virtuals token lost nearly half its value in one week. The rule that emerges is cruel and instructive: the rare tokens that held had a genuinely functional agent generating on-chain activity; the "narrative first" projects lost more than 90% in a few months. The other grounds for failure are structural. AI is intrinsically centralising and capital-hungry, against crypto's decentralising ideal: training a large model will always favour whoever owns the factory, not the network of amateurs. Compute networks remain hostage to their token's price. zkML is not mature. And Vitalik's warning stands: as soon as you entrust an AI with a rule role over significant sums, you open an attack surface through adversarial learning that is hard to close. Finally, a good share of projects are solutions in search of a problem: most AI uses have no need of a blockchain, and the marriage is sometimes only a fundraising artifice. The regulatory framework, still hazy The law runs behind the technology, and the gap is gaping on the newest point: the liability of an autonomous agent that transacts. No regulator has yet published a clear doctrine on the application of KYC and anti-money-laundering when an AI agent executes a cross-border payment on its own. Liability regimes presuppose human intent and direct causality, two notions that wobble when the decision is taken by a machine. The emerging doctrine shifts the burden onto the deployer, the company that integrates the model into a product, which cannot pass the buck to its model provider. A concept is rising, "Know Your Agent", the 2026 equivalent of KYC: cryptographically verify an agent's identity before any transaction, looping back onto crypto's identity-attestation uses. The rest of the framework comes together in blocks. In the United States, the GENIUS Act frames agents' money and the CLARITY Act their asset environment. In Europe, the MiCA regulation already governs crypto assets, while the AI regulation likely classifies autonomous agents making financial decisions among high-risk systems, with obligations of human oversight and auditability, whose heaviest requirements apply from 2 August 2026. Several recent academic works, from the International Monetary Fund on agentic payments to legal analyses of the autonomous agent under European law, map this vacuum in the process of being filled. The rule, here, is behind the machine, and this lag is itself a risk factor. TradFi follows, but not everywhere The question that will decide is that of adoption by established finance, and the answer is nuanced: it follows the plumbing, not the casino. On the infrastructure side, the commitment is massive and documented. Visa, Mastercard, BlackRock, JPMorgan, Fidelity, State Street, Stripe and more than one hundred and forty Fortune 500 companies are deploying stablecoin rails and tokenised products on a shared layer whose market cap has passed $322 billion. On 30 June 2026, a joint stablecoin initiative, Open USD, brought together Stripe, Visa, Mastercard, Coinbase, BlackRock, BNY, Standard Chartered, Google and Shopify among others. Stripe bought the Bridge infrastructure and distributes wallets to agents; we described this shift from crypto to infrastructure in our pieces on Ethereum and TradFi and on Hyperliquid. But this adoption is selective, and that is the whole point. Traditional finance absorbs stablecoins, tokenisation and agent-payment rails, because they cut costs and open markets. It keeps its distance from the speculative layer, the agent tokens and the decentralised narratives, whose collapse it witnessed. Telling sign, Visa is building technology allowing banks to turn their deposits into programmable money while keeping the funds on their balance sheet: TradFi wants the stablecoin's advantages without ceding its intermediation. It adopts the function, not the ideology. It is the best proof that the convergence has a real substance, and the best reason to doubt it will keep all its rupture promises. The dividing line The right way to read crypto and AI is neither the prospectus's enthusiasm nor the sceptic's contempt, it is the sorting. On one side, a plumbing that solves real problems: agents paying below the cost of a card, a compute market arbitraging a commodity, a verification starting to pierce the black box. On the other, a casino that has already burned its players, and a share of projects where the blockchain is only a wrapper to raise funds. The dividing line is clear once you look for it: the convergence works when crypto solves a coordination, money or trust problem that AI genuinely poses, and it fails when it is only a ticker stuck onto a model. The test, for the analyst as for the investor, holds in one simple question, the same that separated the survivors from the dead in the 2025 bubble: is there, behind the token or the protocol, a working agent, a running GPU, a verifiable proof? If so, the convergence has a future. If not, it is a buzzword, and buzzwords always end up emptying. Sources 1. Vitalik Buterin, "The promise and challenges of crypto + AI applications", 30 January 2024: the four-use framework and the warning on AI as a rule: https://vitalik.eth.limo/general/2024/01/30/cryptoai.html 2. Amazon Web Services, "Agents that transact: Introducing Amazon Bedrock AgentCore payments, built with Coinbase and Stripe": agent payments in USDC via x402 and Privy: https://aws.amazon.com/blogs/machine-learning/agents-that-transact-introducing-amazon-bedrock-agentcore-payments-built-with-coinbase-and-stripe/ 3. CoinDesk / Keyrock report, crypto rails as the agent payment layer: ~$73M across 176M transactions, 76% of payments below $0.30: https://www.coindesk.com/business/2026/05/21/crypto-rails-are-becoming-the-default-payment-layer-for-ai-agents-report-says 4. International Monetary Fund, "How Agentic AI Will Reshape Payments", IMF Notes 2026/004: https://www.elibrary.imf.org/view/journals/068/2026/004/article-A001-en.xml 5. "A Survey of Zero-Knowledge Proof Based Verifiable Machine Learning", arXiv:2502.18535: state of the art of zkML (training, testing, inference) and its limits: https://arxiv.org/abs/2502.18535 6. Luca Nannini et al., "AI Agents Under EU Law", arXiv:2604.04604: classification and liability of autonomous agents under the European AI regulation: https://arxiv.org/abs/2604.04604 7. Crypto Economy, correction of AI agent tokens (ai16z and Virtuals): https://crypto-economy.com/ai-agent-tokens-face-market-pressure-as-ai16z-and-virtuals-drop-sharply/ 8. CryptoRank, class action against the creators of ai16z and ElizaOS for alleged fraud: https://cryptorank.io/news/feed/0d429-ai16z-elizaos-creators-sued-fake-ai-hype 9. Yellow Research, the gap between AI compute demand and supply, and crypto GPU networks (io.net, Akash, Render, Bittensor): https://yellow.com/research/ai-compute-demand-crypto-gpu-networks-gap-2026 10. Visa, "New AI, Stablecoin and Token Innovations to Power Intelligent, Programmable Commerce": TradFi's response: https://investor.visa.com/news/news-details/2026/Visa-Announces-New-AI-Stablecoin-and-Token-Innovations-to-Power-Intelligent-Programmable-Commerce-at-Visa-Payments-Forum/default.aspx 11. l0g, guides Stablecoins and the GENIUS Act, Reading on-chain data and articles The GENIUS Act, the 18 July deadline and Ethereum and TradFi. ============================================================================ ANALYSIS: Private credit: redemptions accelerate as mega-IPOs soak up capital URL: https://l0g.fr/en/analysis/private-credit-redemptions-mega-ipos/ Canonical French source: https://l0g.fr/posts/credit-prive-rachats-mega-ipo/ Date: 2026-07-14 (reviewed 2026-07-14) Topics: private credit, systemic risk, BDC, liquidity, IPO, macro ---------------------------------------------------------------------------- Withdrawal requests at the large private-credit funds are hitting a new quarterly record, while fundraising collapses and SpaceX, OpenAI and Anthropic come to compete for the same capital. A quantified overview, cross-checked on primary sources. A record in redemption requests, cross-checked The starting point is a tally published by The Kobeissi Letter from Robert A. Stanger & Co. data: across the five large funds tracked (Golub, Cliffwater, Oaktree Strategic Credit, Blackstone, BlackRock HPS), redemption requests reach about $12bn in the single second quarter of 2026, up $4.3bn (or +56%) on the prior quarter. Two funds concentrate most of the surge. The Cliffwater Corporate Lending Fund (CCLFX, about $31 to 33bn of assets) records the largest rise, +$3.0bn, taking its requests to $5.3bn. The figure squares with Bloomberg: requested redemptions represented nearly 17% of shares in the second quarter, against about 14% in the first; the fund brought its cap down to 5%, serving only about a third of the amounts requested. Next comes the Blackstone Private Credit Fund (BCRED, about $79 to 82bn), whose requests rise by $720m, to $4.5bn, nearly 10% of its shares. The contrast with the previous quarter is marked. In the first quarter, BCRED had honoured 100% of requests (7.9% of shares) after a $400m injection from Blackstone and its executives. Blue Owl, for its part, had seen up to 40.7% of shares requested for redemption on its technology fund and 21.9% on its flagship credit fund, before capping at 5%. The shift from full service to generalised capping is a regime change, not a mere blip. Fundraising collapses faster than exits rise The least-commented part of the file is not the level of redemptions, but the drying up of inflows. According to Stanger, sales of BDCs (business development companies, the retail vehicle of private credit) fell to about $1.6bn in April, down 74% year on year and the lowest monthly figure since May 2023. Over the first four months of the year, cumulative fundraising reaches $10.8bn, down 52%. Above all, the first quarter of 2026 saw, for the first time in the sector's history, honoured redemptions ($6.9bn) exceed fundraising ($4.9bn, down 59% year on year). Stanger speaks of a pullback phase in its liquidity cycle and anticipates a roughly 40% drop in BDC fundraising over the year, by analogy with the reversal of non-traded REITs in 2022-2023. Why holders exit, and why it is delicate The trigger comes from a sector fear: the AI disruption of software vendors, a sector heavily financed by private credit. Blue Owl explicitly attributed the surge in its redemptions to market worries tied to this disruption. Over the October 2025 to February 2026 period, software stocks fell about 30%, BDC shares about 10%. Managers counter these exits with the strength of their portfolios: Cliffwater recalls that its fund received an A rating from S&P Global Ratings in November 2025, citing diversification, measured leverage and asset quality, and considers that the withdrawals owe more to sentiment than to fundamentals. The thesis remains unverifiable in the short term, for lack of observable market prices, which is precisely the heart of the problem. The difficulty is structural. Private-credit assets are valued to model rather than to market, with a 60-to-90-day lag on net asset values (NAV). The gap shows in the indices: in the first quarter, the Stanger index of non-traded BDCs posts -0.03% (its first negative quarterly performance since Q2 2022) and +6.2% over twelve months, while the S&P index of listed BDCs loses 10.1% over the quarter and 14% year on year. When too many holders want out, funds activate their caps (the gates) to avoid selling illiquid assets at a loss; the flip side is that a holder served 5% for 17% requested recovers only about 29 cents per dollar requested. The contagion channel: insurers and banks This is the point on which regulators concentrate their attention. On 6 May 2026, the Financial Stability Board (FSB) estimated the market between $1,500bn and $2,000bn at end-2024, heavily concentrated in the United States, euro area and United Kingdom, and warned that its complexity, leverage and interconnections could amplify a shock. The transmission routes are identified. As early as April 2026, the Federal Reserve queried the large US banks on their private-credit exposure to assess the risk of spillover to the rest of the system. The IMF, in its spring report, underlined US insurers' particular exposure to BDC leverage. The New York Fed, for its part, recalls that banks remain the providers of financing and liquidity to non-bank actors (NBFI): the risk, ultimately, comes back to them, in a pattern reminiscent of the ABCP conduits and SIV vehicles of before 2008. The IMF notes in passing that these actors now hold nearly half of world financial assets. The warnings are not new: Jeffrey Gundlach compared the enthusiasm for private credit to 2006 conditions as early as June 2025, and Jamie Dimon regularly points to the lack of transparency and the quality of valuations. On the opposite side, SEC chair Paul Atkins played down the systemic risk of the non-bank sector. This divergence of appraisal, between cautious regulators and a reassuring market authority, is in itself a factor to watch. The agencies, for their part, have begun to adjust their reading: Moody's cut the outlook on Blue Owl's OCIC fund to negative on 7 April 2026, because of redemption requests markedly higher than those of its peers. The mega-IPOs, a drain at the worst moment The calendar adds pressure. On 12 June 2026, SpaceX made its stock-market debut at a targeted valuation of about $1,750bn, closing above $2,000bn: the largest listing in history, with a float of only about 4.3% and nearly 30% of the offer reserved for retail (about $22bn). OpenAI filed on 8 June for a valuation of up to $1,000bn, targeting a listing in the second half. Anthropic is aiming for October, for more than $60bn. Together, this trio could raise $200 to 240bn. The effect on liquidity is debated. For Ed Yardeni, $200bn remains a fraction of the S&P 500's roughly $60,000bn of market cap: the issue would be less a global drain than a supply shortage on SpaceX's tiny float. Conversely, a BNP Paribas note relayed by CNBC puts at up to $50bn the potential liquidations (crypto, semiconductors, leveraged ETFs) to fund the SpaceX tranche alone; bitcoin has lost nearly a third since the start of the year, with ETF outflows of $3.1bn. The junction point with private credit sits there. These listings court, through their broad retail tranches, the same dollar of discretionary risk that fuelled the rise of BDCs, precisely when that fundraising collapses. The dominant risk is therefore not a mechanical drain of the whole market, but the reinforcement of a rotation already under way out of private credit. And a fund whose exits exceed its inflows depends on fresh flow to honour its redemptions without selling assets; drying up that flow brings the capping deadline closer. To note, as a caution signal on the seller side: more than 600 OpenAI employees and former employees sold about $6.6bn of shares on the secondary market ahead of the listing. What to watch Four signals deserve close tracking: the quarterly redemption rate relative to the effective cap at the large funds; the persistent gap between model-based NAVs and listed comparables; the share of capitalised interest (PIK) in declared income, a sign of borrowers no longer paying in cash; and the evolution of the bank credit lines extended to managers. Against a backdrop of stretched equity valuations (S&P 500 P/E near 32x against a fifteen-year average of 20x, Shiller CAPE ratio around 42, and a top 10 representing nearly 38% of the VOO ETF and half of the QQQ), the coincidence of an exit from private credit and a wave of record listings sketches the playing field of the coming quarters. Sources - The Kobeissi Letter, Q2 2026 tally (Robert A. Stanger & Co. data). - Robert A. Stanger & Co. / AltsWire, BDC fundraising -74% in April; redemptions exceeding fundraising in Q1, Stanger indices. - Bloomberg, Cliffwater, 17% redemption requests in Q2 (2 June 2026). - Seeking Alpha, cap brought down to 5%. - Investment Executive, BCRED 7.9% honoured in Q1 and $400m injection, Blue Owl, Moody's OCIC outlook. - Financial Stability Board, report on private-credit vulnerabilities (6 May 2026). - Fortune, the Fed queries banks on their exposure (April 2026). - IMF, GFSR spring 2026 transcript; non-banks and stability. - Federal Reserve Bank of New York (Cetorelli), banks as liquidity providers to non-banks. - Reuters / Seeking Alpha, SpaceX listed above $2,000bn, valuations and concentration (15 June 2026). - Yahoo Finance, OpenAI IPO filing (8 June 2026). - Stocktwits / Yardeni Research, the drain fears would be exaggerated. - 99Bitcoins / CNBC / Reuters, BNP Paribas note, about $50bn of potential liquidations. ============================================================================ ANALYSIS: Strategy: Saylor's bitcoin bet URL: https://l0g.fr/en/analysis/strategy-saylor-bitcoin-bet/ Canonical French source: https://l0g.fr/posts/strategy-pari-bitcoin-saylor/ Date: 2026-07-14 (reviewed 2026-07-14) Topics: crypto, markets, bitcoin ---------------------------------------------------------------------------- A Minnesota software company become the largest listed bitcoin holder on the planet: that is the story of Strategy, Michael Saylor's former MicroStrategy. At the end of June 2026, it owns 847,363 bitcoins, paid around $64 billion. And an event has just happened that its supporters judged impossible: its market capitalisation fell below the market value of its own treasury. The model, brilliant as long as bitcoin rose, faces its first real test. Here is how it works, and where it can break. A software company become a bitcoin reserve Renamed Strategy in February 2025, the company led by Michael Saylor, its executive chairman, and Phong Le, its chief executive, has made bitcoin its reason for being. Its historic software activity still exists, but it has become marginal against a treasury of nearly $64 billion. Over some $50 billion raised in five years on the markets, the company has accumulated 847,363 bitcoins at an average cost of about $75,650 per unit, and openly targets a million. To buy a Strategy share is first to buy leveraged exposure to bitcoin. The engine: the premium to net asset value The heart of the mechanics comes down to one ratio, mNAV, the market cap relative to the value of the bitcoin held. As long as the stock trades above this net asset value, issuing new shares is accretive: the company raises dollars at a price above the underlying value, buys bitcoin, and the amount of bitcoin per share rises for existing shareholders. This is what Strategy measures through its own metric, "BTC Yield", at 9.4% since the start of 2026. The paradox is only apparent: as long as the dollars raised buy more bitcoin per share than dilution removes, the operation enriches the incumbent shareholder, what the company calls accretive dilution. The stronger the premium, the faster the machine runs. Two fuels add to this engine: more than $7 billion of convertible bonds, and a stack of five preferred shares issued in 2025, the so-called "Digital Credit" instruments. The 2026 turn In June 2026, bitcoin fell back to around $65,000, below Strategy's average cost of $75,650: the treasury is at an unrealised loss. Above all, mNAV fell below 1.0 for the first time in the company's history, its market cap now worth less than its bitcoins. The signal is heavy, because no treasury company had ever durably held below its net asset value. Chain consequences: at the end of May, Strategy sold bitcoin for the first time in four years, 32 units, to pay a preferred dividend, and management acknowledged it could sell more if the premium stayed absent and financing closed. It is the explicit abandonment of the "never sell" dogma that had been Saylor's brand. At the same time, its STRC preferred was trading below its $100 par, a sign that even the preferred channel is seizing. Cash commitments, with no cash to pay them Here is the knot of risk. The five preferreds pay cash dividends, from 8% for STRK to 12% for STRC, and the convertible debt bears interest. In all, Strategy faces about $800 million a year of interest and dividends. Yet its software business does not generate such cash. To honour these payments, the company must therefore issue more securities, which the discount makes costly, or sell bitcoin, which it has just started to do. Strategy replies that its treasury covers 71 years of dividends at the current value, when a Grayscale analyst judges it prudent to sell at least $3 billion of bitcoin to secure two years of bonds. Both readings coexist, but one thing is certain: the common shareholder is served last, behind the debt and all the preferreds. To this fragility is added an accounting volatility: since January 2025, a new standard obliges Strategy to value its bitcoin at market price in its income statement, so that a quarter of decline translates into massive displayed losses, unrelated to its real cash. A reflexive bet The model is reflexive: the premium feeds bitcoin buying, which feeds the premium, both ways. On the way up, it is a remarkable amplifier. On the way down, it is a trap, as in 2022 when the stock fell nearly 89% while the S&P 500 lost 25%. Three structural pressures aggravate the picture today. Spot bitcoin index funds, launched in 2024, offer the same exposure without the debt or the preferreds, which erodes the premium's reason for being. The index provider MSCI backed off, in January 2026, from excluding digital-treasury companies, and the stock kept its place in the Nasdaq 100 at the December 2025 review, but MSCI maintained a re-examination, JPMorgan sizing a possible exclusion at several billion of passive outflows. And the dependence on a single man, Saylor, as on a single asset class, concentrates the risk. At bottom, Strategy is neither a fraud nor a sure thing: it is a leveraged bet, transparent and assumed, on the durable rise of bitcoin. As long as the asset rises and the premium holds, the mechanics enrich the shareholder. When bitcoin ebbs and the premium fades, the debt, the dividends and the discount reassert themselves, and the company finds itself forced to choose between selling its treasury or diluting its holders. The bet stays binary, and 2026 delivers its first real-scale test. For anyone who wants to understand what lies behind such exposure, knowing how to read on-chain data and an issuer's capital structure has never been so useful. --- Primary sources: Strategy 8-K filings with the SEC (bitcoin holdings, IPOs of the STRK, STRF, STRD, STRE and STRC preferreds, capital-allocation framework, first bitcoin sale of May 2026); CoinDesk (tracking of weekly purchases, mNAV falling below 1, May 2026 sale); strategy.com, Notes section (definition of mNAV, terms of the STRC preferred at 12% and STRK at 8%); first-quarter 2026 earnings call (remarks by Michael Saylor and Phong Le on a possible bitcoin sale); Grayscale, Zach Pandl (bond coverage); JPMorgan (estimate of passive outflows in case of MSCI exclusion); Forbes, Yahoo Finance and BeInCrypto (analysis of the premium compression and performance compared with 2022). ============================================================================ ANALYSIS: Hyperliquid: the on-chain exchange that buys back its own token and knocks on TradFi's door URL: https://l0g.fr/en/analysis/hyperliquid-onchain-exchange/ Canonical French source: https://l0g.fr/posts/hyperliquid-bourse-onchain-tradfi/ Date: 2026-07-14 (reviewed 2026-07-14) Topics: crypto, markets, regulation, tech ---------------------------------------------------------------------------- Most crypto exchanges are black boxes: you see the prices, never the workings. Hyperliquid takes the opposite tack. Its order book, its matching engine and its liquidations live entirely on a public blockchain, verifiable in real time. In under two years, this protocol has become the leading decentralised derivatives market, with a business model that stands out: nearly all its fees serve to buy back its own token on the market. And it is starting to build bridges toward traditional finance, from ETFs to equity perpetuals. Here is the machine, its fuel, its doors, and its fragilities, put into data. Hyperliquid is not a mere application, it is a layer-1 blockchain designed for one thing, running an exchange. It rests on two engines sharing the same consensus, named HyperBFT. The first, HyperCore, is an on-chain order book able to process up to 200,000 orders per second, with one-block finality. The second, HyperEVM, launched on 18 February 2025, is an Ethereum-compatible environment that allows deploying smart contracts accessing the order book's liquidity directly, with no bridge between two chains. Where centralised exchanges keep their innards secret, every order, every transaction and every liquidation is publicly traceable there. It is a promise of radical transparency, exactly the terrain this journal seeks to illuminate. The real weight: volumes and market shares The figures situate the scale of the phenomenon. Over a thirty-day window in spring 2026, Hyperliquid processed about $172.63 billion of perpetuals volume, nearly 32% of the volume recorded across all decentralised derivatives exchanges, and 3.3 times the volume of the second player, Aster. Open interest exceeded $9 billion. On a daily basis, some estimates put its dominance above 50% of the segment, the gap owing to methodologies and to suspicions of artificial volume at some competitors. On valuation, and these figures move fast, the HYPE token had a market cap of around $14 billion at the end of June 2026, for a fully diluted value of about $60 billion, tenth among crypto assets, after an all-time high of $76.70. The protocol generates real revenue, on the order of several million dollars of fees a day, an annualised pace above $1.3 billion. This is not a use-less token backed by a promise, it is a market infrastructure that collects commissions. The business model: the buyback loop This is where Hyperliquid really stands out. The HYPE token was distributed at the end of November 2024 through an airdrop, with no venture capital, no private sale, about 31% going directly to the community. But the originality lies in the value engine. Nearly all transaction fees, around 97% per the parameters, are paid to an assistance fund that buys back HYPE on the open market. In parallel, the gas fees paid on HyperEVM are burned. The result is a mechanical loop: the higher the volume, the higher the fees, the heavier the buybacks, which supports demand for the token, independently of speculation. This logic brings HYPE close to a stock that would devote all its earnings to buying back its own shares. Over a recent ninety-day period, the protocol bought back about $135 million of token, helping to absorb the selling pressure of unlocks. It is a rarity in the crypto universe, a direct and automated link between a platform's real activity and the value of its token. Whether this link holds at scale remains to be seen, and that is the subject of the final part. HIP-3 and HIP-4: the market factory Hyperliquid does not stop at crypto perpetuals. In October 2025, the HIP-3 update opened permissionless creation of perpetual markets: anyone can now launch a market by locking up HYPE, on commodities, equities, currencies or indices. Four months after its launch, this function already represented about 10% of the protocol's revenue, driven notably by perpetuals on silver and oil. In February 2026, HIP-4 added opinion markets, binary contracts on events, further broadening the audience beyond crypto traders alone. The trajectory is clear: to move from a perpetuals exchange to an on-chain financial operating system, able to host almost any asset. The doors to traditional finance This is the heart of the matter. Hyperliquid opens several distinct accesses to the world of classic finance. The first is the regulated wrapper. In May 2026, the managers Bitwise and 21Shares launched spot ETFs on HYPE, which gathered more than $137 million of assets by early June. These products let a traditional investor gain exposure to the token without custody, and inject regulated capital, less sensitive to the cycles of crypto retail trading. The second access is the repatriation of traditional assets onto the chain. Via HIP-3, perpetuals on equities, indices, commodities and currencies now trade in the same transparent order book as cryptos. Traditional finance does not only come to buy the token, it sees its own underlyings become tradable products on Hyperliquid's infrastructure. The third access is distribution. The Builder Codes mechanism lets third-party platforms build their own interfaces on Hyperliquid by routing their users to it, in exchange for a fee share. It is a white-label infrastructure logic, which turns the protocol into rails rather than a mere application. Add ongoing institutional integrations on custody and settlement. Put end to end, these doors draw a shift: from a decentralised-exchange token to a market-infrastructure asset, valued as such. For the regulatory context of this convergence, see our analyses of MiCA and the GENIUS Act on stablecoins. The limits and the risks Enthusiasm does not exempt one from a severe examination, and several fragilities deserve to be laid out plainly. The first is dilution. With about 220 million tokens in circulation out of one billion, regular unlocks will keep feeding supply, and the buyback mechanic only supports the price if it absorbs those unlocks. Yet buybacks depend on volume, itself cyclical and sensitive to risk appetite. In a bear market, fees ebb, buybacks weaken, and the loop seizes at the worst moment. The second fragility touches governance and centralisation. In March 2025, a large position on a token named JELLYJELLY caused a liquidation risk that the whole set of validators resolved by removing the market. The episode showed that a network presented as decentralised could intervene in an emergency in a very concentrated way, and it raised the question of the risk carried by the community liquidity fund that serves as a buffer. The third is regulatory. Access is restricted in several jurisdictions, including the United States, and established exchanges like the CME are pressing the US regulator to frame these platforms, invoking the risks of manipulation and sanctions evasion. A customer-identification requirement would collide head-on with the permissionless model. There remains, finally, a concentration risk. That a single protocol dominates decentralised perpetuals to this extent creates a single point of failure for the whole segment. And the very measure of this dominance is debatable, since some competitors' volumes are suspected of being inflated. On-chain transparency precisely allows these figures to be verified, which changes the nature of the debate compared with opaque exchanges. What Hyperliquid lets us see Hyperliquid is a rare case where a crypto token corresponds to real cash flows and to publicly verifiable data, where the sector usually runs on narrative. Its doors to traditional finance are concrete, from ETFs to equity and commodity perpetuals, through white-label distribution and institutional custody integrations. They remain, however, curbed by regulation and threatened by dilution. The real question is not whether this bridge exists, it does, but whether it will hold without recreating the opacity and centralisation Hyperliquid claims to replace. The good news, for anyone who wants to judge on the record, is that everything happens on a public chain. The verdict will read there in data, not in promises. --- Primary sources: DefiLlama, Hyperliquid dashboards (perpetuals volume, fees, revenue, open interest, market shares of decentralised derivatives exchanges, routing of about 97 to 99% of fees to the assistance fund); CoinGecko (price, market cap of about $14 billion, fully diluted value of about $60 billion, circulating supply of about 220 million out of one billion, all-time high of $76.70, end June 2026); Hyperliquid documentation and HIP-3 (October 2025) and HIP-4 (February 2026) proposals, HyperEVM (18 February 2025), HyperCore and HyperBFT; specialist-press reporting on the Bitwise and 21Shares ETFs (May 2026, more than $137 million of assets), on token buybacks and on the JELLYJELLY episode of March 2025. Market figures are dated to mid-2026 and evolve rapidly; competing platforms' volumes are subject to caution. ============================================================================ ANALYSIS: The CLARITY Act: the bill set to redraw crypto regulation in the United States URL: https://l0g.fr/en/analysis/the-clarity-act-us-crypto-regulation/ Canonical French source: https://l0g.fr/posts/clarity-act-regulation-crypto-etats-unis/ Date: 2026-07-14 (reviewed 2026-08-06) Topics: regulation, crypto, us politics ---------------------------------------------------------------------------- Update, 7 August 2026. The CLARITY Act has not become law. The official tracker still labels it Passed House, not Passed Senate. The Senate's planned state work period from 10 August through 11 September does not close the procedure, but it leaves fourteen scheduled session days, from 14 September through 2 October, for the pre-midterm sequence. Our new status article separates facts from scenarios: CLARITY Act: the window before the midterms. A ten-year battle nearing its end On 14 May 2026, at 10:30 in the Senate's Dirksen room, the Banking Committee adopts by 15 votes to 9 the Digital Asset Market Clarity Act. Two Democrats, Ruben Gallego (Arizona) and Angela Alsobrooks (Maryland), cross the aisle to join the thirteen Republicans. For the first time since the origin of the crypto debate in Congress, a complete market-structure text clears the critical stage of a Senate committee with a documented bipartisan vote (source: TheStreet Crypto, 14 May 2026). The scope of the event exceeds mere parliamentary procedure. The CLARITY Act is the first federal text that structurally resolves the turf war that has poisoned the industry for ten years: the SEC, historically led by Gary Gensler, considered nearly all tokens as securities (relying on the 1946 Howey precedent), while the CFTC claimed jurisdiction over the main assets like Bitcoin and Ethereum, which it qualifies as commodities. This ambiguity had pushed tens of billions of dollars of capital and talent out of the US market, a flight documented by Coinbase, Andreessen Horowitz, and multiple crypto think tanks. The CLARITY Act settles it. Genesis and legislative timeline The original text was introduced by French Hill, chair of the House Financial Services Committee, on 29 May 2025. It was adopted by the House of Representatives on 22 July 2025 by 294 votes to 134, the largest bipartisan vote ever recorded on a crypto text in Congress (source: Latham & Watkins, US Crypto Policy Tracker). The text then stayed stuck in the Senate for ten months, while the Senate Banking Committee worked its own version (under the name Responsible Financial Innovation Act of 2025), with successive discussion drafts published in July 2025 (Tim Scott and Cynthia Lummis), September 2025 (182 pages), then January 2026 (278 pages). On 11 May 2026 at midnight, the final 309-page version is made public before the 14 May markup (CoinDesk, 12 May 2026). On 14 May, after six hours of contentious hearing marked by the frontal attacks of Elizabeth Warren, the bill clears the markup and is now on its way to the Senate floor. To become law, it will still need to: - Reconcile the Banking version with that of the Senate Agriculture Committee (which covers CFTC jurisdiction) - Reach the 60 votes on the Senate floor, which will require at least seven additional Democrats - Reconcile the final text with the House version - Receive President Trump's signature The realistic timeline aims for enactment before the end of 2026, ideally before the November midterms. The core of the mechanism: the SEC / CFTC split The bill institutes a fundamental legal dichotomy between two categories of digital assets, each falling under a distinct regulator. Securities under the SEC Tokens sold within an investment contract (in the sense of the Howey precedent) remain by default subject to SEC jurisdiction. This is the "default" status of new tokens at their primary launch. Digital commodities under the CFTC A new category of "digital commodities" is created and placed under the exclusive jurisdiction of the CFTC, including on spot markets, a major change, since the CFTC historically had jurisdiction only over derivatives. According to Gibson Dunn (November 2025 analysis), for an asset to qualify as a digital commodity, it must cumulatively: 1. Be intrinsically linked to and derive its value from a "mature blockchain system" 2. Be sufficiently decentralised 3. Not confer property rights (debt, equity, liquidation rights, interest, dividends) The key concept: the "mature blockchain" This is probably the most important conceptual innovation of the bill. According to Tangem's analysis (February 2026) relaying the legislative text, a mature blockchain is defined as "a blockchain system, with its associated digital commodity, that is not controlled by any person or group of persons under common control". For a network to qualify as mature, it must satisfy specific criteria: - Full operational functionality - Effective decentralisation: no entity controls more than 20% of the supply or voting power - Absence of unilateral update authority by the founders or the initiating company This qualification creates a migration path: a token can start as a security (primary launch) and "graduate" to digital commodity when the network becomes sufficiently decentralised. This is what Davis Wright Tremaine calls, in its January 2026 analysis, the "tokenized continuum". The "ancillary asset" concept A third, hybrid category was introduced: the "ancillary asset", defined as "a network token whose value depends on the entrepreneurial or managerial efforts of an ancillary originator or a related person" (bill text, 11 May 2026 version). The bill creates a rebuttable presumption that a network token is an ancillary asset, unless the originator submits to the SEC a written certification with reasonable evidence demonstrating the contrary. The SEC has 60 days to reject the certification on the basis of factual elements. When an asset is certified as non-ancillary, it escapes SEC jurisdiction and moves under the CFTC. Regulation Crypto: the new registration exemption The bill creates a registration exemption to the Securities Act named "Regulation Crypto" (Section 103 of the text). According to the section-by-section summary published by the Senate Banking Committee: - An issuer may raise the greater of: (1) $50 million per calendar year for four years, or (2) 10% of the total dollar value of ancillary assets in circulation - The absolute cumulative cap is $200 million in gross proceeds - Secondary transactions on these tokens become free once certification is done This is a major regulatory innovation. Today, launching a token in the United States requires either a Regulation D exemption (reserved for accredited investors, so retail excluded), or a full S-1 registration (extremely costly and legally risky). Regulation Crypto creates a viable middle path. A second exemption, "Regulation DA" (Digital Assets), frames the semiannual disclosure obligations for blockchains in the process of maturing. Issuers must report to the SEC the state of the blockchain, the efforts of the issuer and related persons, as well as the financial information tied to the blockchain (Patomak Global Partners, August 2025 analysis). The stablecoin compromise: the subject that blocked everything For months, the main sticking point was not classic tokens, but stablecoins and their yield. The debate pits: - Traditional banks (the American Bankers Association leading), who fear that a yield-bearing stablecoin becomes a direct competitor to interest-bearing deposit accounts - The crypto industry (Coinbase, Circle, Tether), which defended the freedom to remunerate stablecoin holders The final compromise in the 11 May 2026 version prohibits rewards on the mere passive holding of stablecoins when these rewards are "economically or functionally equivalent to the payment of interest or yield on an interest-bearing bank deposit" (CoinDesk). But the bill allows stablecoin rewards or incentives tied to an activity, that is, loyalty programmes, cashback, or yields tied to the active use of the stablecoin in DeFi protocols. This dividing line, negotiated by the skin of its teeth, partly satisfies both camps. It is what unblocked the markup. The DeFi victory: developer protection This is one of the most important advances for the ecosystem. The bill integrates the Blockchain Regulatory Certainty Act (BRCA), which establishes that: - Non-custodial developers (those who do not control users' funds) are not considered money transmitters - Fully decentralised DeFi protocols can operate without a FinCEN licence - Smart-contract developers do not bear responsibility for users' actions This is a fundamental legal protection that responds directly to cases like Tornado Cash (prosecutions against developers) or Samourai Wallet (arrests). The DeFi Education Fund welcomed this integration: "we are encouraged by the direction of recent negotiations and note that the most important provisions for developers and infrastructure providers, the BRCA and the protections under the Exchange Act, are in this bill" (CoinDesk, 12 May 2026). The text also covers important technical elements: - Legal recognition of DAOs: a DAO is not treated as a "single person" for the purposes of the Securities Law's control provisions - Bankruptcy safe harbor: digital-commodity transactions benefit from protections similar to those of conventional derivatives in a custodian's bankruptcy - Portfolio margining: the SEC and CFTC are required to jointly issue rules allowing portfolio margining across securities, swaps, futures and digital-commodity accounts - NFTs: generally excluded from the Securities Laws' perimeter - Tokenized securities: remain securities for all purposes The flashpoints: what was not settled The Trump ethics question This is the most politically charged point. According to Elizabeth Warren's official statement at the markup, President Trump and his family reportedly accumulated at least $1.4 billion of gains on crypto deals since he took office (Senate Banking Committee, 14 May 2026). World Liberty Financial (the Trump family's crypto company) and the TRUMP memecoin are at the heart of the controversy. Warren and Democratic senator Jack Reed wrote to the Department of Justice and the Treasury to request an investigation into World Liberty Financial. They cite a Wall Street Journal article on a partnership with a firm named AB, which had previously tried to develop a resort in East Timor under the direction of two individuals sanctioned by the Treasury for "pig butchering scams". World Liberty denied any direct association with these sanctioned individuals (CoinDesk markup liveblog). Several Democratic amendments to create ethics guardrails applicable to the president, vice-president, lawmakers and senior officials were rejected. Patrick Witt, White House crypto adviser, publicly declared that the negotiating position was to establish rules applying "to everyone, from the president to the Capitol Hill intern", but to reject any provision specifically targeting an officeholder. The ethics question partly falls under other Senate committees (notably Ethics), which complicates resolution. Senator Mark Warner explicitly conditioned his final vote on the inclusion of these guardrails. The rejection of the Warren amendments The 14 May markup was the occasion for dozens of Democratic amendment votes, almost all rejected: - Restoration of sanctions authority over DeFi platforms (reference to the Tornado Cash case): rejected 11-13 - Prohibition on investing certain digital assets in retirement accounts: rejected 11-13 - Publication of bank-supervision information tied to Jeffrey Epstein: rejected 11-13 - Restriction of the Federal Reserve on master accounts for uninsured deposit institutions engaged in digital assets These serial rejections set the tone of the text's political orientation: favourable to the industry, restrictive on regulatory expansion. The mapping of positions The supporters - The Trump administration and the White House (Patrick Witt, David Sacks former crypto czar) - Senate Banking Republicans (Scott, Lummis, Boozman) - The major crypto industry: Coinbase (CEO Brian Armstrong called the bill a "real compromise" that "could redraw how Americans interact with money and financial markets"), Circle, Ripple, Andreessen Horowitz - DeFi Education Fund and The Digital Chamber The opposition and critics - Senator Elizabeth Warren (D-MA, ranking member Banking): "this bill puts investors, our national security and our entire financial system at risk, and it will turbocharge Donald Trump's crypto corruption" - NASAA (North American Securities Administrators Association, which groups the regulators of the 50 US states): documented opposition in January 2026 and an official statement after the vote (14 May 2026) deploring "the advancement of a bill with provisions that bad actors will seek to exploit" - Unions: AFL-CIO, SEIU, AFT, NEA, AFSCME, who fear the exposure of retirement accounts and pension funds to crypto volatility - American Bankers Association: pressure to harden stablecoin restrictions (partly satisfied) The Democrats open to negotiation Three names to watch to reach 60 votes: Ruben Gallego, Angela Alsobrooks (who voted for it in committee), and Mark Warner. On the side of progressive senators like Bernie Sanders or Ed Markey, the opposition stays strong. Expected consequences for the crypto ecosystem For centralised exchanges This is probably the sector that gains the most from enactment. Coinbase would see a major share of its listings slide under the CFTC (digital-commodity status), considerably less costly to serve than the SEC. Brian Armstrong has quantified the stake several times: the end of multi-year proceedings with the SEC frees several hundred million dollars in legal costs and regulatory fees. Exchanges will have to register as "digital commodity exchanges" with the CFTC, with obligations to protect client funds, monitor markets, and report. More constraining than the de facto status quo, but radically more predictable. For DeFi protocols Structural victory. Non-custodial developers are exempted from money-transmitter status, so escape FinCEN jurisdiction. Truly decentralised protocols (Uniswap, Aave, Compound) now operate in a clear legal framework. The criminal risk weighing on developers (the Tornado Cash case) is neutralised for non-custodial architectures. For ETFs and asset management The spot crypto ETFs already approved (Bitcoin, Ethereum, and more recently XRP, Solana) now operate in a statutory rather than case-law framework. The conditions for approving new ETFs are clarified. BlackRock, Fidelity, Franklin Templeton and VanEck actively lobbied for this text. For token issuers Regulation Crypto creates a fundraising vehicle able to partly replace the ICOs (Initial Coin Offerings) of the 2017-2018 period, but this time in an explicit legal framework. US-based crypto startups could raise up to $50M a year for 4 years (cumulative cap $200M) without a full S-1 registration, while respecting disclosure obligations. For stablecoins The stablecoin market (USDT, USDC, RLUSD) now operates under a federal framework (combined with the already-adopted GENIUS Act). The prohibition of passive yields forces issuers to rethink their business models. Circle (USDC) and Ripple (RLUSD) are structurally more exposed than Tether (USDT), which operates mostly outside the United States. For Layer-1 and Layer-2 innovation Blockchains able to demonstrate their maturity (20% decentralisation, absence of unilateral control) move under the CFTC. For Bitcoin and probably Ethereum, this is settled. For younger Layer-1s (Solana, Avalanche, Sui), certification becomes a major strategic process. Layer-2s (Arbitrum, Optimism, Base) are in a greyer zone. For the market: the US geographic premium The most important macro effect could be the return of capital and talent to the United States. Singapore, Dubai and Switzerland attracted tens of billions of dollars of crypto industry during the Gensler years. With a clarified and competitive US framework, this flow could reverse. International comparison: where does the CLARITY Act stand? The CLARITY Act is explicitly presented as a response to MiCA (Markets in Crypto-Assets Regulation), the European regulation that entered full application in December 2024. According to KuCoin Research, the bill "aligns with global efforts like MiCA and addresses risks such as terrorism financing". Key comparisons: - MiCA is more protective (consumer protection paramount, strict framing of CASPs, Crypto-Asset Service Providers) - CLARITY is more pro-innovation (more generous Regulation Crypto, DeFi explicitly protected) - Hong Kong and Singapore have comparable frameworks but with less retail penetration - The United Kingdom is more advanced on stablecoins (BoE) but lags on market structure The CLARITY Act, if it becomes law, could give the United States back global regulatory leadership on digital assets, leadership lost during the Gensler years (2021-2025). Risks and grey zones Several serious technical criticisms deserve consideration: 1. The definition of "ancillary asset": NASAA explicitly pointed to an internal contradiction in the text. The ancillary asset is defined as a subtype of network token (therefore a non-security digital commodity), but remains subject to an SEC certification. The text stipulates that its value depends on the "entrepreneurial or managerial efforts of others", exactly the definition of a security per Howey. This ambiguity risks generating litigation. 2. The 20% decentralisation bar: a potentially gameable threshold. How do you verify a founder's real holdings via multiple wallets? How do you count voting power in a DAO with complex governance structures? 3. The absence of ethics guardrails: if the Ethics committee does not take up the subject downstream, the final bill could leave open the scenario where a sitting president directly profits from the regulation he signs. This is unprecedented in the modern history of US financial regulation. 4. The fragility of the 60 Senate votes: without a substantial ethics agreement, several Democrats who voted favourably in committee could block the floor passage. The realistic timeline According to industry sources (CoinDesk, Fortune, The Block) and the senators' own statements: - May-June 2026: Banking-Agriculture negotiations to merge the two versions - July-September 2026: Senate floor vote (60 votes needed) - September-October 2026: conference committee between House and Senate to reconcile - November 2026: final vote in both chambers - December 2026 · January 2027: presidential signature expected Any agenda slippage risks pushing the final vote beyond the November 2026 midterms, which would potentially alter the balance of power in Congress. Conclusion: a regulatory Rubicon Beyond the technical details, the CLARITY Act represents a paradigm shift. It signs the end of the Gensler doctrine ("regulation by enforcement", systematic assimilation of tokens to securities), in favour of a more predictable regulation by categorisation. For the ecosystem, it is a signal of normalisation and maturity. Institutional capital, the real flows that can take the market from $4 to $20 trillion, needs a stable legal framework. The CLARITY Act, despite its imperfections, provides it. For the text's opponents, it is a gift to an industry that has not demonstrated its capacity to self-regulate, and that now benefits from a framework more permissive than the traditional banking system. The criticisms of Warren, NASAA and the AFL-CIO deserve to be taken seriously: an asset that can produce $1.4 billion of personal gains for the president in one year, in a grey regulatory zone, is not an ordinary industry. History will judge the balance of the text. But one thing is certain: after ten years of turf war, the US crypto industry will finally operate in a structured federal framework. It is the most important regulatory event since the Securities Act of 1933. --- Primary sources: - Senate Banking Committee, Section-by-Section Summary of the Digital Asset Market Clarity Act, 11 May 2026. - CoinDesk, "Clarity Act, in the flesh, unveiled by U.S. Senate Banking Committee before hearing", 12 May 2026. - CoinDesk, markup liveblog, 14 May 2026. - TheStreet Crypto, "Markets surge as Clarity Act clears Senate committee in landmark 15-9 vote", 14 May 2026. - Fortune, "The crypto industry's Clarity Act hits a critical juncture", 13 May 2026. - Bitcoin Magazine, "Senate Banking Committee Opens Historic Crypto Bill Markup", 14 May 2026. - Senator Warren Opening Remarks, Senate Banking, 14 May 2026. - NASAA Statement on Senate Banking Committee Vote, 14 May 2026. - Gibson Dunn, "Update on the U.S. Digital Assets Regulatory Framework", November 2025. - Davis Wright Tremaine, January 2026 analyses. - Latham & Watkins, US Crypto Policy Tracker (April 2026 update). - Patomak Global Partners, "The Future of U.S. Crypto Regulation", August 2025. - The Block, "More than 100 amendments filed…", 14 May 2026. - Tangem Blog, "Which Crypto Assets will Benefit from the CLARITY Act?", February 2026. - CCN, "Tornado Cash, Epstein, Iran, Chokepoint 2.0: Warren Throws Everything at CLARITY Act", 14 May 2026. ============================================================================ ANALYSIS: Private credit, June 2026: a record default, and liquidity closing off URL: https://l0g.fr/en/analysis/private-credit-record-default-liquidity-closing/ Canonical French source: https://l0g.fr/posts/credit-prive-juin-2026-defaut-record-gating/ Date: 2026-07-14 (reviewed 2026-07-14) Topics: private credit, gating, hlend, blackstone, bdc, valuation, fed ---------------------------------------------------------------------------- The headline figure fits on one line: default is at its highest since the indicator existed. The useful information is elsewhere, in the gap between what the figures show and what they leave unsaid. On 15 June, Fitch published its monthly update on US private credit. The headline stayed the same as the previous month: the default rate holds at its record. This is a non-event, and that is precisely what makes it interesting. A market that many described in December as due to ease in 2026, thanks to rate cuts, instead shows default stuck at its highest, while the exit closes off for the savers who would like their money back. Nothing spectacular, no resounding bankruptcy this month, and yet three counters are rising in parallel. This article takes stock, on primary sources, in the continuity of the journal's private-credit coverage and of the book Auditing Opacity (Auditer l'Opacité). No catastrophism: a methodical, risk-oriented assessment. A record default rate, and what it keeps out of sight Fitch's benchmark index, the Private Credit Default Rate (PCDR), tracks about 1,200 middle-market borrowers followed by the agency. It stands at 6% over the trailing twelve months. This level, reached over the period ending in April, was first reported by CNBC on 21 May; the 15 June update shows it holding for the period closed at end-May. It is the highest since the indicator's launch, in August 2024. Its component covering the largest LBO borrowers, the PMR, is higher still, around 9.5% according to Fitch's 15 June report. The detail matters more than the figure. The default events recorded by Fitch in May are, for the most part, not outright payment defaults. They are the introduction of payment-in-kind interest (PIK, which stacks debt instead of paying cash) and maturity extensions. In other words, liability management. A borrower who can no longer pay does not necessarily go bankrupt: it renegotiates, swaps cash for PIK, pushes back the maturity. Default becomes an engineering operation, and it leaves the most visible counter to land in one that few people watch. This is what I called, in the deep analysis on private credit and shadow banking, the silent default. The public leveraged-loan market offers the best illustration of this gap, because there we have two measures side by side. On the Morningstar LSTA index, at 31 May, the payment default rate by issuer comes in at 1.42%, against 1.24% in April. But the dual-track rate, which adds distressed exchanges and liability-management operations, climbs to 3.11%, against 2.84% the previous month. Over twelve months, these liability-management operations represent 54% of total defaults, and they have exceeded classic payment defaults every month since January 2024. The norm has flipped: restructuring out of court has become the default mode. This machinery has an identifiable fuel. Lincoln International, which values about a third of US private-credit portfolios, measures the share of PIK loans: it has doubled, from about 5% of the market in early 2022 to 11% at end-2025. More telling still, "bad PIK", that is, loans initially paid in cash and converted to PIK because the borrower can no longer keep up, has gone from 2% to 6.4% over the same period. PIK is not a default in itself. It is a deferral. But a generalised deferral is a signal about borrowers' real ability to service their debt in a rate environment that stays high. The exit closes off: semi-liquid funds put to the test The test of the moment is not about the default figures, which rise slowly, but about the liquidity of the vehicles sold to the general public. Perpetual BDCs and semi-liquid funds promise a quarterly redemption window, generally capped at 5% of net asset value per quarter, while holding illiquid loans that do not sell in a day. When too many investors want out at the same time, the manager activates the cap. That is what is happening, quarter after quarter. The most closely watched case is HLEND, the $26bn fund BlackRock inherited from the HPS acquisition. In the first quarter, it received redemption requests equal to 9.3% of net asset value, crossed its 5% cap for the first time since inception, and served exits only up to that cap. In the second quarter, requests rose to 13.3%, about 50% more, and the fund again capped at 5%. This is the scenario I described from the first HLEND gating: the liquidity window holds as long as no one uses it en masse. HLEND is not alone. Blackstone's credit fund, BCRED, after honouring a record 7.9% of redemptions in the first quarter by injecting $400m of its own capital, ended up capping in turn. Cliffwater saw 14% of requests on its flagship $33bn fund and gated for the second quarter running. Morgan Stanley capped one of its vehicles after requests at 10.9%. And Blue Owl went further in February, simply closing the redemption window of its OBDC II fund. Moody's cut its outlook on the BDC sector to negative in early April, noting that unlisted perpetual BDCs recorded their very first net outflow in the first quarter. A cold clarification is needed here, because it separates signal from noise. What is exploding is redemption requests, not actual redemptions. Real exits remain capped at 5% by construction. The distinction is crucial: a spike in requests is a sentiment signal, the admission that retail holders want their money back. It is not, in itself, proof of a deterioration in the underlying credit. To confuse the two is to read a panic where there is, for now, a nervousness contained by the very design of the funds. No forced asset sale, no downward spiral in valuations has been observed. Still, the cap protects the fund, not the saver. And it raises the question private credit has avoided for years: that of the freshness of valuations. When 19% of HLEND's portfolio sits in software, a sector hit by the AI-disruption thesis, is the value carried at net asset value the market value? We are squarely on the terrain of zombie funds and private valuations: as long as you do not sell, the NAV stays smooth, and the displayed stability is partly a function of the absence of transactions. The systemic channel: banks, under the Fed's eye The real systemic-risk question is not the fate of a retail fund, it is the interconnection with banks. US banks had lent close to $300bn to the private-credit sector by mid-2025, according to Moody's. These are financing lines to funds, and shared borrowers. That is the route through which a problem lodged in the non-bank sector could climb back toward the core of the system. The Fed's Supervision and Regulation Report, published in June, is measured, and that is what makes it credible. It describes a solid banking system: more than 99% of banks well capitalised at end-2025, CET1 ratio around 13%, total delinquency rate at 1.6%, below the long-run average of about 3%. On exposures to non-depository financial institutions, it notes that delinquency data remain limited. But it adds, and this is the passage to retain, that several high-profile defaults among these actors have raised concern about private credit, and that some banks are reviewing their collateral-management practices on these lines. In the first quarter, large banks cited a strengthening of their monitoring of private-credit exposures. The tone is not alarmist. It has shifted from observation to vigilance. The catalysts have names: the failures of First Brands and Tricolor at end-2025 served as revealers. At the international level, the Financial Stability Board published on 6 May a report dedicated to private-credit vulnerabilities, which sizes the market between $1,500bn and $2,000bn at end-2024 and points to the same blind spots: banking interconnections, opacity of valuations, sector concentration in tech, healthcare and services, leverage in layered structures, and liquidity mismatches in funds with a redemption option. Its most accurate sentence fits in few words: private credit has not yet been tested by a prolonged recession. The FSB now proposes a set of monitoring indicators (fund size, borrower leverage, redemption frequency, retail-versus-institutional investor ratio, sector concentration) and opens four work streams. In parallel, in the United States, the FSOC put out for consultation until 14 May a framework for designating systemic non-banks, and the Office of Financial Research published a note on measuring counterparty exposures to private credit via Form PF. This regulatory ramp-up is the direct continuation of the silent contagion I documented. Risk is not measured by defaults alone: it is measured by the authorities' ability even to see what is happening, and that ability remains, by their own admission, lacking. The other end of the market: the AI megadeals You have to hold both ends of the chain, because private credit is not a homogeneous block. While default concentrates on the smallest borrowers, those with EBITDA below $25m, where the Fitch rate exceeds 11%, the top of the market is living its most spectacular moment. On 5 June, Apollo and Blackstone closed a $35bn financing for Anthropic, one of the largest private-credit deals ever assembled. The structure is worth pausing on: a dedicated vehicle buys Google's TPU chips and leases them to Anthropic, the lease payments servicing the debt, all backed by residual-value guarantees from Broadcom and payment guarantees from Google. The hardware stays off Anthropic's balance sheet, which is convenient for a company preparing its IPO after a $65bn raise in May, at a $965bn valuation. It is the same financial-engineering logic seen in the loan refused to SoftBank or around the SpaceX IPO: you back it, you guarantee it, you deconsolidate it. Private credit has thus become, in the same movement, what a BofA strategist called at end-2025 the lowest-quality asset class in the entire leveraged-finance universe, and the vehicle for the most colossal bets on AI infrastructure. The risk is not uniform: it is structured and backed at the very top, raw and unmanaged at the bottom. This bifurcation is the true face of the market in 2026. The macro backdrop weighs on both ends. The war between the United States, Israel and Iran pushed Treasury yields higher, and since private-credit loans are floating rate, refinancing costs more, which tightens the vise further on fragile borrowers. The rate context that the Fed and the debate over its balance sheet come to complicate is not a neutral backdrop: it is the multiplier of the stress. The picture as of 17 June As of 17 June 2026, the picture is coherent and without immediate drama. The broad default rate is at its record and stays there, at 6%. The true rate, the one that includes liability management, is mechanically higher than the visible counter, because the way distress is managed has changed: you defer and exchange rather than default. The test under way is about the liquidity of retail vehicles, where exit requests are rising and the barriers are coming down, but where actual redemptions remain capped and no forced sale has taken place. The banking channel, for its part, is closely watched, benign on the data, but the regulatory tone has turned to vigilance. The coming quarters will be judged on four elements. The trajectory of redemption requests, to know whether nervousness turns into durable exit. The freshness of valuations, especially on portfolios exposed to software and AI, where the gap between NAV and market would be most likely. Banks' collateral-management practices, which are the leading indicator of the moment the non-bank becomes a banking problem. And finally, whether or not the stress of small borrowers climbs toward the larger ones. None of this heralds 2008: there is not, in the banking system, the high-speed leverage that produced the domino effect. But the opacity is intact, and it is what prevents answering with certainty the only question that matters: what are these portfolios really worth on the day they have to be sold? As long as the answer stays "we'll see when we sell", the record default is only the part of the counter someone chose to switch on. Sources - Fitch Ratings, U.S. Private Credit Default Rate, update of 15 June 2026 (default at its record in May). Press pickup: The Epoch Times, 15 June 2026, - CNBC, Private credit defaults hit record high as interest rates soar, 21 May 2026 (PCDR at 6%, TTM April), - PitchBook LCD, Dual-track leveraged loan default rate jumps amid heavy LME activity (Morningstar LSTA data as of 31 May 2026), - Lincoln International, on the share of PIK and "bad PIK" (cited via crypto.news / MEXC, 13 March 2026), - Reuters, BlackRock fund limits withdrawals as redemptions rattle private credit, 6 March 2026, - ZeroHedge, BlackRock's Private Credit Fund Gates Investors Again (HLEND, requests at 13.3%), 12 June 2026, - CAIA, Private Credit Redemptions, Defaults, and Wrappers, Oh My! (distinction requests / actual redemptions), 20 April 2026, - Financial Stability Board, Report on Vulnerabilities in Private Credit, 6 May 2026, and press release - Federal Reserve, Supervision and Regulation Report, June 2026 (NDFI exposures, collateral management), - Office of Financial Research, Measuring Counterparty Exposures to Private Credit, brief 26-02, 12 March 2026, - Federal Register / FSOC, Authority To Require Supervision and Regulation of Certain Nonbank Financial Companies (consultation closed 14 May 2026), - Bloomberg, Apollo Wraps Up $35 Billion Debt to Buy AI Chips for Anthropic, 5 June 2026, ; Axios, - Reuters, US private credit defaults to ease in 2026 but fragility to persist, says BofA, 9 December 2025 (forecast 4.5%, Neha Khoda quote), ============================================================================ ANALYSIS: When $110 billion of collateral isn't enough to borrow $6 billion URL: https://l0g.fr/en/analysis/softbank-margin-loan/ Canonical French source: https://l0g.fr/posts/softbank-margin-loan/ Date: 2026-07-14 (reviewed 2026-08-06) Topics: AI, Private credit, Finance ---------------------------------------------------------------------------- ... And it's a worrying symptom for all of BigTech --- Update, 6 August 2026. SoftBank has since disclosed a $10bn loan agreement entered into on 5 August for a Vision Fund 2 vehicle. The public document provides for a cash-collateral account and mandatory prepayments if, among other things, the fair value of OpenAI preferred shares falls significantly. It does not establish that this facility is the same transaction as the talks reported below, or that June’s stalled talks were a general refusal to monetise the OpenAI stake. Our narrower, updated analysis is here: OpenAI’s private valuation becomes a liquidity risk. On Wednesday 10 June, Bloomberg revealed that SoftBank's talks to raise at least $6 billion via a margin loan backed by its stake in OpenAI had stalled. The stock fell as much as 9.7% in Tokyo. From afar, a funding hiccup. Up close, one of the coldest signals the credit market has sent since the start of the AI cycle. Let us analyse this refusal Timeline: on 23 April, Bloomberg reports that SoftBank is seeking $10 billion over 2 years with a one-year extension option, at about SOFR + 425 basis points, around 7.9%. In May, faced with some creditors' reluctance over OpenAI's valuation, the target is cut 40%, to $6 billion. About $5 billion of commitments are gathered, without knowing whether they were firm or verbal. Then the talks stall. Bloomberg specifies that the reason for the blockage is not known and that the file could come back later. The collateral: about 13% of OpenAI, the fruit of a cumulative investment of about $64.6 billion once the last payment is closed. At the price of the latest round, closed on 31 March at $852 billion post-money valuation for $122 billion raised, figures announced by OpenAI itself, this stake is worth notionally some $110 billion. The exact perimeter of the pledge has not been made public, but the order of magnitude speaks: $6 billion sought against a $110 billion notional position, a loan-to-value below 6% if the whole position served as collateral. And the banks said no. When a lender refuses an LTV of this order, it is not saying the asset is a little overvalued. It is saying its recoverable value, in case of trouble, is close to zero. A collateral that is not one OpenAI shares tick every box of bad collateral. Unlisted: no continuous price, no liquidity to sell them in a margin call. Their valuation comes not from a market but from private rounds. And who co-leads these rounds? SoftBank, which entered at a $300 billion valuation in April 2025 per Reuters, then co-led the $852 billion round of March 2026, with $30 billion invested. In one year, the borrower nearly tripled the price of its own collateral by injecting its billions into it. Any credit committee spots the circularity in thirty seconds. A detail noted by Bloomberg: of the $50 billion promised by Amazon in that same round, $35 billion is conditional on an IPO or the reaching of AGI. The loan's motivation completes the picture: to finance the continuation of the group's AI offensive, OpenAI commitments included. Borrowing against your OpenAI shares to keep buying OpenAI is leverage stacked on a self-sustaining valuation, whose guarantee is the very object of the bet. The symptom no one wants to see The equity market, for its part, still prices the dream: on 1 June, SoftBank gained 14% in the session, passed ¥48 trillion of market cap and dethroned Toyota as Japan's most valuable stock, a first since 2003. On 4 June, the stock lost 11.3%, about ¥5 trillion wiped out in one session. The credit market, less lyrical, has just rendered its own verdict: the private marks of AI are not monetisable. Yet the whole edifice rests on it. To honour its $22.5 billion tranche to OpenAI at the end of 2025, SoftBank sold its entire Nvidia position, $5.8 billion, sold $4.8 billion of T-Mobile and froze most Vision Fund deals, any ticket above $50 million requiring the personal sign-off of Masayoshi Son, Reuters reported in December. The group's standalone interest-bearing debt reportedly reaches ¥16.3 trillion at the end of 2025, about $104 billion per S&P, which placed the BB+ rating on negative outlook in March. Above all, a record $40 billion bridge loan, this one unsecured, announced on 27 March with JPMorgan, Goldman Sachs, Mizuho, SMBC and MUFG, matures in March 2027. Its 12-month maturity says everything: the lenders are betting on an OpenAI IPO within the year. The confidential filing was reported in May by CNBC and the WSJ, with a listing targeted as soon as September 2026 above $1 trillion. There is the mechanics laid bare: banks carry $40 billion of unsecured exposure whose repayment depends on an IPO, while refusing to advance $6 billion against the shares of the same company with a massive margin of safety. The two positions are coherent only on one condition: that the market window opens on schedule and at the hoped price. If the IPO slips, or validates a price far below the $852 billion of the latest round, the phantom collateral and the very real debt will find themselves face to face. Crisis historians will recognise the pattern. Japan in 1989 lent against land whose price could not fall. 2007 securitised super-senior tranches rated AAA by construction. In both cases, the collateral was worth its price as long as no one asked to convert it into cash. SoftBank has just asked. The answer came. Let us add, still per Bloomberg, that some SoftBank executives worry internally about the scale of the commitment, at a moment when Anthropic's progress puts OpenAI's competitive position into perspective. When doubt reaches even the main shareholder, the price of private rounds becomes a fragile convention. What this refusal really says The consensus will want to see a technical incident, merely postponed. It is possible. But the simplest reading remains the most disturbing: faced with the most celebrated private asset on the planet, the banks preferred to abstain. They have just quietly drawn the border between a valuation and a price. All the AI capex, its hundreds of billions of debt and cross-financing, is built on the wrong side of that border. We will soon know whether June 2026 was a hiccup or a warning. Sources 1. Bloomberg, 10 June 2026, blockage of the $6bn margin loan, $5bn of commitments, internal doubts, Anthropic: https://www.bloomberg.com/news/articles/2026-06-10/softbank-s-attempt-to-get-6-billion-openai-margin-loan-stalls 2. Bloomberg, 23 April 2026, initial search for $10bn, 2-year + 1 structure: https://www.bloomberg.com/news/articles/2026-04-23/softbank-seeks-10-billion-margin-loan-backed-by-openai-shares 3. Bloomberg, 31 March 2026, OpenAI round of $122bn at $852bn, Amazon $50bn of which $35bn conditional, Nvidia and SoftBank $30bn each: https://www.bloomberg.com/news/articles/2026-03-31/openai-valued-at-852-billion-after-completing-122-billion-round 4. OpenAI, statement of 31 March 2026, $122bn raised, $852bn post-money: https://openai.com/index/accelerating-the-next-phase-ai/ 5. CNBC, 31 March 2026, SoftBank co-leads the round: https://www.cnbc.com/2026/03/31/openai-funding-round-ipo.html 6. Bloomberg, 27 March 2026, unsecured $40bn bridge loan, March 2027 maturity, bank syndicate: https://www.bloomberg.com/news/articles/2026-03-27/softbank-secures-record-40-billion-bridge-loan-for-openai-stake 7. Reuters, 19 December 2025, $22.5bn tranche, Nvidia sale $5.8bn, T-Mobile $4.8bn, Vision Fund freeze, entry at $300bn valuation in April 2025 (accessible copy): https://www.marketscreener.com/news/softbank-races-to-fulfill-22-5-billion-funding-commitment-to-openai-by-year-end-sources-say-ce7d50ddd089f027 8. Kyodo via Japan Today, 1 June 2026, market cap above ¥48 trillion, Toyota overtaken, +14% in the session: https://japantoday.com/category/business/update1-softbank-overtakes-toyota-to-become-japan's-most-valuable-company 9. CNBC, 4 June 2026, stock down more than 11%: https://www.cnbc.com/2026/06/04/softbanks-shares-are-down-10percent-amid-broader-tech-sell-off.html 10. The Next Web, 8 May 2026, SOFR + 425 bps rate, notional value of about $110bn, stake of about 13%, cumulative investment of $64.6bn, S&P BB+ rating negative outlook: https://thenextweb.com/news/softbank-10b-margin-loan-openai-stake-collateral 11. S&P Global Ratings, standalone debt of about ¥16.3 trillion (~$104bn) at the end of 2025, cited by TradingKey, 5 June 2026: https://www.tradingkey.com/analysis/stocks/more/261946147-softbank-openai-arm-softbankstockpriceplummeted 12. CNBC and WSJ via Investing.com, OpenAI confidential IPO filing, listing targeted as soon as September 2026 above $1 trillion: https://www.investing.com/analysis/the-trilliondollar-ipo-test-spacex-and-openai-face-public-markets-200680688 ============================================================================ ANALYSIS: Chinese real estate: anatomy of a risk that settles in URL: https://l0g.fr/en/analysis/chinese-real-estate-risk/ Canonical French source: https://l0g.fr/posts/immobilier-chinois-anatomie-d-un-risque-qui-s-installe/ Date: 2026-07-14 (reviewed 2026-07-14) Topics: china, real estate, risk, macro, debt ---------------------------------------------------------------------------- import InfographicBars from '../../components/infographic/InfographicBars.astro'; Chinese real-estate risk has changed nature. The spectacular phase, that of resounding defaults and stalled construction sites, has given way to a slower and more diffuse sagging, less photogenic but just as heavy in consequences. In 2026, the question is no longer whether the sector will correct, it has been correcting since 2021, but to measure where the risk now lodges and how fast it propagates to the rest of the economy. A weight that recedes but stays central First thing to frame: the size. Real estate was long the engine of Chinese growth, and it remains so by its very mass even in decline. Its contribution to GDP went from about 24% in 2018 to nearly 19% in 2024, an important decline but one that leaves the sector at the heart of the economy. Above all, property concentrates around 70% of Chinese households' gross wealth, a proportion that makes the price of housing a variable of national wealth, not just a sector indicator. This double characteristic, a weight that falls but stays enormous and household savings massively backed by property, explains why the correction can be neither ignored nor rushed. Housing prices have fallen about 30% from their 2021 peak per market estimates, a purge of a rare scale for an asset held by hundreds of millions of households. The early-2026 drop The official data of early 2026 confirm that the fall has found no floor. Over the first two months of the year, the National Bureau of Statistics measures real-estate investment down 11.1% year on year, at ¥961 billion. Residential home sales fall 21.8% by floor area, and housing starts 23.1%, a sign that developers are not restarting the machine. New-home prices fall a further 3.2% year on year in February, their sharpest fall in eight months. These figures describe a well-identified vicious circle: weak sales dry up developers' cash, they slow construction, which erodes buyer confidence and weighs further on sales. The Chinese specificity aggravates the mechanism, because a large share of homes were sold off-plan, before construction. When a developer wobbles, it is households who have already paid who risk never receiving their home, hence the priority Beijing gives to completing already-sold projects. Developers on life support The wave of defaults opened by Evergrande has not spared the players reputed to be the most solid. China Vanke, long considered the reference developer, narrowly avoided a default on $284 million of debt, with more than ¥9.4 billion of bonds maturing over six months and revenue down 27% year on year in the third quarter. Country Garden, once the country's top builder by sales, continues its debt restructuring. The risk here is not only that of an isolated bankruptcy. A default by a player perceived as a survivor would send a signal to creditors across the whole sector, making refinancing even harder for already-fragile private developers. It is this risk of contagion through confidence, more than through balance sheets, that worries. The mechanism is not specific to China: it recalls the dynamic described in our analysis of the silent contagion of private credit, where the weak link is not the biggest but the most exposed to a turn in sentiment. The hidden link: local governments The most underestimated channel passes through local finances. For two decades, Chinese local governments financed their development by selling land-use rights, through ad hoc vehicles, the LGFVs, which borrowed by pledging this land and repaid thanks to land sales. The real-estate collapse dried up this source: state land-rights sales fell back to around ¥4,150 billion, about $601 billion, down 14.7%. Yet the debt accumulated by these vehicles is colossal, and its exact scale is debated, which is in itself a risk factor. The central-bank governor, Pan Gongsheng, puts LGFV operational debt at about ¥14,800 billion, when the International Monetary Fund uses a far broader measure, on the order of ¥58,000 billion, nearly half of Chinese GDP. This gap measures everything that stays off balance sheet. Beijing reacted at the end of 2024 with a ¥12,000 billion debt-swap plan, about $1,700 billion, meant to convert hidden local debt into better-monitored official debt. Hidden debt thus reportedly fell back to ¥10,500 billion at the end of 2024, and the number of LGFVs on the official list shrank 71% between March 2023 and September 2025. The procedure stabilises liquidity and pushes out maturities, but it moves the problem rather than erasing it, a point documented in our reading of non-bank financial intermediation. An overhang of unsold homes that caps the recovery Even in the event of a demand rebound, the accumulated supply would weigh on prices for years. At the end of 2025, the time to clear new homes in the top 100 cities reached 27.4 months, well above the 12-to-14-month range judged healthy. The stock of unsold homes would represent, if fully built, on the order of ¥93,000 billion, about $13,000 billion, a figure to be taken as an order of magnitude of the overhang rather than a market value. Faced with this excess, 2026 policy has changed axis. Rather than massively stimulating demand, Beijing seeks to reduce supply: control new launches, buy back unsold homes to convert them into social housing, and prioritise completing already-sold projects. The logic is assumed, it is an organised landing, not a stimulus. S&P Global Ratings accordingly anticipates new-home prices still falling 1.5 to 2.5% in 2026, and existing-home prices 4 to 5%, with one to two more years before a trough. The other reading: a managed decline, not a Lehman For the sake of objectivity, the opposite thesis must be carried, because it is solid. The analogy with the Lehman Brothers bankruptcy, often brandished, holds up poorly under examination. Three differences matter. First, the correction is partly intended. In 2020, Beijing deliberately tightened credit to developers with the "three red lines", precisely to deflate the bubble. The deflation is painful, but it is not accidental. Second, Chinese debt is financed mostly by domestic savings and intermediated by state banks, in a system partly closed by capital controls. The risk of brutal international contagion, the core of the 2008 shock, is therefore limited. Finally, the state has levers a market economy does not: it can recapitalise banks, direct credit and spread losses over time. This reading is right, and it bounds the catastrophe scenario. But it has a limit its proponents sometimes forget: if the state can prevent the collapse of banks, it cannot compel households, companies and investors to durably consider an overvalued home a safe store of value. Management avoids the crash; it does not invent demand. The risk, in chronic version Chinese real-estate risk in 2026 is therefore not that of a sudden blast, but of a chronic pressure exerted through three measurable channels. The first is wealth-related: with 70% of their wealth in property, households who see their home depreciate reduce their consumption, which feeds deflationary pressures and weakens domestic demand. The second is fiscal: deprived of land revenue, local governments cut spending and struggle to service their debt, transferring the strain from the developer to the local state. The third is financial: regional banks and the non-bank intermediation system carry a real-estate and local exposure whose quality slowly deteriorates. None of these channels produces a spectacular moment. Together, they describe a durable brake on Chinese growth, with repercussions beyond the borders, on commodity demand in particular, as the fall in crude imports illustrates. The real risk is not that China has its 2008. It is that it has a long, grey version of real-estate stagnation, the one hardest to exit precisely because it never forces the decision. Sources 1. China National Bureau of Statistics (NBS), real-estate investment down 11.1% to ¥961.2 billion, residential home sales down 21.8%, housing starts down 23.1%, new-home prices down 3.2% year on year in February, January-February 2026: https://www.stats.gov.cn/english/PressRelease/202603/t202603171962803.html 2. Bloomberg, persistent difficulties of China Vanke, Country Garden and the sector, real-estate weight from about 24% of GDP in 2018 to about 19% in 2024: https://www.bloomberg.com/news/articles/2026-03-27/china-vanke-country-garden-navigate-persistent-property-headwinds 3. ABC News / Associated Press, Vanke narrowly avoids a default on $284 million, ¥9.4 billion of bonds maturing over six months, revenue down 27%: https://abcnews.go.com/Business/wireStory/china-vankes-default-exposes-fragility-faltering-recovery-property-128798090 4. Atlantic Council, LGFV debt, divergent estimates (¥14.8 to ¥58 trillion), collapse of state land sales to around ¥4,150 billion ($601 billion, -14.7%): https://www.atlanticcouncil.org/blogs/econographics/beijing-extends-and-pretends-to-deal-with-its-mountain-of-local-government-debt/ 5. Caixin Global, ¥12,000 billion debt-swap plan at the end of 2024, hidden debt down to ¥10,500 billion, number of LGFVs down 71% from March 2023 to September 2025: https://www.caixinglobal.com/2026-05-29/in-depth-as-chinas-hidden-local-debts-shrink-a-new-challenge-emerges-102449016.html 6. IMF, conclusion of the 2025 Article IV consultation with China (February 2026), real-estate and local-debt risks: https://www.imf.org/en/news/articles/2026/02/18/pr-26053-china-imf-executive-board-concludes-2025-article-iv-consultation 7. Caixin Global, effort to offload unsold homes, clearance time of 27.4 months in the top 100 cities, priority to social housing: https://www.caixinglobal.com/2025-12-15/china-ramps-up-effort-to-offload-vast-supply-of-unsold-homes-102393476.html 8. S&P Global Ratings, 2026 forecasts: new-home prices down 1.5 to 2.5%, existing down 4 to 5%, one to two years before a trough, unsold stock of about ¥93,000 billion: https://www.spglobal.com/ratings/en/regulatory/article/china-property-watch-supply-glut-to-impede-recovery-s101667227 9. NPR / Associated Press, Lehman comparison put in perspective, debt financed by domestic savings, the 2020 "three red lines" scheme: https://www.npr.org/2024/01/30/1227554424/evergrande-china-real-estate-economy-property-collapse 10. South China Morning Post, debate on a possible "Lehman moment", limits of state management against household distrust: https://www.scmp.com/economy/china-economy/article/3231900/chinas-property-crisis-plagues-its-economy-and-financial-system-lehman-moment-looming 11. Global Property Guide, history of Chinese housing prices, a fall of about 30% from the 2021 peak: https://www.globalpropertyguide.com/asia/china/price-history ============================================================================ ANALYSIS: Commercial real estate: the 2026 refinancing wall, and the regional link URL: https://l0g.fr/en/analysis/commercial-real-estate-refinancing-wall/ Canonical French source: https://l0g.fr/posts/immobilier-commercial-mur-refinancement-2026-maillon-regional/ Date: 2026-07-14 (reviewed 2026-07-14) Topics: real estate, banks, credit, risk, markets ---------------------------------------------------------------------------- Some risks explode, others settle in. US commercial real estate is the second type: not a sudden crash, but a wall borrowers have seen coming for months and yet will have to cross. Close to a trillion dollars of debt matures in 2026, taken out when money cost 3 to 4%, to be refinanced today at 6 or 7%, against buildings that have lost a fifth to two-fifths of their value. The epicentre is the office, emptied by remote work. And transmission to the financial system runs through a precise link, the regional banks, which have tripled their bets on bricks and mortar. This piece dismantles the mechanics of this slow risk, and weighs its two readings, the one that judges it contained and the one that fears it contagious. The wall, in figures The starting point is a timing problem. A large part of commercial real-estate debt, CRE, is short-dated and must be refinanced regularly. Yet the loans taken out during the free-money years all mature at the same time, forming what the trade calls a maturity wall. Per the estimates, between $875 billion and close to a trillion dollars of CRE loans must be repaid, refinanced or extended in 2026 alone, the peak expected in 2027. On the securitised segment alone, CMBS, about $77 billion face so-called hard maturities, with no extension option. The danger is not the volume alone, it is its combination with the rate shock. A building financed at 75% of its value, at a rate of 3 to 4%, refinances today at 6 or 7%, on a value that has fallen, and often at an LTV cut to 55 or 60%. The borrower must then find the difference in fresh equity, sometimes tens of percent of the debt, on pain of handing back the keys. As long as the Fed holds high rates, this wall stays high; its height depends directly on the rate trajectory, which makes it a risk suspended from monetary policy. The epicentre: the office It would be wrong to treat commercial real estate as one block. The divergence between segments is the most important fact of this cycle. Data centers, carried by artificial intelligence, and logistics are doing well; multifamily and retail are under moderate strain. The epicentre of the risk is the office, hit by a structural rather than cyclical change: remote work has durably reduced demand for space. Vacancy rates run around 19 to 20%, and values have fallen 20 to 40% from their peak, more in some submarkets. The figures of already-materialised stress are severe. Among office loans that matured and remained unresolved, more than 80% are in default, and nearly 93% have gone into special servicing, the procedure reserved for troubled loans. The office is therefore not a risk to come, it is a claim in progress, whose scale for the system depends on who holds these loans. The answer brings us back to the banking core. The weak link: regional banks Here is the transmission that turns a real-estate problem into financial risk. The bulk of CRE debt is not securitised, it sleeps on banks' balance sheets, and very unevenly. Regional and community banks, those under $10 billion of assets, are over-exposed: they have nearly tripled their commercial-real-estate loans in a decade, and collectively hold more than $1,600 billion of them. The figure that should raise the alarm is not the volume, it is the concentration. When the national average CRE-to-capital ratio is around 30%, the median ratio of regional banks reached 312% at the end of 2024, and 54.8% of them exceeded the 300% threshold beyond which the regulator imposes enhanced supervision. A Wharton study sums up the peril in a phrase, that of banks "too many to ignore": none is systemic alone, but all carry the same risk at the same time. It is exactly the mechanism of the 2023 panic, which we described in our piece on regional banks and liquidity reform, and that is why this risk must be read with the grid of our bank-health guide. Extend and pretend, the art of buying time How is it, then, that the system has not cracked yet? Through a practice as old as credit, extend and pretend. Rather than recognising the loss on a loan a borrower cannot refinance, the bank extends the maturity and acts as if the problem were solved. This avoids booking the loss, mobilising capital and alarming the market. Above all, it postpones the day of truth, in the hope that rates fall or values recover by then. This mechanism is a direct cousin of the restructurings that mask defaults in leveraged credit, described in our CLO guide. It has a virtue, avoiding a wave of foreclosures that would collapse prices at once, and a vice, sustaining an accounting fiction that hides the real scale of the stress. The displayed default rate on commercial real estate thus understates true distress, exactly like the corporate-credit default rate. Extension buys time; it creates no value, it bets that time will. The private-credit channel A second channel now doubles the banks': private credit. As regional banks turned cautious after 2023, private-debt funds rushed into commercial-real-estate financing, including on the riskiest assets. The risk does not disappear, it moves toward less transparent and less regulated vehicles, the non-bank financial institutions that we made the through-line of our second-quarter bank-earnings preview. The loop is the one we keep finding in credit plumbing. Banks finance the private-credit funds, which finance commercial real estate; the exposure leaves the bank balance sheet through the direct-lending door and comes back through wholesale funding. It is the silent contagion of private credit applied to bricks and mortar, and it makes mapping the risk harder, because a real-estate loss can now strike where you do not look for it, in a semi-liquid fund rather than in a listed bank. Contained or contagious There remains the question that decides everything, or at least to weigh it: is this risk contained or contagious? Both readings have their arguments, and honesty commands laying out both. The contained-risk thesis is solid. Real-estate losses materialise slowly, loan by loan, and not in a single shock; the office represents only a fraction of the stock and of balance sheets; extension and an eventual rate cut can give values time to recover; and the largest banks showed, in June's stress test, that they would absorb a severe shock. Several analysts judge, moreover, that the worries are easing, a few idiosyncratic pockets aside. The opposite thesis is no less supported. The concentration of regional banks is a fact, not a hypothesis, and the history of 2023 showed how fast one of these banks can crack. The refinancing wall is dated and quantified, and as long as the Fed stays restrictive, as a dot plot leaning toward a hike and sticky inflation suggest, refinancing happens at the worst rate. Extension resolves nothing, it defers, and private credit makes the real exposure harder to map. Our reading, measured, is that the deciding variable is the rate trajectory: at durably high rates, the time bought by extension worsens the wound instead of closing it. The points to watch To follow this risk without yielding to either panic or denial, a few dials. The CMBS default rate, particularly on offices, gives the temperature of already-materialised stress. CRE concentration ratios relative to capital, bank by bank, say where the risk is lodged. The maturity-wall calendar, year by year, to the 2027 peak, indicates the pace of the ordeal. The signs of extension, loan modifications and prolongations, reveal the scale of what balance sheets do not show. Finally, the valuation marks of private-credit funds exposed to real estate, when they filter through, say whether the risk moved off the banks is correctly priced. And above all, the Fed's trajectory, which sets the height of the wall. A risk that settles in US commercial real estate is not the subprime of 2008, and repeating it is useful: losses there materialise slowly, the system digests them in small doses, and nothing there resembles the correlated securitisation that blew up the world. But it is also no non-event. It is a risk that settles in, as we said of Chinese real estate in another register, concentrated in a precise link of the banking system, sustained by an accounting fiction, and suspended from the one variable no one really controls, the level of rates. It will probably not make the front page of a crash Monday. It will weigh, quarter after quarter, on the solidity of dozens of regional banks, until rates fall and relieve it, or a link gives way and reveals it. To watch it is to accept following a risk that does not have the courtesy to explode. Sources 1. S&P Global Market Intelligence, commercial-real-estate maturity wall ($950bn in 2024, peak in 2027): https://www.spglobal.com/market-intelligence/en/news-insights/research/commercial-real-estate-maturity-wall-950b-in-2024-peaks-in-2027 2. CoStar, "Why commercial property pros say a looming $1.26 trillion debt wall can be scaled": scale of the wall and rate shock at refinancing: https://www.costar.com/article/1122236114/why-commercial-property-pros-say-a-looming-1-26-trillion-debt-wall-can-be-scaled 3. CRE Daily, "Maturing Debt Drives 2026 CRE Distress" and "CMBS Maturity Wall Tests Refinancing in 2026": office defaults (80%), $77bn of hard-maturity CMBS: https://www.credaily.com/briefs/maturing-debt-drives-2026-cre-distress/ 4. BankHealthData, "Commercial Real Estate Bank Risk 2026": CRE concentration of regional banks (median ratio 312%, 54.8% above 300%), $1,600bn on balance sheet: https://www.bankhealthdata.com/blog/commercial-real-estate-bank-risk-2026 5. Wharton (F. Hinzen, S. Van Nieuwerburgh et al.), "Too-Many-to-Ignore: Regional Banks and CRE Risks": the systemic risk of regional concentration: https://wifpr.wharton.upenn.edu/wp-content/uploads/2025/10/HSV-Regional-Banks-and-CRE-Risks.pdf 6. Federal Reserve, 2026 stress-test scenario: 39% fall in commercial-real-estate prices, about $75bn of losses: https://www.federalreserve.gov/publications/2026-stress-test-scenarios.htm 7. l0g, Regional banks and liquidity reform, The silent contagion of private credit, guides Bank health and CLOs and leveraged loans. ============================================================================ ANALYSIS: Uranium: a market in deficit, the promise of AI, and hidden bottlenecks URL: https://l0g.fr/en/analysis/uranium-market-deficit-ai-bottlenecks/ Canonical French source: https://l0g.fr/posts/uranium-marche-en-deficit-promesse-de-l-ia-et-goulots-caches/ Date: 2026-07-14 (reviewed 2026-07-14) Topics: uranium, nuclear, ai, commodities, risk ---------------------------------------------------------------------------- import InfographicBars from '../../components/infographic/InfographicBars.astro'; Uranium has become, in a few quarters, one of the most talked-about commodities. The story is seductive: structurally growing demand, driven by the nuclear revival and now by the electricity thirst of artificial intelligence, against a supply slow to respond. This story is largely well founded. But a commodity market is never read through demand alone, and uranium hides, behind the ore, a far more binding bottleneck: enrichment. Let us lay out the thesis, then its antithesis, with the figures. For the fundamentals, see our guide on reading the uranium market. A market already tight before AI The price signal leaves little doubt about the tension. The spot price jumped about 25% in January 2026 to move back above $100 a pound of uranium oxide (U3O8), a first in two years. Above all, the long-term price, the one that matters for reactor operators' contracts, reached $93 a pound at the end of March 2026 according to TradeTech's indicator, its highest level in more than eighteen years. Investor surveys point to a $100 to $120 range, with expectations up to $135. Behind the price, a fundamental imbalance. World mine production stands around 60,000 tonnes of uranium a year, when reactor needs already approach 69,000 tonnes. The gap is filled by secondary supply, that is, inventories, reprocessing and various non-mining sources. Yet demand is on a steep upward path: the World Nuclear Association projects installed nuclear capacity rising from 398 gigawatts electric mid-2025 to 746 in 2040, and uranium requirements climbing from about 68,900 tonnes in 2025 to more than 150,000 in 2040. At the same time, 78 reactors are under construction worldwide, China in the lead with about 38 units. AI plugs nuclear back in It is against this already-tight backdrop that the electricity demand of artificial intelligence arrives. According to the International Energy Agency, world data-centre consumption would rise from 415 terawatt-hours in 2024 to 945 in 2030, a hyperscale data centre dedicated to AI potentially demanding 300 to 500 megawatts, the equivalent of a mid-sized city. This electricity, the tech giants want it dispatchable and decarbonised, and they have turned to nuclear. The announcements have multiplied. By mid-2026, every major hyperscaler had signed at least one nuclear deal: in total, thirteen announced projects commit more than 9.8 gigawatts of capacity to power AI. Microsoft secured a twenty-year, $16bn power-purchase agreement for the restart of Three Mile Island's unit 1, some 835 megawatts expected around 2027. Meta targets up to 6.6 gigawatts by 2035, relying on advanced reactors from TerraPower and Oklo and existing plants from Vistra and Constellation. Google signed with Kairos Power and NextEra to restart a plant in Iowa. Amazon is developing small modular reactors (SMRs) with X-energy and Energy Northwest. A point of method imposes itself here, because it conditions the whole reasoning. These deals cover electricity delivered between 2027 and 2035, via restarts, existing plants or small reactors still to be built. The impact on physical uranium demand is therefore real but deferred, and it concerns the fuel as much as its transformation. AI's promise supports the long-term story; it does not create an immediate call on uranium. A concentrated supply, slow to respond Facing this demand, supply has two flaws: it is concentrated and it responds slowly. The world's leading producer, Kazakhstan's Kazatomprom, and Canada's Cameco together supply more than 40% of production, Kazakhstan alone weighing about 40% of world extraction. This concentration is a risk factor in itself. Kazatomprom has moreover lowered its ambitions: its 2026 production is guided between 27,500 and 29,000 tonnes on a 100% basis, below its nominal capacity, owing notably to a shortage of sulphuric acid, a key input for in-situ leach mining. Opening or restarting a mine takes years and a lot of capital. In the United States, a few projects are restarting, such as Uranium Energy Corp's Burke Hollow in Texas or Ur-Energy's Shirley Basin in Wyoming, but domestic production remains marginal. Over the longer term, supply eventually responds to price, which tempers the most bullish scenarios. In the short term, it does not: the World Nuclear Association notes that rising demand runs into a supply base that is little elastic to current price levels. Notably, the geography of supply is also recomposing, Kazakhstan steering a growing share of its sales toward China, Russia and India, via long-term contracts. The bottleneck the market underestimates: enrichment The real point of fragility is not always where one looks. Extracting uranium is not enough: it must be converted then enriched before it becomes fuel. Yet enrichment is an oligopoly where Russia holds a dominant place, with about 44% of world capacity. The United States depends heavily on it: Russia supplied about 35% of its enriched-uranium imports, and US operators bought 4,141 thousand separative work units (SWU) from it in 2023. Washington has decided to sever this link. The law banning Russian uranium imports, enacted in May 2024, will take full effect in 2028, with waivers possible until then. It unlocks $2.72bn to build a domestic supply chain. The problem is acute for advanced reactors, which require a more enriched fuel, HALEU, of which Russia was until 2024 the only commercial supplier. In the United States, only Centrus currently produces HALEU, at small scale, with more than 900 kilograms delivered to the Department of Energy and a $900m contract signed on 1 July 2026. In other words, the AI-driven nuclear revival could stumble less on the ore than on the capacity, to be rebuilt, to enrich it outside Russia. The antithesis: why the story could disappoint Rigour demands taking the objections seriously, because uranium is a market with a cyclical and volatile history, prone to bouts of enthusiasm as much as to disillusion. Several factors invite caution. First, the mining deficit is not an immediate shortage. It has been filled for years by secondary supply, and operators long preferred to draw on their inherited inventories rather than contract anew. This cushion delays the reckoning. Next, AI demand is largely a story of the 2030s. The commercialisation of small reactors has fallen behind, a flagship project having been abandoned in the United States in 2023, and the first SMRs would not be operational before 2030 or 2031. The tech giants' deals deliver electricity, not uranium tomorrow. To this is added an execution risk on the producer side, with licensing delays and forecast revisions on several projects, and a precedent that should breed humility: after Fukushima in 2011, nuclear demand collapsed and uranium spent a decade in lethargy. Finally, AI demand itself rests on spending expectations we have learned to view with caution, from the real productivity of AI to the risk of a bubble in valuations. If AI investment slows, part of the promised electricity demand evaporates with it. It should be noted, lastly, that some of the bullish talk comes from actors selling uranium-exposure products, which invites distinguishing analysis from the sales pitch. Outlook, and the signals to watch A balanced reading emerges. The uranium market is structurally tight, with credible rising demand and a concentrated, rigid supply, which argues for durably firm prices. AI demand reinforces this story, but on a 2030 horizon more than the immediate one, and the most binding bottleneck sits at enrichment, not the mine. The risk is not so much that the thesis is wrong as that its direction, probably right, gets confused with its timing, highly uncertain. For anyone wanting to follow this commodity, a few indicators concentrate the information. TradeTech's long-term price, more telling than the spot, says operators' conviction. The pace of utility contracting signals the end of the inventory cushion. Kazatomprom's and Cameco's guided production give the pulse of supply. The milestones of restarts and SMRs, along with the ramp-up of enrichment outside Russia, notably HALEU, will say whether the promise materialises. And the trajectory of AI spending will remain the arbiter of the expected electricity demand. It is this whole, more than any single figure, that deserves sustained watching. Sources 1. World Nuclear Association, uranium markets, supply and demand: installed capacity from 398 GWe mid-2025 to 746 GWe in 2040, requirements rising from about 68,900 to more than 150,000 tonnes of uranium, supply little elastic at current prices: https://world-nuclear.org/information-library/nuclear-fuel-cycle/uranium-resources/uranium-markets 2. TradeTech, uranium price indicators: long-term price at $93.00 a pound of U3O8 at end-March 2026, highest in more than eighteen years: https://www.uranium.info/pressreleases.php 3. Sprott, "Uranium Outlook 2026": spot jump above $100 in January 2026, contracting dynamics, structural tension (source from an issuer of uranium products, to be read as such): https://sprottetfs.com/insights/uranium-outlook-2026/ 4. International Energy Agency, data-centre electricity consumption from 415 TWh in 2024 to 945 TWh in 2030, hyperscale data-centre needs: https://www.iea.org/reports/energy-and-ai 5. smrintel, census of tech giants' nuclear deals: thirteen projects, more than 9.8 GW for AI, Microsoft, Meta, Google, Amazon detail: https://smrintel.com/nuclear-data-center-deals/ 6. Forbes, 19 February 2026, Microsoft and Amazon turn to nuclear for AI, $16bn contract for Three Mile Island: https://www.forbes.com/sites/rrapier/2026/02/19/why-microsoft-and-amazon-are-turning-to-nuclear-power-for-ai/ 7. World Nuclear News, 2025 production results for Cameco and Kazatomprom, Kazatomprom's 2026 guidance (27,500 to 29,000 tonnes, sulphuric-acid constraint): https://world-nuclear-news.org/articles/Cameco-Kazatomprom-release-2025-figures 8. World Nuclear Association, uranium and nuclear power in Kazakhstan, the country's share of world production: https://world-nuclear.org/information-library/country-profiles/countries-g-n/kazakhstan 9. U.S. Department of Energy, ban on Russian uranium imports and development of the domestic supply chain, $2.72bn: https://www.energy.gov/ne/articles/russian-uranium-ban-will-speed-development-us-nuclear-fuel-supply-chain 10. Nuclear Regulatory Commission, context of the ban: Russia at about 44% of world enrichment capacity, about 35% of US imports, 4,141 thousand SWU bought in 2023: https://www.nrc.gov/reading-rm/doc-collections/fact-sheets/uranium-import-ban 11. Utility Dive, law banning Russian uranium, full effect in 2028, $2.7bn unlocked for HALEU and enrichment: https://www.utilitydive.com/news/congress-passes-russian-uranium-import-ban-haleu-nuclear-fuel-advanced-reactors/715256/ 12. Power Magazine, Centrus, sole US producer of HALEU, more than 900 kg delivered to the DOE, $900m contract of 1 July 2026: https://www.powermag.com/centrus-completes-900-kg-haleu-delivery-to-doe-in-u-s-nuclear-fuel-enrichment-milestone/ 13. Sprott, "Uranium's Tale of Two Markets": SMR delays (first reactors operational around 2030-2031), utilities drawing on inventories, execution risks: https://sprott.com/insights/uranium-s-tale-of-two-markets/ ============================================================================ ANALYSIS: Gilts: the market the Bank of England wants to deleverage before the next accident URL: https://l0g.fr/en/analysis/gilts-repo-leverage-bank-of-england/ Canonical French source: https://l0g.fr/posts/gilts-repo-levier-banque-angleterre/ Date: 2026-07-14 (reviewed 2026-07-14) Topics: macro, markets, liquidity, central banks, regulation, United Kingdom ---------------------------------------------------------------------------- The Bank of England is not tackling a back-office detail. In seeking to frame hedge-fund leverage in the gilt market, it is putting its finger on a more general fragility: a heavier public debt, carried by shorter, more competitive intermediation more sensitive to margin calls. The risk is not a British default tomorrow morning. It is in the liquidity of a sovereign market reputed to be deep, but financed by positions that can shrink very fast. In early July, the Financial Times and The Times report that the Bank of England is advancing on a mechanism to limit the leverage hedge funds can take in the gilt market, British sovereign bonds. The central path: impose minimum repo haircuts, in other words prevent a fund from borrowing almost the entire value of a security by pledging it as collateral. The gilt market weighs close to £3 trillion per the orders of magnitude relayed by the financial press, and hedge funds reportedly account for about 30% of activity. Their role is useful: they take the other side of flows, arbitrage the gaps between cash bonds and derivatives, and smooth the market. But this liquidity is often financed by very short repo, sometimes with haircuts close to zero. The leverage hidden in the haircut A haircut is a safety discount. If an investor pledges 100 of gilts as collateral and the lender applies a 2% haircut, it lends only 98. If the haircut falls to zero, the same collateral finances almost the whole position. The gap looks tiny; its effect on leverage is not. The Bank of England is targeting precisely this zone. Per the FT, the institution judges that competition between prime-brokerage banks and large clients can push funding conditions too far, notably through portfolio margining. A well-calibrated cross margin can reduce redundant margin calls. But in a systemic sovereign market, a zero haircut is not just a commercial price. It amounts to offering a lot of liquidity to a private trade. The mechanism resembles the Treasury basis trade, already documented on l0g: a small price difference between cash and futures becomes a big position because it is financed very cheaply. As long as markets are calm, the arbitrage helps. When volatility rises, it can reverse its role: instead of absorbing the shock, it amplifies it. Why the United Kingdom matters The United Kingdom is not an exotic case. That is precisely why it is interesting. It has a secondary reserve currency, a large financial centre, high debt, a deep bond market, and a recent memory of crisis. The 2022 LDI episode was not the same trade, but it already showed the same mechanics: margin calls, forced gilt sales and a central bank obliged to intervene to prevent a market dysfunction. The new episode is elsewhere, in hedge-fund repo. Per the elements relayed by the Guardian from the Bank of England's December 2025 Financial Stability Report, net gilt repo borrowing approached £100 billion in November 2025. A small handful of funds represented more than 90% of this net borrowing, often at very short maturities and zero or near-zero haircuts. The risk is therefore not only size. It is the combination of concentration, short maturity and leverage. The FT adds a more recent signal: during the sell-off tied to the war in Iran, the Bank of England observed a rapid deleveraging of about £19 billion, with a level remaining around £74 billion after the episode. This is not a collapse. It is a preview of the behaviour of a heavily financed trade when the market becomes less comfortable. The BIS reading: public debt and NBFIs converge The cleanest framework comes from the BIS Annual Economic Report 2026, published on 28 June. The Bank for International Settlements describes there a new link between fiscal risk and financial stability. In the old world, sovereign risk passed mainly through banks. In the new world, public-debt markets are more intermediated by NBFIs: hedge funds, open-ended funds, money market funds, insurers, market vehicles. The BIS's central figure is clear: in advanced economies, the NBFI share of sovereign-debt holding reportedly rose from 44% in 2021 to 53% in 2025. Over the same period, the share of domestic central banks in these holdings reportedly fell from 27% to 17%, and that of the foreign official sector from 15% to 13%. This shift does not say that every non-bank investor is fragile. It says the marginal buyer has become more private, more yield-sensitive, more dependent on funding conditions. This reading explains why the BoE is acting now. When states issue more, they need markets able to absorb the volumes. If central banks shrink their balance sheets and banks limit their market intermediation, hedge funds take more space. Liquidity does not disappear; it changes carrier. And that carrier can be funded overnight. The dilemma: more safety, less apparent liquidity The regulatory project is delicate. Imposing minimum haircuts reduces maximum leverage, but makes certain trades more expensive. Encouraging central clearing can make exposures more transparent, but also concentrates risk in clearing houses and raises visible margin requirements. The market fears a simple consequence: fewer hedge funds, therefore less liquidity, therefore higher borrowing costs for the state. This objection is serious. It is not enough to close the debate. A liquidity that exists only as long as haircuts stay at zero is not robust liquidity. The question is not whether hedge funds are useful. They are. The question is how much leverage a sovereign market can accept to obtain that usefulness. The BIS frames the problem in broader terms: central-bank backstops must stay temporary, targeted and reversible, otherwise they risk encouraging the very leverage they will then have to rescue. If operators believe the central bank will always intervene, they can fund shorter and bigger. If the central bank promises never to intervene, a technical shock can become a macro shock. Between the two, you have to reduce the probability of needing the firefighter. The signals to watch Three signals matter now: the exact shape of the BoE's proposals, the reaction of prime-brokerage banks, and the transatlantic treatment of the same problem. The United States already has its own project on central clearing of Treasuries and repo, in the continuation of the debate on the basis trade. Europe is also looking at margins on sovereign repos. This is therefore not a British story. It is the same problem in several currencies: how to finance more-indebted states with less-banked markets, without turning sovereign debt into a margin-call machine. A gilt remains a British sovereign bond, not an exotic emerging debt. But modern risk is not always in the issuer. It is in the way the asset is financed, rehypothecated, arbitraged and sold when everyone reduces their balance sheet at the same time. The Bank of England is trying to remove a little leverage before the market does it itself. In one case, the deleveraging is negotiated. In the other, it arrives under margin call. --- Primary sources: Financial Times, "Bank of England to push ahead with plan to limit hedge fund leverage", 2 July 2026; The Times, "Bank of England to limit debt that hedge funds can use to buy gilts", 2 July 2026; BIS, Annual Economic Report 2026, 28 June 2026, chapters I and II on the fiscal-financial link, NBFIs and repo funding; The Guardian, summary of the Bank of England's Financial Stability Report, 2 December 2025. Figures relayed with their perimeter: hedge-fund activity in gilts, net repo borrowing, deleveraging observed during the Iranian sell-off, BIS sovereign-holding shares. This is not investment advice. ============================================================================ ANALYSIS: No, rising French rates do not signal France leaving the euro zone URL: https://l0g.fr/en/analysis/french-rates-no-frexit/ Canonical French source: https://l0g.fr/posts/taux-francais-pas-de-frexit/ Date: 2026-07-14 (reviewed 2026-07-14) Topics: bonds, macro, france ---------------------------------------------------------------------------- A seductive intuition, a faulty reading For a few weeks, a narrative has been gaining visibility on social media: the rise in interest rates on French debt supposedly signals that markets are "pricing" an imminent exit of France from the euro zone. The scenario, popularised by figures such as Marc Touati or François Asselineau, posits that investors anticipate the euro's explosion, the return to the franc, a wave of monetisation by the Banque de France, therefore galloping inflation and a devaluation of bonds. QED: the rise in the OAT would be the prelude to the great rupture. This reading is intellectually seductive but factually false. It confuses a classic sovereign risk premium with a pricing of exit from the monetary union, two distinct mechanisms that show up through very different indicators. When you look at the figures, the Frexit scenario is nowhere in the pricing. Here is why. The OAT rises, but not alone On 15 May 2026, the 10-year French OAT stood at 3.81%, against 3.11% for the German Bund, a spread of 70 basis points (source: Idéal Investisseur, Banque de France and Deutsche Bundesbank data). Over a year, this spread oscillated between 59 and 85 bps, for an average of 71.8 bps. The current level is therefore within the recent average, not above it. More important: the rise in French rates is part of a coordinated European and global move. The German Bund crossed 3% this week, the UK 10-year reached its highest since 2008, the US 30-year touched 5.12% on Friday 15 May (source: CNBC, 15 May 2026). If the French rise were caused by Frexit risk, you would observe a decoupling: OAT rising, Bund stable or falling (because capital would flee France for the safety of Germany). The opposite is happening. The real cause is documented by all mainstream analysts and by the central banks themselves: the Iran-Israel-United States war triggered at the end of February 2026 pushed Brent above $100, the Strait of Hormuz remains closed, and the energy shock is diffusing into inflation expectations. Christine Lagarde, ECB president, publicly acknowledged in April that "high energy costs have deflected the euro zone from its baseline economic trajectory" (Trading Economics, 16 April 2026). Markets now price at least two ECB rate hikes by the end of 2026, whereas they anticipated cuts at the start of the year. If markets priced a Frexit, the spread would be at 300 bps, not 70 This is the simplest test to run. During the 2011-2012 sovereign debt crisis, when markets actually priced a serious risk of euro-zone fragmentation, the OAT-Bund spread peaked at 225 basis points on 17 November 2011 (source: Banque de France via Idéal Investisseur, full history since 2005). For Greece, the spread against the Bund exceeded 3,000 basis points at the moment exit was seriously contemplated. The current spread at 70 bps is therefore at 31% of the 2011 peak, and nearly 45 times lower than Greece's in the depths of its existential crisis. The threshold considered an alert signal by markets is 80 bps (source: Idéal Investisseur). We are below it. The thesis that markets are pricing an exit does not survive a second of examination against the data. The mechanical drivers of a Frexit pricing are totally absent When markets genuinely price a risk of monetary rupture, several indicators move simultaneously. None does today: French sovereign CDS (the cost of insuring against a French default at 5 years) stay at moderate levels, with no exponential rise. In a Frexit pricing, they would explode, as was the case for Greece in 2012. The EUR/CHF and EUR/USD futures market shows no massive discount on the euro. On the contrary, EUR/USD navigates within its range of the last 12 months. If a French exit were anticipated, the euro itself would be under direct pressure. Intra-euro-zone TARGET2 balances (the payment system between European central banks) show no massive capital flight from France to Germany. It is this indicator that signalled the Greek risk in real time in 2015. Deposits in French banks stay stable. BNP Paribas, Société Générale and Crédit Agricole are not suffering capital flight. During the real stress episode of spring 2024 (dissolution of the National Assembly), these banks had lost nearly 10% of market capitalisation in a few days (source: Club Patrimoine, September 2025). Today, nothing of the sort. Agence France Trésor auctions remain largely oversubscribed. IFRAP notes in its January 2026 analysis that "despite the political tensions, the option of a fall in the auction cover ratio, revealing a form of distrust, has not for now manifested itself". Yet that is precisely what would happen first if markets anticipated an exit: structural buyers (insurers, pension funds, foreign central banks) would start to withdraw. The real reason for the French premium is fiscal, not existential What explains the 70-bps spread (and not 30-40 bps as before 2024) is a real and identified subject: French fiscal deterioration. The official figures (INSEE, Directorate General of Public Finances, Agence France Trésor) are clear: - Public debt: 117.4% of GDP in the third quarter of 2025 (INSEE, November 2025 release) - 2025 public deficit: 5.4% of GDP, or €152 billion (DGFiP, end-of-management finance law of 8 December 2025) - 2025 interest bill: €52 billion (Agence France Trésor) - Debt issuance planned in 2026: €530 billion total, of which €270 billion of medium- and long-term OATs, an absolute record, above 2020 (the Covid year at €400 billion) (IFRAP, January 2026) S&P Global downgraded France from AA- to A+ in October 2025, and Fitch followed in September 2025. These downgrades do not anticipate a euro exit, but note a public-finance trajectory judged unsustainable. S&P even projects that French debt will reach 121% of GDP in 2028, against 112% anticipated at the end of 2024. It is this sovereign risk premium, comparable to that markets demand on Italy or Spain, that explains the differential with Germany. Nothing to do with a pricing of monetary rupture. Leaving the euro would be financially catastrophic, and markets know it This is the argument the promoters of the Frexit scenario systematically evade. A euro exit would trigger a cascade of immediate and documented effects: 54.7% of French debt is held by non-residents (Banque de France, Q1 2025 data). A redenomination into francs would amount to a technical default against these creditors, what jurists call a violation of the international lex monetae. Rating agencies would downgrade France several notches into speculative territory (sub-investment grade), as happened for Greece. Immediate consequence on new financing: the post-exit spread would be estimated at between 400 and 600 basis points above the Bund for several years, according to the work of Jacques Sapir himself, yet favourable to exit. The annual interest bill, already projected at €78 billion in 2026, would explode. On purchasing power: even Sapir, the economist who defends the project, acknowledges a cumulative inflation of 8% over three years, of which 4.5% in the first year (2012-2017 publications). Mainstream estimates (Patrick Artus at Natixis, Bertelsmann Stiftung) are more pessimistic: -8 to -12% of purchasing power in the first year. On the French banking system: BNP, SocGen and Crédit Agricole hold massive European assets in euros, their liabilities (French deposits) would move into francs. A balance-sheet mismatch of several hundred billion euros, which would require a public recapitalisation or a nationalisation. Without ECB support since France would have left. Markets know all this. No rational investor would hold OATs if they anticipated an exit: they would sell them massively. The maintenance of structural demand for French debt demonstrates, a contrario, that this scenario is not priced. The "experts" cited in support of the Frexit scenario are activists, not analysts Marc Touati has predicted the explosion of the euro and of French debt since 2010. Like any binary prediction repeated every year for fifteen years, it will statistically end up landing on a market event… but that does not validate the method. François Asselineau is president of the UPR, a party whose central and exclusive programme is precisely Frexit. He is not a neutral economist analysing data: he is an activist for whom leaving the euro is the conclusion to which all analysis must lead. No bond strategist at a major bank (Natixis, BNP CIB, Société Générale, Crédit Agricole CIB, Goldman Sachs, JPMorgan, Morgan Stanley) supports the thesis of a Frexit pricing in the current spread. All explain it by the combination of the Iranian energy shock, the repositioning of ECB expectations, and the French fiscal premium. Conclusion: a real worry, but not the one being told The serious subject behind the rise in French rates exists, but it is not that of a monetary rupture. It is that of fiscal sustainability. France has entered the dangerous zone of the snowball effect: when the apparent interest rate on the debt exceeds the nominal growth rate, debt accumulates mechanically, even without new public spending. The Caisse des Dépôts and the OFCE documented it in their 2025 work. If the fiscal trajectory does not straighten out, if deficits stay around 5% of GDP, if growth stays listless, if the interest bill keeps rising, France will eventually have to choose between a painful fiscal consolidation (tax rises, spending cuts) and a real financing crisis. But this eventual crisis, several years out, would take the form of a European assistance programme (with conditionality), not a euro exit. No country has ever voluntarily left the monetary union, and all the institutional mechanisms (ESM, OMT, the ECB's TPI) are designed precisely to prevent a state from being forced into it. To confuse a fiscal risk premium with a monetary-exit pricing is to confuse a reasonable worry with an apocalyptic scenario. The first is documented and deserves a serious debate on public finances. The second is prophecy, not market analysis. --- Primary sources: - Idéal Investisseur, OAT/Bund spread on 15/05/2026. - IFRAP, "2026: record year for France's debt issuance", January 2026. - Club Patrimoine, "OAT-Bund spread: France under political pressure", September 2025. - Trading Economics, 10-year OAT yield. - Putsch Media, "10-year OAT at 3.81%", March 2026. - CNBC, "Treasury yields surge as inflation data points to tricky rates path", 15 May 2026. - INSEE, public debt Q3 2025 (November 2025 release). - S&P Global Ratings, France downgrade to A+, 17 October 2025. - Jacques Sapir, publications on the costs of a euro exit, 2012-2022. - Banque de France, holding of public debt by non-residents, Q1 2025. ============================================================================ ANALYSIS: RealT in liquidation: the token that did not own the house URL: https://l0g.fr/en/analysis/realt-liquidation-token-without-the-deed/ Canonical French source: https://l0g.fr/posts/realt-en-liquidation-le-token-qui-ne-possedait-pas-la-maison/ Date: 2026-07-14 (reviewed 2026-07-14) Topics: crypto, rwa, real estate, tokenisation ---------------------------------------------------------------------------- import InfographicBars from '../../components/infographic/InfographicBars.astro'; Tokenised real-world assets, RWAs, have never weighed so much: around $26 to 32bn of on-chain value in 2026, roughly four times more than a year earlier, according to sector-tracking data (rwa.xyz). Yet one of their most visible pioneers has just collapsed. In early July 2026, during a community call broadcast on YouTube then relayed on Telegram, RealT announced its voluntary liquidation. Some 14,000 French investors are said to be affected, according to the Delomel law firm, within a global base spread across more than 150 countries. The model: an LLC, a house, some tokens RealT, based in Florida and founded by brothers Remy and Jean-Marc Jacobson, had industrialised a simple idea. For each property, a US limited liability company, an LLC, was supposed to hold the house. The shares of that LLC were converted into ERC-20 tokens issued on Ethereum then on Gnosis, and sold in fractions. The token cost about $50, a property counted more than a thousand of them, and each house often gathered several hundred investors. In exchange, the holder collected each week their share of rent, paid in the USDC stablecoin, for an advertised yield of around 10% a year. Barred to US investors, the platform targeted the rest of the world, and France in particular. Since 2019, the portfolio had grown to about 650 tokenised houses in Detroit according to the deeds recorded by the local press, the company claiming close to a thousand. This promise of rental real estate made liquid, fractional and global had made RealT a showcase for real-estate RWA. The liquidation makes it the inverted textbook case. The fall: when the rent dries up The machinery seized up from the bottom, where the bricks meet the real world. In 2024, the city of Detroit brings against the company what its lawyer, Conrad Mallett, describes as the largest nuisance-abatement proceeding in its history. The complaint targets 408 properties and demands compliance certificates within 90 days, failing which the city reserves the right to do the work at RealT's expense. The local entity, Michigan Realtoken, then owes the city at least $2m in tax arrears and blight fines, is behind on taxes on more than 300 properties, and more than 200 properties face foreclosure. Because these properties were in poor condition. The investigation by local outlet Outlier Media counts more than a thousand blight fines and, from postal data, more than a hundred vacant homes. In 2025, a judge places tenants' rents in escrow, reserved for repairs alone. Deprived of collectible rent, the model collapses at its source: in February 2026, RealT suspends almost all distributions, a move some investors call theft. An independent fiduciary, Charles Bullock, takes over the properties in April, a trial is set for 27 May, and in early July the liquidation marks the end of the game. The token was not worth the deed This is where the case touches the core of what a tokenised real-world asset is. By cross-checking the Wayne County records, Outlier Media and local channel WXYZ uncover a disconnect between the token sold and the registered title. The clearest case: RealT collected $2.72m from investors for 39 houses in eastern Detroit, when it had only agreed to pay $1.1m to the seller. More than a year after the last tokens were sold, the deeds to these 39 houses still bore the name of the original seller, Brewer Park Homes, and not RealT. The registered owner, Kathy Makino-Leipsitz, confirmed it: the property was "under contract for more than a year, but the sale was never completed". More broadly, of 25 properties offered to investors in January, only three appeared under RealT's name in the county register. The token therefore promised its holder the status of owner, but the title, enforceable before a judge, often stayed elsewhere. The token lived on the blockchain; the house, in the law of the State of Michigan. Nothing guaranteed that the two coincided. Bricks and mortar, the RWA blind spot This flaw sheds light on a paradox. The RWA market is thriving, but not on the bricks-and-mortar side. Six categories of tokenised assets have crossed the billion-dollar mark on chain: private credit, commodities, US Treasuries, corporate bonds, non-US sovereign debt and institutional alternative funds. Tokenised Treasuries alone account for around $15bn, nearly half the total. Tokenised real estate, long presented as the sector's flagship application, is its exact counter-example: added together, platforms like RealT or Lofty have never exceeded $100m of on-chain value, a crumb at the scale of the market. Why does this category thrive when the bricks break? Because the nature of the underlying differs. A tokenised Treasury is a near-liquid claim, held by a regulated custodian that actually controls the asset: the token is the fund share, settlement is clean, and there is no roof to fix, no property tax to pay, no tenant to evict. Rental real estate, by contrast, requires the token to enforce, off chain, a title governed by a land registry, a tax authority, courts and the physical state of a building. The blockchain has no grip on that substrate: it faithfully records who holds the token, without being able to guarantee that the token commands the brick. This is the limit any serious reading of on-chain data recalls: a faithful register is not a true register. The liquidation balance The balance looks thin. According to the specialist press, the city's escrow account held less than $640,000, when the fiduciary's fees alone reached $178,000 in two months. Once unpaid taxes, administrative costs and the legal fees weighing on the group's companies are settled, there will be little left to distribute among token holders, and many properties face tax foreclosure. One must avoid condemning all of RWA from this single case: the findings of Outlier Media and WXYZ are journalistic, not settled by a court, and RealT is only one actor. But the structural lesson holds. Tokenising an asset does not create ownership; it records a claim on a register. The value of an RWA depends on the strength of the bridge between the chain and the legal world: title, taxation, justice, physical management. That bridge, RealT did not hold. Tokenised bricks and mortar did not fail because they were on a blockchain, but because the blockchain did not fix the roofs and did not hold the deeds. Sources 1. Outlier Media, "The real estate scheme gobbling up Detroit, one digital token at a time": more than 500 properties in Detroit tied to RealT (close to 1,000 claimed), purchases since 2019, token at $50.72, more than 250 investors per property, advertised yield around 10%, more than 1,000 blight fines, more than 100 vacant homes, founders Remy and Jean-Marc Jacobson: https://outliermedia.org/crypto-real-estate-realt-cryptocurrency-detroit/ 2. WXYZ Detroit, "Crypto real estate company RealT collected millions from investors for Detroit properties it doesn't own": $2.72m raised for 39 houses against $1.1m agreed, deeds still under Brewer Park Homes, Kathy Makino-Leipsitz quote, 3 of 25 properties under RealT's name, about 650 tokenised properties, city proceeding targeting 408 properties (Conrad Mallett): https://www.wxyz.com/news/crypto-real-estate-company-realt-collected-millions-from-investors-for-detroit-properties-it-doesnt-own 3. Outlier Media, management and collapse of the model: distributions halted in February 2026, rents in escrow, more than 300 properties behind on taxes, more than 200 facing foreclosure, at least $2m owed to the city: https://outliermedia.org/realt-crypto-real-estate-detroit-landlord-property-management/ 4. Cryptoast, liquidation in early July 2026, about 14,000 French investors (Delomel firm), fiduciary Charles Bullock, escrow below $640,000, fiduciary fees of $178,000: https://cryptoast.fr/immobilier-tokenise-realt-liquidation-francais-concernes/ 5. PYMNTS, tokenised real-world asset value up fourfold to about $26bn, six categories above the billion: https://www.pymnts.com/blockchain/2026/tokenized-real-world-asset-value-jumps-fourfold-to-26-billion/ 6. rwa.xyz, tokenised-asset market tracking data: on-chain value around $31 to 32bn mid-2026, weight of Treasuries, tokenised real estate remaining below $100m: https://app.rwa.xyz/ ============================================================================ ANALYSIS: CLARITY Act: Trump, the first obstacle to his own crypto law URL: https://l0g.fr/en/analysis/clarity-act-trump-conflict-of-interest/ Canonical French source: https://l0g.fr/posts/clarity-act-trump-obstacle-conflit-interets/ Date: 2026-07-14 (reviewed 2026-08-06) Topics: crypto, regulation, us politics, stablecoins ---------------------------------------------------------------------------- Update, 7 August 2026. The conflict described here remains a negotiation issue, not enacted law. H.R. 3633 is still officially Passed House, not Passed Senate. The next pre-midterm opportunity depends on fourteen scheduled session days between 14 September and 2 October, after the August break. Our new article separates that calendar constraint from political scenarios: CLARITY Act: the window before the midterms. The CLARITY Act is the number-one priority of the US crypto industry, the text meant to divide the regulation of digital assets cleanly between the SEC and the CFTC. It has cleared its Senate committee, it is on the calendar, the White House is pushing it. And yet it could fail for a reason few had anticipated: its biggest obstacle is neither a banking lobby nor a technical disagreement, it is the president himself. The provision Democrats demand in exchange for their votes directly targets the crypto interests of Donald Trump and his family, and the administration refuses any text that singles him out. Update 21 July 2026. The obstacle described here has just been lifted in principle. On the evening of 20 July, Donald Trump agreed to an ethics clause in the CLARITY Act, with the White House sending the language to Senate Republicans. Three days earlier, the Senate's merged draft had come out without that clause, drawing public opposition from Senators Murphy, Van Hollen and Merkley. The final text is not yet public and Democrats have not seen it; about ten session days remain before the 7 August recess. We devote a dedicated article to this endgame and its dated scenarios: Trump concedes on ethics, the August countdown begins. Update 7 July 2026. Since this article was published, the file has become better documented. The Office of Government Ethics published on 30 June 2026 Donald Trump's certified 2025 annual report. The official file mentions a 45-day extension and late fees for transactions that had not previously been disclosed on 278-T forms; it also lists, within CIC Digital LLC, $635,068,835 in royalties tied to Celebration Coins. MarketWatch notes that this 927-page filing is nearly four times longer than the previous one. In the same sequence, The New Yorker aggregates the report to more than $2.2bn of income declared in 2025, of which more than $1.4bn is associated with tokens or crypto investments. The Wall Street Journal reports that Democratic senators are calling for hearings into a secret $500m investment in World Liberty Financial from a group led by a senior Emirati official. The White House disputes the conflict: Business Insider quotes Anna Kelly, according to whom the president's assets are managed by independent third-party institutions. Methodologically, this update does not replace the original article; it adds a primary, dated piece to the same diagnosis: the ethics clause is no longer merely a bargaining argument, it now bears on income declared in a public filing. The Digital Asset Market Clarity Act organises a division of authority: the CFTC would gain jurisdiction over the spot markets of digital commodity assets, such as bitcoin, while the SEC would keep assets deemed investment contracts. This legal clarity is the holy grail of a sector that considers it indispensable after years of regulation by enforcement, in the continuity of the GENIUS Act on stablecoins. The problem is not in the architecture of the text, it is in one line that still does not appear in it. The 60-vote wall The text has genuinely advanced. The House passed its version, H.R. 3633, on 17 July 2025. The Senate Banking Committee passed its own on 14 May 2026 by 15 votes to 9, the thirteen Republicans joined by only two Democrats, who immediately warned that their committee vote was no commitment on the floor. On 1 June 2026, the text was placed on the Senate calendar under number 423. Then comes the wall. On the floor, 60 votes are needed to break the filibuster, so at least seven Democrats joining the fifty-three Republicans. Yet Senator Kirsten Gillibrand, though crypto-friendly, set a public condition: no ethics provision, no Democratic votes. An ethics clause tailored to Trump The provision in question would bar senior public officials from holding personal interests in the crypto industry they regulate. Its genesis is explicit: it was born of the president's crypto activities. In committee, an ethics amendment from Senator Chris Van Hollen, which targeted the president and vice president, was rejected by 13 votes to 11, on party lines. Republicans argued that ethics was outside the text's scope and could be added later on the floor; Democrats reply that deferring it means burying it. The White House holds a clear line. Its crypto adviser, Patrick Witt, repeats that a rule applying "to everyone", from the president to the last Capitol intern, would be acceptable, but that any wording targeting a specific office would be rejected. The formula is clever: a general rule with a long transition period might never force the president to divest his positions. The paradox is complete: the man whose administration carries the text is also the one whose interests block its passage. The scale of the stakes, in numbers The amounts explain the tension. According to a Reuters investigation, the crypto ventures tied to the president generated about $2.3bn in pre-tax revenue between November 2024 and April 2026. Senator Jamie Raskin's report, published in November 2025, values the family's crypto holdings at up to $11.6bn. At the heart of the setup is World Liberty Financial, which passes a large share of token-sale proceeds to the family, complemented by the memecoin bearing the president's likeness. Reuters underlines the zero-sum nature of the operation: the family's gains face about $2.25bn of net losses on the retail-investor side, a mechanic dissected in our analysis of the "presidential scam". These ventures are, moreover, interlaced with foreign capital, linked to Gulf states and to actors under surveillance, which raises the ethics debate to the level of national security, something the White House refuses to see named. The other obstacles, quite real The conflict of interest is the main bone, but not the only one. The Banking Committee text must first be merged with that of the Agriculture Committee, which handles the CFTC's powers, a merger still disputed. The question of yield on stablecoins, long explosive, seems settled by compromise. Above all, time is short: the realistic window closes on the August recess, beyond which the midterm campaign absorbs everything, and the text could require up to a week of floor time against budget priorities. An open call Analysts do not settle it. The investment firm Galaxy puts the chances of passage in 2026 at roughly 50-50, owing not to a single hard point but to the number of questions to resolve in sequence under calendar pressure. The negotiators say they are 80 to 85% aligned on substance, which leaves ethics as the only decisive variable. The irony deserves to be stated plainly. If the text passes without a safeguard, it enshrines a president active in an industry he regulates. If it includes a real safeguard, it might never receive the presidential signature. Between the two, an "everyone" wording paired with a long grace period would offer a political exit, at the price of a compromise the firmest Democrats deem cosmetic. The fate of the CLARITY Act, the priority of an entire sector, depends on a trade-off that the law's chief beneficiary is also best placed to make fail. --- Primary sources for the original article: Senate Banking Committee (14 May 2026 release, 15-9 passage); Congress.gov, H.R. 3633, 119th Congress; CoinDesk (markup sequence, amendments, calendar, May and June 2026); Fortune and Elliptic (markup and Van Hollen amendment); Reuters (Trump family crypto revenue estimated at $2.3bn, investor losses at $2.25bn, May 2026); report by Senator Jamie Raskin, House Judiciary Committee Democrats, "Trump, Crypto, and a New Age of Corruption" (25 November 2025); Public Citizen (entanglement with Binance and foreign interests, May 2026); Galaxy Research and Astraea Law (forecasts). Figures and dates were verified one by one. Sources for the 7 July 2026 update: Office of Government Ethics, publication of the certified 2025 annual report; Donald Trump's OGE 278e report; MarketWatch, 1 July 2026; The New Yorker, 2 July 2026; Wall Street Journal, 1 July 2026; Business Insider, 3 July 2026. ============================================================================ ANALYSIS: US import prices: +1.9% in a month, but read the fine print URL: https://l0g.fr/en/analysis/us-import-prices-read-the-fine-print/ Canonical French source: https://l0g.fr/posts/prix-import-mai-2026-petits-caracteres/ Date: 2026-07-14 (reviewed 2026-07-14) Topics: macro, inflation, fed, tariffs ---------------------------------------------------------------------------- This morning, the Bureau of Labor Statistics published its foreign-trade price indexes for May 2026. The figures are already circulating, often summarised like this: import prices up 1.9% on the month, an annualised pace of more than 25%, and 6.7% year on year; export prices up 1.3% on the month and 11.2% year on year. A frequent conclusion attached to these numbers: that this would be a far more reliable measure of inflation than the CPI. First good news for those who like rigour: all four figures are accurate. I checked them line by line against the official release. Imports, all categories: +1.9% in May, +6.7% over twelve months. Exports, all categories: +1.3% in May, +11.2% year on year. None is invented or shifted. The bad news is that the headline hides the essential, and the sentence about the CPI is false. Let us take it properly. Fuel does all the work The +1.9% "all categories" is an average that blends two worlds. On one side fuel, on the other everything else. And the gap is gaping. // Import prices, May 2026 (monthly change, %) Fuels +12.5% All categories +1.9% Ex-fuels +0.8% Source: BLS, Import/Export Price Indexes, May 2026 (Table A). The fuel component, highly volatile, pulls the average up. Imported fuel prices jumped 12.5% in the single month of May. Excluding fuels, import prices rise only 0.8%. The energy item, which weighs a fraction of the basket, therefore explains most of the acceleration on display. Over twelve months, the contrast is even more spectacular: imported fuel climbs 45.1% year on year, against 3.7% for prices excluding fuels. This detail changes everything, because it says where the shock comes from. It is not a diffuse, generalised inflation that would have settled everywhere. It is mainly an oil shock, consistent with the surge in the barrel tied to the Strait of Hormuz. Energy is by nature the most volatile item of any price index, capable of reversing the following month. Building a runaway-inflation narrative on a single month of fuel is to mistake a spark for a fire. The annualisation trap The claim "+1.9% on the month, or more than 25% annualised" is arithmetically correct. Compounding 1.9% over twelve months gives about 25.3%. But it is a classic rhetorical manipulation, and one has to know how to spot it. To annualise is to assume that a monthly move will repeat identically twelve months in a row. Yet we have just seen that this +1.9% is driven by one item, fuel, whose defining feature is precisely never to repeat identically: it rises hard one month, falls back the next. Annualising the most volatile month produces the most spectacular and least informative figure. The proof by the facts: the actual year-on-year rate is 6.7%, not 25%. When you want the pace over a year, you read the figure over a year, you do not extrapolate a single month. Imports versus exports: the real story The most interesting point of this release lies elsewhere, in the divergence between what America pays for its imports and what it charges for its exports. // Year on year, May 2026 (%) Imports +6.7% Exports +11.2% Import fuel +45.1% Import ex-fuel +3.7% Source: BLS, May 2025 to May 2026. Exports rise faster than imports excluding energy. Export prices rise faster than import prices excluding energy: +11.2% year on year against +3.7%. This is a sign that US producers have pricing power on world markets, driven notably by agricultural exports (+5.5% year on year) and non-agricultural exports (+11.8%). For the terms of trade, this is rather favourable: the United States sells its exports at a higher price. But it is also a signal of domestic inflation diffusing outward, and a point of attention for the trading partners that import these American goods. Why this is not a substitute for the CPI That leaves the most problematic claim: that these indices would be a far more reliable measure of inflation than the CPI. This is false, and confusing the two leads to erroneous conclusions. Import/export price indexes measure the prices of goods at the border, at the moment they enter or leave the country. They include neither distribution margins, nor taxes, nor above all services, which make up the bulk of the American consumption basket (housing, health, education, leisure). The CPI, for its part, measures what the household actually pays, services included. The two do not measure the same thing and are not interchangeable. What import prices really bring is a leading signal: they capture pressures on input costs before they feed through into consumer prices. As such, they are valuable for anticipating, and the 3.7% ex-energy rise year on year deserves attention, because it can feed the CPI in the months ahead. It is, moreover, a useful complement to what I described in the great US inflation comeback. But a leading indicator is not a "truer" measure of lived inflation. It is another instrument, to be read for what it is. Beyond the headline The figures in the release are accurate, and that is to the credit of those who relay them. But the honest reading is more nuanced than the headline. The monthly rise is overwhelmingly driven by fuel, a volatile item; the 25% annualisation is an artifice; the real information is the firmness of prices excluding energy and the strength of export prices; and these indices complement the CPI, they do not replace it. To track these dynamics over time rather than on a single release, the US macro risk dashboard aggregates the inflation and price-pressure series. One figure does not make a trend. One month of fuel does not make an inflation regime. --- Sources: U.S. Bureau of Labor Statistics, Import/Export Price Indexes, release of 16 June 2026 (May 2026 data), Table A and press release. Year-on-year changes from May 2025 to May 2026. This is not investment advice. ============================================================================ ANALYSIS: Dollar-yen: the risk is not only the level, it is the unwind URL: https://l0g.fr/en/analysis/dollar-yen-intervention-carry-unwind/ Canonical French source: https://l0g.fr/posts/dollar-yen-intervention-risque-carry-2026/ Date: 2026-07-14 (reviewed 2026-07-14) Topics: yen, boj, dollar, carry trade, macro, markets ---------------------------------------------------------------------------- Dollar-yen is back in a politically flammable zone. The official FRED DEXJPUS series gives 161.37 yen to the dollar on 18 June 2026; WSJ/LSEG quotes still put USD/JPY around 161.94 on 25 June then 161.65 on 26 June. This is not just a round number. It is a zone where Japan has already shown it can sell dollars to buy yen. The institutional nuance matters. An FX intervention is decided by the Ministry of Finance (MoF), not by the Bank of Japan (BoJ) alone, even if the central bank can act as the operational agent. The register of the Japanese Ministry of Finance shows net intervention of ¥11,734.9bn between 28 April and 27 May 2026. In other words, Tokyo has already spent ammunition. The market is now testing the willingness to do it again. // USD/JPY: back in the intervention zone 150 155 160 165 161-162 zone May 18 Jun 25 Jun 26 Jun Sources: FRED DEXJPUS, WSJ/LSEG. Data accessed 27 June 2026. The paradox is that the BoJ has already tightened. Its monetary policy decision of 16 June 2026 took the policy rate to 1.0%. But a Japanese rate at 1% remains low against still-high US yields, in the wake of Warsh's first FOMC and a Fed constrained by inflation. As long as the yield gap stays massive, selling yen to buy dollars remains a rational trade. That is where the risk becomes systemic. The yen carry works like an implicit short position on volatility: you borrow in cheap yen, buy better-yielding assets, and all is well as long as the yen does not rise sharply. A successful intervention therefore does not only create a candle on the FX chart. It can force investors to cut, at the same time, positions in dollars, Treasuries, equities, credit or crypto, depending on how they are funded. // The real risk: a mechanical unwind of the carry weak yen cheap funding asset purchases dollar, rates, risk intervention forced yen buyback unwind sale of liquid assets Reading: intervention becomes dangerous when it turns an FX loss into a forced sale of liquid assets. My assessment: Japanese intervention is likely if the pace of depreciation reaccelerates, but it is not enough to durably reverse the trend without support from rates. The threshold to watch is therefore not only 162 or 165. It is the combination: a fast rise in USD/JPY, MoF rhetoric, yen implied volatility, and speculative positioning published by the CFTC (see reading the COT report). The Yen Carry dashboard is built precisely to track this mechanism. The Treasuries channel also deserves attention. Japan remains a major creditor of the United States according to the US Treasury TIC data. If defending the yen forces sales of dollars or changes the currency hedges of Japanese investors, the effect can transmit to US yields, and thus to global liquidity. This is the same world described in our guide on net liquidity: FX, Treasuries and market funding are not three separate subjects. A sober conclusion: a weak yen helps Japanese exporters, but a yen that breaks too fast becomes a global market risk. Intervention can calm the spot. It can also trigger what it seeks to avoid: a disorderly exit from the carry. --- This is not investment advice. Market data accessed 27 June 2026. Primary sources: FRED, DEXJPUS, last available observation as of 18 June 2026; WSJ/LSEG, USD/JPY historical prices, quotes for 25 and 26 June 2026; Bank of Japan, monetary policy decisions, 16 June 2026; Ministry of Finance Japan, Foreign Exchange Intervention Operations, April-May 2026 record; U.S. Treasury TIC, major foreign holders of Treasury securities; CFTC, Commitments of Traders. ============================================================================ ANALYSIS: SpaceX goes public, 12 June 2026: the red carpet the SEC rolls out for the most expensive IPO in history by bending its own rules URL: https://l0g.fr/en/analysis/spacex-ipo-sec-red-carpet/ Canonical French source: https://l0g.fr/posts/spacex-ipo-tapis-rouge-sec/ Date: 2026-07-14 (reviewed 2026-07-14) Topics: markets, macro, crypto, us politics ---------------------------------------------------------------------------- On Friday 12 June 2026, Space Exploration Technologies Corp. takes its first steps on the Nasdaq under the ticker SPCX. A fixed price of $135 a share, about 556 million shares, $75bn raised, a valuation of $1,750bn. This is, by a wide margin, the largest initial public offering ever completed, more than double the record set by Saudi Aramco in 2019, according to Reuters. The staging is grand. The regulatory machinery that made it possible, far less flattering. An accelerated review and a take-it-or-leave-it price SpaceX filed its confidential S-1 on 1 April 2026, then its public version on EDGAR on 20 May (registration 333-296070). Between the two, the SEC wrapped up its review faster than expected, which allowed the calendar to be brought forward from an offering initially set for late June, according to three sources cited by Reuters. For the heaviest listing in the US equity market, the express handling raises questions. Two structural anomalies accompany this favourable treatment. First, SpaceX set a single price of $135, take it or leave it, instead of the indicative range that moves with demand, as convention dictates. Second, up to 30 percent of the offer, about $22.5bn, is reserved for retail, against 5 to 10 percent normally. Ordinary savers are thus exposed, en masse, to the most opaque and most expensive listing on the market. Then there is the index question. The inclusion of SPCX in the major indices, and the tens of billions of forced buying that come with it, turns a speculative bet into an imposed exposure for millions of passive fund holders. S&P Global declined to play along as things stand, but the mechanism remains explosive. A moonshot S-1 The prospectus itself is worth the detour. In it, SpaceX promises AI compute satellites in sun-synchronous orbit, meant to process inference at a scale beyond terrestrial data centres, with a first deployment announced as early as 2028. The syndicate banks project $140bn of Starlink revenue in 2030. The $1,750bn valuation, as Elizabeth Warren writes, demands many acts of faith. Above all, what investors are being asked to value is no longer the profitable space company we knew. In February 2026, SpaceX absorbed xAI, Musk's AI company, renamed internally. Consolidated result: a loss of about $5bn over 2025, of which nearly $4.94bn is directly tied to the xAI merger, and a cash burn on the order of $1bn a month (Investing.com, Yahoo Finance). The Space and Connectivity segments remain profitable. The cash furnace is AI, consolidated by force into a single vehicle rather than isolating the mature activities, as management had nonetheless hinted. All of this against a backdrop of political frenzy around AI, the one where a phrase promising "the American people as shareholders of AI" moved semiconductors by tens of billions. Warren rings the alarm On 9 June, Elizabeth Warren, the ranking Democrat on the Senate Banking Committee, sent a twelve-page letter to SEC chair Paul Atkins. In it she asks that the listing be delayed until investors are protected. Her grievances: potentially misleading accounting around the xAI acquisition, Musk's uniquely unchecked power through a multiple-voting-rights structure, and rigged stock indices that would force millions of savers to hold SpaceX without having chosen to. Her formula sums up the affair: major risks for small holders, enormous advantages for insiders. At the time of listing, neither SpaceX nor the SEC had responded on the substance. The financial risks, unvarnished The work of Jay Ritter (University of Florida) has been a reminder for decades: the best very large IPOs underperform the S&P 500 in the years that follow. SpaceX may be the exception, but the base rate is not flattering. Add a float estimated between 3 and 5 percent, which amplifies volatility, lock-up periods that release paper later, conflicts of interest between Musk entities (Tesla, xAI, SpaceX), and an overweight retail base quick to sell if the first session disappoints. The cocktail is a known one. And that huge short position On the synthetic side, the show has already begun. With no shares available before the bell, pre-IPO perpetuals on SPCX have been trading since mid-May (Hyperliquid, Kraken up to 5x), where you can be long or short on a price, not on the company. Arkham Intelligence flagged precisely a $5.7m short position at 2x leverage opened by an account under the pseudonym wenyu8888888, which it describes as the largest SpaceX short it has tracked. The thesis is clear: the IPO premium will deflate once the public listing is under way. Facing it, Arkham and Onchain Lens spotted a record long of $16.6m posted by address 0x9cc. A useful reminder: these perpetuals never become shares, they only track the price. The red carpet is rolled out, the rules softened, the S-1 cosmic. Good luck to investors. Good luck to traders. --- Primary sources: SEC EDGAR (S-1, registration 333-296070), Senate Banking Committee (Warren letter to Atkins, 9 June 2026), Reuters, CNBC, Yahoo Finance, Investing.com, Arkham Intelligence. Data as of 12 June 2026. This article is journalistic analysis and does not constitute investment advice. ============================================================================ ANALYSIS: 13FLOW: the money that is heavy and the money that knows URL: https://l0g.fr/en/analysis/13flow-institutional-and-insider-signals/ Canonical French source: https://l0g.fr/posts/13flow/ Date: 2026-07-14 (reviewed 2026-07-14) Topics: 13flow, SEC EDGAR, 13F, Form 4, markets, fundamental analysis ---------------------------------------------------------------------------- The 13F and the Form 4 are two of the rare clean signals that US regulation makes public. The first exposes, every quarter, the long positions of managers above $100m. The second captures, within two business days, the trades of executives on the stock of their own company. Both are in the public domain and readable on EDGAR. 13FLOW does not claim to give access one would not otherwise have: it industrialises the reading and, above all, it crosses the two signals, something the historical databases (Dataroma, WhaleWisdom) do badly or not at all. The value of the crossing rests on an asymmetry. The 13F is dense but late: up to 45 days after the quarter's close, it is an already stale snapshot by the time it lands. The Form 4 is near real time but individually weak: a single executive's purchase, in isolation, has no predictive value. Their intersection corrects both flaws at once. A stock that several institutions are accumulating and that executives are buying at market price, within the same window, is the coincidence of two populations that share neither the same information nor the same constraints. The noise cancels out, the conviction remains. Building the score The Confluence Score, from 0 to 100, aggregates four explicit components. Institutional breadth: the number of funds adding to the position, weighted by their conviction, that is, the weight of the line in the portfolio and not merely its presence. Insider conviction: the number of distinct buyers, seniority (a CEO or CFO purchase weighs more than a director's), amount committed. The dollars actually mobilised on both sides. And an agreement term that rewards the directional alignment of the two signals. Each card exposes its breakdown pillar by pillar: the score is auditable, not declarative. Signal hygiene Two methodological choices make the quality of the whole. Time, first: each insider purchase decays with a half-life of about 30 days, a filing from 3 days ago does not weigh like one from 80. Relative size counts (a line increased by 30% is not a symbolic purchase), as does concentration: several insiders within a 14-day window form a cluster, and that is the strong signal. Filtering by transaction code, next: only market-price orders count, purchases (P) and sales (S). Awards, option exercises (M) and tax withholdings (F) are parsed but excluded from the score, because they reflect no discretionary decision to enter or exit. That is precisely the noise most insider trackers let through. Data and surface The data comes directly from the 13F-HR and Form 4 filings on EDGAR, read at source, with no aggregator in between. The CUSIP-to-ticker mapping goes through OpenFIGI. Four screens organise the tool: Consensus, Funds, Compare, and the Confluence table that orders the universe by convergence. The tool is live at 13flow.eu and its code is public on GitHub. The framing remains, stated plainly: 13FLOW is a screen, not investment advice, and the crossing of two public signals guarantees no performance. What it produces is a defensible reduction of the universe on verifiable data. Analysis begins where the tool stops. ============================================================================ ANALYSIS: The GENIUS Act: the 18 July deadline, between settled rules and a bet on the debt URL: https://l0g.fr/en/analysis/the-genius-act-stablecoins-and-the-debt/ Canonical French source: https://l0g.fr/posts/genius-act-echeance-18-juillet-regles-actees-pari-sur-la-dette/ Date: 2026-07-13 (reviewed 2026-07-13) Topics: stablecoins, crypto, regulation, debt, us treasury ---------------------------------------------------------------------------- 18 July 2026 marks the first anniversary of the GENIUS Act, the US stablecoin law, and above all the legal deadline for its implementing rules. Six federal agencies are right now finalising the framework that will govern a private money of more than $230 billion. Around this date, two orders of reality must be distinguished: what the law sets in stone, which is a fact, and what the market expects from it, which remains a bet. Confusing the two would be the best way to misread the event. A framework reminder is in order. A payment stablecoin is a digital token meant to be worth a dollar, backed by reserves and redeemable at par. We described how this market and the law work in our guides on stablecoins and the GENIUS Act and on its enforcement architecture. This article does not redo that pedagogy: it focuses on what the 18 July deadline changes, and on the share of narrative that surrounds it. The timeline and the settled rules Let us start with the facts. The GENIUS Act was enacted on 18 July 2025. The law required its implementing rules to be published no later than a year after, that is, 18 July 2026. Six agencies are concerned: the OCC, the FDIC, the NCUA, the Treasury, FinCEN and OFAC. Each published a proposed rule between March and April 2026, and all the public consultations closed on 9 June, placing publication of the final texts in the June-July window. The content of these rules is already written into the law. A payment-stablecoin issuer must hold 100% reserves, backed by cash, insured bank deposits and short-maturity Treasury bills, of 93 days or less. It must publish the composition of its reserves every month. It is forbidden to pay interest to holders. Above $10 billion of outstanding, it mandatorily moves under federal supervision; below it, it can opt for a state regime deemed equivalent. Finally, foreign issuers are excluded from the US market, except by a reciprocity agreement negotiated by the Treasury. This base is solid and, for the most part, uncontested. The battle of comments focused mainly on implementation details, notably OFAC's reach over foreign issuers. But the structure will not move anymore. It is from this base that the uncertain part begins. The Treasury-bill channel From the reserve rule follows a mechanical fact: a compliant stablecoin issuer is a near-automatic buyer of short Treasury bills. For each token issued, it must place a dollar in cash or government debt of less than 93 days. The stablecoin thus becomes a cousin of the money market fund, with the same reserve mechanics, but without the right to pass the yield to its holders. This point is settled, and it is not trivial: it directly links private digital money to the financing of the state. Where we leave the fact for the bet is on the scale. The stablecoin market today weighs about $230 billion. The projections, for their part, soar. Treasury Secretary Scott Bessent estimated the US market could exceed $2 trillion by the end of 2028, and Standard Chartered puts at about $1 trillion the new Treasury-bill demand that would result. The same bank calculates that, added to the other needs, this demand could exceed the expected net bill supply, forcing the Treasury to issue more short-term debt. These figures must be held for what they are: scenarios, issued by analysts and officials who have an interest in a flourishing market. The mechanism is real, the order of magnitude is not yet. A compliant issuer does buy Treasury bills; that they collectively be a trillion depends on an adoption trajectory that is nothing written. The link between stablecoins and debt auctions is a credible hypothesis, not an observed fact. The ban on paying yield, and its side effects One provision deserves particular attention, because it will shape the market more than the others: the ban on issuers paying a yield to holders. The intent is clear, to prevent stablecoins from becoming disguised savings accounts, escaping bank regulation and able to siphon banks' deposits. But the side effect is predictable. If a stablecoin pays nothing while its reserves, for their part, produce interest pocketed by the issuer, two dynamics kick in. On one side, compliant issuers see their business model confirmed: they keep the reserve yield, very lucrative at this scale. On the other, holders seek yield elsewhere, which pushes toward tokenised money market funds and other interest-bearing vehicles, in direct competition with payment stablecoins. The rule does not remove the appetite for yield, it displaces it. It also creates a geographic asymmetry: per several analyses, the same issuer can offer yield on its tokens issued outside the United States, but not on those regulated in the United States, a border that invites circumvention. The Tether case, the framework's blind spot No honest reading can ignore the elephant in the room. Tether, the issuer of USDT, represented about two-thirds of the global stablecoin supply in mid-2026. Yet Tether is domiciled outside the United States, and the question of whether OFAC can really constrain a foreign issuer serving Americans is precisely one of the points the consultation was to settle. The reciprocity determination that would open the US market to the group has not been issued to date. Two readings clash. In the first, 18 July puts Tether against the wall: without compliance, access to the US market closes, and the group has moreover launched a dedicated token, USAT, designed for the US rules. In the second, more sceptical, the bulk of Tether's activity will simply stay offshore, beyond the US regulator's reach, and a non-compliant dollar-stablecoin market will keep thriving internationally. US regulation would then frame the domestic stablecoin without reducing the global dollar stablecoin. Which one wins remains open, and it is the framework's heaviest uncertainty. The possible trajectories Several continuations emerge. These are scenarios, not forecasts. The first scenario is the clear framework and the wave of demand: the rules are finalised in time, compliant issuers like Circle thrive, and Treasury-bill demand begins to materialise, validating the narrative of state financing by digital money. The second is the timeline that slips: the agencies publish interim or incomplete rules, and implementation spreads out, because regulatory sprints rarely keep all their promises on time. The third is circumvention by Tether, already mentioned. The fourth, the least commented and the most serious, is the run. The risk we watch least: the run Here is the counter-thesis, the one that tempers both enthusiasm and alarmism. The dominant narrative presents the GENIUS Act either as a revolution in state financing, or as a systemic bomb in gestation. Both probably overestimate the event, and neglect the real weak point. On the optimistic side, some perspective is needed. A $230 billion market stays modest against the $28 trillion of US marketable debt and the $8 trillion of money market funds. The stablecoin as a great creditor of the state is a projection to 2028, not a 2026 reality, and the history of adoption forecasts invites caution. The framework, on the other hand, brings a real and underestimated progress: by requiring reserves in short Treasury bills and monthly transparency, it cleans up a long-opaque market, where Tether once held commercial paper and ended up settling with the New York justice over its reserves. From this viewpoint, the law reduces a risk rather than creating one. On the danger side, the error would be to look for the threat in the wrong place. The risk is not that stablecoins buy too many Treasury bills, it is that they cannot sell them fast enough on the day of a run. A stablecoin is a money market fund without the money market fund's protections: neither the liquidity cushions of rule 2a-7, nor above all a central-bank net. Yet the structure invites flight, as for a fund or a bank: at the first doubt, better to exit at par before the others. The precedent exists. In March 2023, Circle's USDC depegged to about $0.87 when the market learned that part of its reserves was frozen at the failing Silicon Valley Bank. The token only regained its peg after the federal rescue of the bank. Safe reserves do not prevent a depeg if redemptions go faster than liquidity, and the GENIUS Act, as it stands, provides no lender of last resort for issuers. We detailed this run mechanic in our money market funds guide; it holds, worse, for stablecoins. The fact and the bet The 18 July deadline is a real milestone, and it must be read without excess in either direction. The fact is a framework finally written: full reserves in short, safe assets, transparency, ban on paying yield, a regulatory border. It is a clear improvement on the era of opacity, and that deserves to be said. The bet is the narratives grafted onto it: the trillion-dollar wave of debt demand, Tether's submission, the monetary revolution. They are possible, not settled, and the honest analyst flags them as hypotheses. As for the risk, it is not in the excess safety of the reserves, but in the absence of a net the day confidence wavers. The right reading of 18 July holds in one sentence: the rule is settled, the bet begins, and the question we ask least, that of the run, is the one that matters most. Sources 1. Congress.gov, text of the GENIUS Act (S.1582, 119th Congress), enacted 18 July 2025: https://www.congress.gov/bill/119th-congress/senate-bill/1582/text 2. Stablecoin Insider, six federal agencies at the 18 July 2026 deadline, consultations closed 9 June: https://stablecoininsider.org/six-federal-agencies-have-35-days-to-finalize-genius-act-stablecoin-rules-by-july-18/ 3. OCC, proposed implementing rule (12 CFR Part 15), March 2026: https://www.occ.gov/news-issuances/bulletins/2026/bulletin-2026-3.html 4. U.S. Department of the Treasury, proposed GENIUS Act anti-money-laundering rule: https://home.treasury.gov/news/press-releases/sb0435 5. Brookings, "Next steps for GENIUS payment stablecoins": https://www.brookings.edu/articles/next-steps-for-genius-payment-stablecoins/ 6. The Block, Standard Chartered projection of about $1 trillion of stablecoin-linked Treasury-bill demand: https://www.theblock.co/post/390783/stablecoins-could-drive-1-trillion-in-t-bill-demand-giving-treasury-room-to-shift-issuance-standard-chartered 7. The Block, Scott Bessent estimates the US stablecoin market could exceed $2 trillion by the end of 2028: https://www.theblock.co/post/357872/us-stablecoin-market-could-exceed-2-trillion-projection-by-end-of-2028-thinks-treasury-secretary-bessent 8. l0g, guides Stablecoins and the GENIUS Act and Who enforces the GENIUS Act. 9. l0g, Reading money market funds. ============================================================================ ANALYSIS: Private credit in 2026: the new king of shadow banking starts to cough (and stammer on withdrawals) URL: https://l0g.fr/en/analysis/private-credit-the-new-shadow-banking/ Canonical French source: https://l0g.fr/posts/private-credit-2026-shadow-banking/ Date: 2026-07-13 (reviewed 2026-07-13) Topics: macro, private credit, finance ---------------------------------------------------------------------------- Private credit, that charming euphemism for "direct loans to companies by non-bank funds", that eldorado that grew like a mushroom after the rain of zero rates. In this early April 2026, the market is nearing or exceeding $2 trillion of assets under management (AUM), with Moody's forecasting a tidy breach of the $2trn mark this year and a nice $4trn by 2030. Preqin even talks of $4.5trn if you count the semi-liquids and evergreen funds. The United States alone weighs between $1.3trn and $2.1trn depending on the source. It's beautiful, it's big, it's… exactly the kind of size that makes regulators say "nothing to do with the shadow banking of 2008, promise". The yields? Still sexy on paper: directly originated first-lien loans should land around 8-8.5% in 2026, even after spread compression. Not bad for an "illiquid" asset that promised the illiquidity premium. Except that… the cracks are starting to show. And not just a little. March-April 2026 will go down in the annals as the moment the "zero-loss fantasy" started to look its age. Redemption requests are exploding at BDCs (Business Development Companies). Several large managers had to put caps on withdrawals to avoid the haemorrhage. Bloomberg headlined at the end of March: "Why investors are rushing to exit the private credit market now". The war in Iran, AI wrecking software business models, rates staying high… it all adds up. Result: PIK (payment-in-kind, in other words "we pay the interest in… additional debt") has doubled, reaching 11% of the market at the end of 2025. Real defaults? Around 5.4% over 12 months in February 2026 per Fitch (slightly down, phew), but Morgan Stanley sees 8% as possible, UBS up to 15% in the worst AI scenario. Analysts are already talking of "shadow defaults": maturity extensions, covenant waivers, quiet restructurings. The kind of thing you don't see in the headlines but that hurts the portfolio. The tastiest part? The banks, those nasty regulated players who had fled the middle-market, now lend $300 billion to private-credit funds (Moody's). They've become the managers' best friends… while starting to take back market share on leveraged loans. Translation: private credit filled the void left by post-2008 bank regulation, and now it's getting so big that even the banks are coming back to nibble. It's almost poetic. On the investor side, retail and HNW individuals are rushing in via interval funds and semi-liquid structures (nearly a third of the US direct-lending market). The assets of semi-liquid credit funds jumped 22% in the first half of 2025 alone. Great: we democratise illiquidity just as redemptions become… complicated. Managers shout "historic opportunity" while institutional LPs eye the evergreen funds like liquidity saviours. But when the real crisis arrives, we all know how it ends: the queues to exit lengthen and the "quarterly look-throughs" suddenly become very interesting. Innovation is everywhere: asset-backed finance (ABF) becomes the new engine of growth (consumer loans, data centers, infrastructure), private securitisation is exploding, NAV facilities and rated funds are multiplying. In short, we complexify to death to keep the yields up. It's the financial equivalent of "we'll just add a layer of derivatives, it'll be fine". Regulators are watching, of course. They talk of "more transparency" and opening to retail. We know the tune. In 2026, the watchword of the real experts (those not selling LP interests): extreme selectivity. Performance dispersion is going to explode. Forget the easy beta of the 2022-2024 years. You'll have to sort the managers who really know how to underwrite in a world where AI destroys the smallest cash-flow forecast. The "AI-disruption proof" sectors (or at least the less exposed ones) will win out. The rest? That will be the big reset everyone was talking about under their breath. Private credit is not dead. Far from it. It remains the reference financing for the US middle-market and, increasingly, the European and Asian ones. But it is entering its "adult" phase: the one where the promises of risk-free return smash against the reality of cycles. The one where the opacity that made its charm suddenly becomes very inconvenient. The one where the $2trn of outstanding starts to look like a nice systemic leverage disguised as diversification. I mapped these transmission channels toward banks, insurers and crypto in detail in the silent contagion of private credit. Welcome to 2026. Private credit is no longer the institutions' well-kept little secret. It has become Wall Street's big circus, with its BDCs shutting the windows and its managers explaining that "it's just a temporary adjustment". We applaud and keep an eye on the covenants. Because when PIK becomes the norm, it's rarely a good sign. ============================================================================ ANALYSIS: Zombie funds: the great illusion of private valuations hits its limits URL: https://l0g.fr/en/analysis/zombie-funds-private-valuations/ Canonical French source: https://l0g.fr/posts/zombie-funds-valorisations-privees/ Date: 2026-07-13 (reviewed 2026-07-13) Topics: private equity, valuation, finance, private markets, private credit, liquidity ---------------------------------------------------------------------------- Zombie funds: when valuations outlive markets For years, private equity sold a simple story: less volatility, more return and value creation superior to listed markets. But in 2026, one question becomes impossible to avoid: do the valuations displayed by many funds still reflect economic reality? The "zombie funds" phenomenon is not new. Traditionally, the term designated funds reaching the end of their life, unable to sell their holdings but continuing to exist thanks to management fees. Today, the concept has broadened. It now encompasses thousands of private assets whose value is maintained on paper even as exit conditions have sharply deteriorated. The problem is first a liquidity problem. Since the rate cycle turned in 2022, initial public offerings have become scarce, mergers and acquisitions have slowed, and strategic acquirers have become far more selective. As a result, funds hold their stakes much longer than planned. Faced with this absence of exits, managers find themselves in a dilemma. Selling today would often mean accepting multiples lower than those used in their internal valuation models. Not selling, on the other hand, preserves a higher net asset value (NAV) and avoids crystallising losses. This is where the heart of the problem appears. Unlike listed markets, where the price is continuously discovered, private assets are valued periodically according to internal models. As long as no transaction takes place, the displayed value stays largely theoretical. Institutional investors are beginning to wonder. Several large private-credit funds have recently limited their clients' withdrawals in the face of rising liquidity demands. When too many investors want their money back simultaneously, the theoretical valuation runs into the limits of the real market. To buy time, the industry has developed a whole series of financial tools. The most emblematic is the continuation fund. Concretely, a manager transfers an asset from an old fund into a new vehicle, offering partial liquidity to investors while keeping the asset under control. In parallel, NAV loans are seeing explosive growth. These financings use the portfolio's assets as collateral to generate cash without selling the holdings. The market now exceeds $100 billion. This evolution recalls an old financial lesson: liquidity is abundant right up to the moment everyone needs it simultaneously. This does not mean that all private valuations are artificially inflated. Some fund-owned companies keep growing and fully justify their multiples. The large players with quality assets and privileged access to capital seem better equipped to weather this period. The real question for the years ahead is therefore not whether zombie funds exist. They already do. The question is how many current valuations would survive a genuine price discovery on a deep and liquid secondary market. For investors, the lesson is simple: a valuation is not a price. Until an asset has found a buyer, its value remains, above all, a hypothesis. ============================================================================ ANALYSIS: The dollar rebound: how far before the central banks strike back? URL: https://l0g.fr/en/analysis/the-dollar-rebound/ Canonical French source: https://l0g.fr/posts/rebond-du-dollar-jusqu-ou-avant-la-riposte-des-banques-centrales/ Date: 2026-07-13 (reviewed 2026-07-13) Topics: macro, dollar, fed, central banks, yen, rates ---------------------------------------------------------------------------- The dollar posted in June 2026 its best month in more than a year: +2.3% on the DXY index, a thirteen-month high at 101.8. Against it, Japan has already burned 11.7 trillion yen to defend its currency, the BoJ and the ECB are raising rates. The question is no longer whether the central banks react, but whether their responses are enough. On 17 June 2026, the Federal Reserve did nothing, and it is precisely this non-move that relaunched the dollar. Rates unchanged, but projections flipped: for the first time in the cycle, the committee's median anticipates higher rates at year-end. In the days that followed, the greenback broke through its resistances, the yen slid beyond 160, and finance ministries brought back the vocabulary of the big manoeuvres. Three weeks later, a much weaker-than-expected jobs report complicated the story. A state of play, figures in hand. A dated and documented rebound The sequence reads on the DXY index, which measures the dollar against a basket of six currencies. According to MUFG research, the index closed June at 101.155, up 2.3% on the month after +0.9% in May, its best monthly close since March 2025. It touched a thirteen-month peak of 101.8 at month-end, taking the year-to-date gain to about 3% and 5% since late January, per Morningstar. On 7 July, the index still held just above 100.9 per Trading Economics. The move is therefore real, but recent, and it owes mostly to a single catalyst. That catalyst is the FOMC of 17 June, the first chaired by Kevin Warsh. The Fed held its range at 3.50-3.75%, but the dot plot shifted: the median of projections for end 2026 rose to 3.8%, against 3.4% in March, moving from an implicit cut to a hike. Nine participants out of eighteen project at least one increase this year. Futures markets went from about 24% to 77% probability of a hike by December. Warsh, true to his positions, did not add a dot to the cloud, but judged at the press conference that inflation remained clearly above the 2% target, with a price index running around 4.2% year on year. HSBC summed up the surprise in one word: hawkish. The mechanics are classic and worth spelling out: when the market revises the trajectory of US rates upward, dollar yields become more attractive relative to the rest of the world, capital flows in, the dollar rises. It is the rate gap that carries the currency, not a judgement on the health of the US economy. Our dollar reading guide details why the DXY is only a partial thermometer of this phenomenon. The yen, the first front of the response The most visible victim of the rebound is the yen. The dollar crossed 160 yen on 30 April 2026, triggering Tokyo's first FX intervention since July 2024: the Ministry of Finance sold dollars to buy back its currency, bringing it briefly toward 155, per Nikkei Asia. The lull did not last. Per CNBC, Japan spent a total of 11.7 trillion yen, about $73.5 billion, on FX intervention over the April-May period, only to see the yen fall back to 160 and settle there for most of June. At the end of June, the pair traded at 162.53 per MUFG. The BoJ added its stone to the edifice on 16 June by raising its policy rate 25 bps to 1.00%, its highest level in more than thirty years, a decision the Japanese press called a foregone conclusion given how imported inflation and yen weakness weighed. The paradox is cruel: neither two massive interventions nor a historic rate hike brought the yen durably below the red line. The reason lies in the infographic above. At 1% against 3.75%, the yield gap between the yen and the dollar stays gaping, and as long as it persists, selling yen to carry dollars remains a profitable trade. Goldman Sachs drew the consequence on 6 July by revising its twelve-month USD/JPY forecast from 155 to 165, one of the most bearish calls on the yen in the consensus. The MoF nonetheless keeps its finger on the trigger. According to Citigroup, cited by Yahoo Finance, Tokyo is unlikely to re-intervene as long as the yen does not weaken toward the 160-162 zone, in other words the zone where it already is. And the Finance Minister declared on 30 June that Japan and the United States were aligned on FX policy, a formula that, in the muffled grammar of currencies, signals that Washington would not oppose a new operation. ING recalls, for its part, the hierarchy that matters: a unilateral intervention impresses for a few hours, a coordinated intervention among several central banks changes the game. Nothing indicates to date that a joint action is on the table; it is a scenario, not a fact. Europe tightens, without rushing On the euro side, the response takes another form. The ECB raised its deposit rate 25 bps to 2.25% on 11 June 2026, its first hike since 2023, and meets again on 23 July. Markets judge a second hike likely this year, but Christine Lagarde's remarks, deemed dovish, cooled bets on a third, per Cambridge Currency's summary. The euro pays the relative price: around $1.14 in early July, far from its January peak above $1.20, the single currency absorbs the rate differential without drama but without spring. The euro zone does not have an acute FX problem like Japan: a euro at 1.14 even helps its exporters. Its constraint is elsewhere, in inflation imported through energy invoiced in dollars. An ECB tightening moderately, at 2.25% against 3.75% for the Fed, chooses to let a little exchange rate slip so as not to smother a fragile recovery. It is a response, but a quiet one. The 2 July setback Just as the strong-dollar story seemed to settle in, the US jobs report published on 2 July introduced serious doubt. June's job creation came in at 57,000, less than half the consensus of 115,000, the weakest figure in four months, and revisions cut 74,000 jobs from the two previous months, per the BLS and Yahoo Finance. The unemployment rate did fall to 4.2%, but for a bad reason: labour-force participation dropped to 61.5%, its lowest since March 2021. Our jobs report guide details why unemployment falling through participation is not good news. The market reaction was immediate: expectations of a Fed hike as soon as September receded, short yields fell, gold jumped beyond $4,130 an ounce and the dollar gave back part of its gains. There is the dollar rebound caught between two forces: a monetary-policy committee leaning toward a hike because of inflation, and a labour market starting to crack. The FOMC minutes, expected on 8 July, will say how the committee weighed these two risks even before the jobs-report shock. Three possible continuations Three trajectories structure what comes next. These are analyst scenarios, not forecasts, and the probabilities each is assigned are a matter of judgement. The first is the dollar's second leg up. If US inflation stays around 4% and June's jobs prove a statistical accident, the Fed delivers the hike its dot plot sketches, in September or December. The rate gap widens further, the yen breaks 162, and Tokyo finds itself cornered into a third intervention, alone. Recent history suggests it would buy weeks, not a reversal: at 165, Goldman's forecast would become the path of least resistance. The second is the response that moves up a gear. The BoJ has a September meeting where a further hike remains, per MUFG, barely priced by the market, about 5 bps. The ECB can deliver its second hike on 23 July or in September. And if the yen clearly overshoots 162 with the US blessing suggested by the Japanese Finance Minister's remarks, a massive, even concerted, intervention becomes credible again. This scenario does not reverse the dollar, but it bounds its rise: each surge of the greenback would trigger a firmer response, drawing a de facto ceiling. The third is the endogenous ebb, without a spectacular response. It is, by implication, MUFG's central scenario, which sees the DXY come back toward 99.8 in the third quarter and the euro rise toward 1.16: the dollar rebound would fade on its own because the US labour market is deteriorating and the rate-hike premium is deflating, as it began to on 2 July. In this world, central banks do not have to strike back; they only have to wait for the US data to do the work. A single pillar, therefore fragile Finally, one must consider the objection that weakens the whole strong-dollar story. This rebound is short, barely three months, and it rests on an expectation, not an act: the Fed has not raised anything at all yet. A dot plot is neither a plan nor a commitment, as our dot-plot guide recalls, and Warsh himself refused to contribute to it. If the July and August jobs data confirm June's chill, the 77% hike probability can deflate as fast as it inflated, and with it the sole pillar of the rebound. Positioning works the same way: a market that rushed into the dollar in three weeks is a market vulnerable to the slightest counter-signal. The opposite argument, however, deserves its own nuance. Even without a Fed hike, the level of US rates, 3.50-3.75% against 1% in Japan, is enough to keep the pressure on the yen, and US inflation at 4.2% forbids the Fed from cutting quickly. The dollar can stop rising without the yen ceasing to suffer. For the central banks, the real adversary is not the dollar's peak, it is the duration of the plateau. Sources 1. MUFG Research, Monthly Foreign Exchange Outlook, July 2026: DXY at 101.155 end June (+2.3% on the month), USD/JPY at 162.53, Q3 forecasts (DXY 99.77, EUR/USD 1.16), interventions of 11.7 trillion yen, September BoJ hike priced at ~5 bps, Japanese Finance Minister's remarks of 30 June: https://www.mufgresearch.com/fx/monthly-foreign-exchange-outlook-july-2026/ 2. Federal Reserve, FOMC statement of 17 June 2026, rates held at 3.50-3.75%: https://www.federalreserve.gov/newsevents/pressreleases/monetary20260617a.htm 3. Yahoo Finance, June 2026 dot plot: end-2026 median at 3.8% against 3.4% in March, nine participants projecting at least one hike: https://finance.yahoo.com/economy/policy/article/fed-dot-plot-almost-half-of-fomc-members-project-at-least-one-interest-rate-hike-this-year-183645064.html 4. StockTitan, Fed decision of 17 June 2026, hike probability by December up from about 24% to 77% on futures: https://www.stocktitan.net/articles/fed-rate-decision-june-17-2026 5. HSBC Private Bank, "Hawkish stance from Warsh's FOMC surprises markets": https://www.privatebanking.hsbc.com/wih/investments-Insights/market-update/hawkish-stance-from-warshs-fomc-surprises-markets/ 6. Morningstar, "Will the US Dollar Rally Continue?", DXY at a 13-month high of 101.8, +3% since January: https://www.morningstar.com/economy/will-us-dollar-rally-continue 7. Trading Economics, DXY at 100.91 on 7 July 2026: https://tradingeconomics.com/united-states/currency 8. Nikkei Asia, Japanese intervention of 30 April 2026, yen briefly brought from 160 toward 155, first operation since July 2024: https://asia.nikkei.com/business/markets/currencies/japan-launches-fx-intervention-briefly-pushing-yen-to-155-from-160 9. CNBC, BoJ hike to 1.00% on 16 June 2026, highest in over 30 years, 11.7 trillion yen ($73.5 billion) of interventions in May, Goldman Sachs USD/JPY revision to 165 on 6 July: https://www.cnbc.com/2026/06/16/boj-rate-hike-historic-inflation.html 10. Yahoo Finance, Citigroup: no new Japanese intervention expected before the 160-162 zone: https://finance.yahoo.com/markets/currencies/articles/japan-likely-hold-off-fresh-132155677.html 11. ING Think, "JPY intervention: unilateral or joint will be key": https://think.ing.com/articles/jpy-intervention-unilateral-or-joint-will-be-key/ 12. Cambridge Currency, ECB hike of 11 June 2026 to 2.25%, first since 2023, meeting of 23 July, euro toward 1.14: https://cambridgecurrencies.com/euro-forecast/ 13. Bureau of Labor Statistics, Employment Situation for June 2026 (published 2 July): +57,000 jobs, unemployment at 4.2%, participation at 61.5%: https://www.bls.gov/news.release/archives/empsit07022026.htm 14. Yahoo Finance, June jobs report: consensus of 115,000, cumulative revisions of -74,000 over April and May: https://finance.yahoo.com/economy/articles/u-jobs-report-june-2026-123456841.html 15. TradingKey, market reaction to the jobs report, gold beyond $4,130, hike expectations cooled: https://www.tradingkey.com/analysis/economic/indicators/262007217-us-june-jobs-shock-payrolls-add-just-57k-tradingkey 16. l0g, guide "Reading the dollar: DXY, cross-currency basis and offshore dollar debt": https://l0g.fr/en/guides/read-dollar-dxy-cross-currency-basis/ ============================================================================ ANALYSIS: De-dollarisation: the narrative versus the numbers URL: https://l0g.fr/en/analysis/de-dollarisation-narrative-vs-numbers/ Canonical French source: https://l0g.fr/posts/dedollarisation-recit-vs-chiffres/ Date: 2026-07-13 (reviewed 2026-07-13) Topics: macro, dollar, reserves ---------------------------------------------------------------------------- The term "de-dollarisation" has settled into macro commentary as a self-evident fact. Set against the primary data, it nonetheless describes reality badly: the dollar's share erodes slowly, mostly through FX effects, no rival currency emerges, and the only clear move is elsewhere, in gold. A quantified overview. The word and the thing A de-dollarisation worthy of the name would require two things simultaneously: a durable retreat of the dollar in reserves, payments, FX and invoicing, and the rise of a credible substitute. The dominant narrative takes both for granted. The official series tell a more precise, and much less spectacular, story. Reserves: a slow erosion, and mostly an accounting one The dollar's share of global FX reserves stood at 56.92% in the third quarter of 2025 per the IMF's COFER, against nearly 71% in 2000. The long-term decline is real: about fourteen points in twenty-five years. But its quarterly reading invites confusion. The IMF showed that nearly 92% of the decline recorded in the second quarter of 2025 was due to FX effects, not to dollar sales: since reserves are valued in dollars, an appreciating euro mechanically swells the euro's share and reduces the greenback's, without any central bank having moved. At constant exchange rates, the dollar's share had barely fallen. The CEPR went the same way in May 2026: the aggregate often reflects the concentration of large holders and valuation, more than arbitrages against the dollar, whose determinants remain classic, trade exposure and external-debt composition. The yuan is not taking over For there to be de-dollarisation, there would need to be a successor. Yet the currency designated by default, the yuan, does not play that role. Its share of reserves reached 1.93% in the third quarter of 2025, down from 1.99% the previous quarter. In international payments, it weighed 2.73% globally in December 2025 and 2.13% excluding the euro zone, in sixth place, SWIFT itself noting that the Chinese currency "is losing ground". Above all, when the dollar cedes share in payments, it is the euro that recovers most of it, about 60% per SWIFT, followed by the pound, the yen and the Swiss franc. The yuan is not the beneficiary of the move. Diversification goes toward a basket of secondary currencies, COFER's "other currencies" item, which rose to 20.82%, aggregating the Australian dollar, Canadian dollar, Swiss franc, won or Nordic crowns. The plumbing stays in dollars Beyond reserves, the system's infrastructure remains dollarised. The greenback handled about 59% of cross-border payments excluding the euro zone at the end of 2025, far ahead of any other currency, with a modest pullback on the order of 1.8 points in two years. On the FX market, the dollar featured on one side of nearly 88% of transactions per the BIS's latest triennial survey, and it remains the dominant invoicing currency of trade and commodities. None of these compartments shows a shift. What really moves: gold The only clear move is elsewhere. Central banks bought more than 1,000 tonnes of gold a year in 2022, 2023 and 2024, with a record of 1,082 tonnes in 2022, a level unseen since 1950, about double the pace of the 2010-2021 decade. The year 2025, at 863 tonnes, marks a slowdown linked to record prices, but stays well above the historical average, with twenty-three buying countries in the first half. Under the combined effect of purchases and the metal's surge, gold represented about 17% of global reserves at the end of 2024 and approached a quarter at the end of 2025. The drivers are documented: the freeze of about $300 billion of Russian assets in 2022 acted as a signal, and the work of Arslanalp, Eichengreen and Simpson-Bell establishes a link between sanctions exposure and a rise in the gold share. Repatriation follows the same logic: 68% of central banks now store the bulk of their gold on their own soil, against about half in 2020. Poland illustrates the trend, with 550 tonnes representing nearly 28% of its reserves, and a target raised to 30%. Diversification, not substitution Put end to end, the data draw a diversification at the margin, toward gold and a basket of secondary currencies, driven by sanctions risk and by valuation, not the coordinated abandonment of the dollar nor the advent of a rival. The greenback's dominance crumbles at the edges, slowly, without being replaced. "De-dollarisation" over-uses a word for a narrower phenomenon: a hedge, not a rupture. The day the yuan, or any other candidate, durably crosses the threshold of payments and reserves, the term will become accurate. The 2026 figures are not there. --- Primary sources: IMF, COFER (Q3 2025) and blog "Dollar's Share of Reserves Held Steady When Adjusted for FX Moves" (October 2025); CEPR, "The dollar's status through the lens of foreign exchange reserves" (May 2026); SWIFT, Global Currency Tracker (January 2026); BIS, Triennial Central Bank Survey (2022); World Gold Council, Gold Demand Trends; Brookings, "How important are central bank holdings of gold?" (February 2026); Arslanalp, Eichengreen and Simpson-Bell. ============================================================================ ANALYSIS: Gold and central banks: the silent de-dollarisation counted in tonnes URL: https://l0g.fr/en/analysis/central-banks-gold-de-dollarisation/ Canonical French source: https://l0g.fr/posts/or-banques-centrales-dedollarisation-tonnes/ Date: 2026-07-13 (reviewed 2026-07-13) Topics: macro, central banks, hard commodities, geopolitics ---------------------------------------------------------------------------- While attention is on rates and inflation, a slower shift is playing out in central banks' vaults. Since 2022, they have been buying gold at a pace unseen since the end of the Bretton Woods system, and they kept it up in 2025 despite record prices. The metal has quietly overtaken the euro to become the world's second reserve asset, behind only the dollar. This is not a speculative rush, it is a strategic reallocation, partly driven by the fear of sanctions, and partly invisible. Here is that move put into data, with its figures, its blind spots and its limits. A central bank holds reserves to defend its currency, settle its trade and guard against shocks. For decades, these reserves were mostly made of currencies, the dollar in the lead, placed in US Treasury bonds. Gold, judged cumbersome and yieldless, had been relegated to the rank of relic. Since the mid-2010s, and abruptly since 2022, this hierarchy is reversing. The metal is becoming a leading monetary asset again, not for its yield, but for what it protects against. The return of gold to reserves The scale of the move reads in the annual flows. Between 2010 and 2021, central banks bought on average 473 tonnes of gold a year. Then the pace doubled: 1,136 tonnes in 2022, the highest level since records began in 1950, followed by 1,051 tonnes in 2023 and 1,045 tonnes in 2024, three consecutive years above the thousand-tonne mark. In 2025, purchases slowed to 863 tonnes, down 21% year on year and the lowest since 2021, but this figure remains close to double the average of the previous decade, and is still the fourth-largest annual expansion ever recorded. This 2025 slowdown is explained mainly by the surge in prices, which made buyers more cautious, without denting their underlying interest. The fourth quarter confirmed it, with a rebound to 230 tonnes. In all, central banks now hold around 36,000 tonnes of gold, a level close to the peak of 38,000 tonnes reached in the mid-1960s, at the apogee of Bretton Woods. The symbol is strong: official vaults are returning toward the highs of an era when the value of currencies was directly backed by the metal. Who is buying The move is concentrated, and largely led by emerging economies. In 2025, Poland was the top buyer for the second year running, with 102 tonnes added, taking its reserves to 550 tonnes, about 28% of its total reserves, its governor having mentioned a target of 700 tonnes. Kazakhstan added 52 tonnes, its strongest annual rise since 1993. The Czech Republic bought gold for the thirty-fourth consecutive month, reaching 72 tonnes, with a target of 100 tonnes in 2028. Turkey added 27 tonnes, and Brazil returned to the market with 43 tonnes, taking its reserves to 172 tonnes. China's case deserves particular attention. The People's Bank of China declared a rise of 27 tonnes in 2025, taking its official reserves to 2,306 tonnes, a little under 9% of its total reserves, and its fourteenth consecutive month of declared purchases. But these official figures are widely suspected of understating reality, several independent analyses pointing to far higher holdings. Beyond the biggest buyers, a long tail of institutions kept adding gold in small quantities. The World Gold Council's annual survey confirms this dynamic: 95% of surveyed central banks expect a rise in global gold reserves over twelve months, a record, and 43% plan to increase their own holdings, against 29% a year earlier, none anticipating a reduction. Gold, the world's second reserve asset The most spectacular shift is one of level, not flow. According to the European Central Bank's report on the international role of the euro, published in June 2025 on end-2024 data, gold represented about 20% of global official reserves at market value, overtaking the euro for the first time, at about 16%, and sitting just behind the dollar, at about 46%. The trajectory is clear: gold's share was only 11.6% in 2018. Combined with record prices, the accumulation propelled the metal to the rank of second reserve asset in the world. A necessary precision imposes itself here, because two measures coexist and are often confused. The ECB statistic covers total reserves, currencies and gold included, and it is in this frame that gold overtakes the euro. The IMF's COFER base, for its part, measures only the currency composition of FX reserves, excluding gold: in that perimeter, the dollar still weighs around 58%, against nearly 70% in 2000. Both readings say the same underlying trend, the slow erosion of the dollar's place, but they do not refer to the same denominator. The nuance matters, and it distinguishes a gradual reallocation from a collapse that did not happen. Why: diversification and geopolitical shield The engine of this accumulation has changed nature. Historically, the price of gold moved inversely to real rates, rising when inflation ate away at currencies. This relationship broke after Russia's invasion of Ukraine in 2022. Now, the metal follows the logic of geopolitical risk more than that of inflation alone. The ECB's survey of central banks confirms it: two-thirds hold gold to diversify, and two-fifths to protect against a geopolitical risk. The deep reason holds in one word, sanctions. A US Treasury bond can be frozen, access to the interbank payment network can be cut, but a tonne of gold stored in a national vault cannot be confiscated remotely. Gold is an asset with no counterparty and no issuer, independent of any political authority. The ECB notes, moreover, that in five of the ten largest annual rises in gold's share of a country's reserves since 1999, that country had been sanctioned in the same year or the previous one. Russia accumulated more than half of the rise in its official gold reserves since 2014, in full de-dollarisation policy after the annexation of Crimea. Turkey, India and China have, between them, added more than 600 tonnes since the end of 2021. Gold has become again an insurance against being cast out of the dollar-dominated financial system. The blind spot: opaque purchases All this runs into a measurement problem, and it is precisely this journal's terrain. A considerable share of purchases escapes official declaration. For 2025, the gap between the estimates of specialist firms and the publicly reported data suggests that about 57% of central-bank purchases stayed opaque. In other words, some institutions add gold to their reserves without declaring it immediately, a recurring practice in recent years, often to avoid disturbing the market or to preserve strategic room. This opacity has a direct consequence: the real reallocation move toward gold is probably larger than the public figures say, and its mapping remains incomplete. China is the most discussed example, its official declarations looking cautious against independent estimates. Making this hidden share visible is measuring more accurately the real speed of diversification away from the dollar. The limits of the narrative The point remains not to over-interpret. The dollar is not dethroned. It still weighs about 46% of total reserves and 58% of FX reserves alone, and it still dominates global trade, debt and the funding markets described in our series on the plumbing of the dollar. Gold, despite its rise, remains a distant second. The buying pace, moreover, slowed 21% in 2025, and prices above $4,000 an ounce could keep tempering demand. The metal pays no yield, costs to store and insure, and the ECB itself notes that its supply could respond elastically to sustained demand, via the stocks already above ground. This is therefore a gradual diversification, driven by caution and geopolitics, not a sudden monetary shift. It is exactly the nuance we develop in our analysis de-dollarisation, narrative versus numbers, which extends our work on the dollar's international role and its offshore markets. The right reading is neither the triumphalism of the end of the dollar, nor the denial of a real move. It is a slow shift, massive in cumulative terms, partially invisible, and which says less about a new faith in gold than about a growing distrust of a reserve system that can be frozen with a stroke of the pen. The lasting lesson is here. When dozens of central banks start, year after year, to prefer a yieldless asset to the debt of the world's leading power, they are not expressing a market bet, but an insurance premium against political risk. The price of this insurance is counted in tonnes, and much of it is paid in silence. --- Primary sources: World Gold Council, Gold Demand Trends, full year 2025 and monthly central-bank statistics (net purchases of 863 tonnes in 2025, 2010-2021 average of 473 tonnes, records of 1,136 tonnes in 2022, 1,051 in 2023 and 1,045 in 2024, unreported share of about 57%, annual central-bank survey); European Central Bank, "The international role of the euro" (June 2025) and the box "Gold demand: the role of the official sector and geopolitics" (gold's share at about 20% of official reserves end 2024, ahead of the euro, behind the dollar; link between sanctions and accumulation); International Monetary Fund, COFER base (dollar share of FX reserves) and International Financial Statistics; Arslanalp, Eichengreen and Simpson-Bell, "Gold as international reserves: a barbarous relic no more?" (2023). Figures and dates verified one by one; reported data are subject to revision and to a share of unreported purchases. ============================================================================ ANALYSIS: US inflation: the March 2026 energy shock, read in the numbers URL: https://l0g.fr/en/analysis/us-inflation-comeback/ Canonical French source: https://l0g.fr/posts/inflation-us-grand-come-back/ Date: 2026-07-13 (reviewed 2026-07-13) Topics: macro, inflation, fed ---------------------------------------------------------------------------- In March 2026, the US consumer price index jumped to 3.3% year on year, its highest level since April 2024. Almost all the rise comes from energy, propelled by the war in Iran. Beneath this surface, inflation excluding energy and food stayed calm. This distinction is not a technical detail: it is the essential of the analysis. The March figure According to the Bureau of Labor Statistics, in its report of 10 April, the CPI rose 0.9% on the month in seasonally adjusted data, after 0.3% in February, taking annual inflation to 3.3% against 2.4% the previous month. The driver is unequivocal: energy rose 10.9% on the month, its strongest monthly rise since September 2005, and gasoline 21.2%, the strongest monthly increase ever recorded since the series began in 1967. On its own, gasoline explains nearly three-quarters of the monthly rise in the headline index. Beneath the surface, the core stays calm It is the report's second figure that says the most. Excluding energy and food, core inflation rose only 0.2% on the month and 2.6% year on year, a tenth of a point below consensus. Housing, the heaviest item in the index, rose 0.3% on the month and 3.0% year on year, its slowest annual pace since August 2021. Food was flat on the month. The reading that imposes itself is therefore that of a concentrated energy supply shock, not a generalised re-acceleration of prices. The nuance matters because the two situations call for different responses. An oil shock acts like a tax on the consumer: it mechanically weighs on purchasing power, more so on lower-income households for whom gasoline represents a higher share of the budget, but it does not reflect an overheating of demand. In a central bank's analytical framework, such a shock is treated differently from inflation driven by services or wages. Why the distinction matters The risk lies in the transmission. An energy shock, even temporary, can diffuse to the core of the index over six to nine months, as in 2022, through transport and production costs. The favourable scenario has it fade before contaminating the core, which was precisely on the way to moderation before March. The March report does not settle between these two trajectories; it only sets the stakes. For households, the effect is immediate: real hourly wages fell 0.6% on the month. For the Federal Reserve, whose new chair Kevin Warsh argued for lower rates, the rise in inflation combined with a resilient labour market made a quick cut unlikely, markets pricing almost none for 2026. Update, since March The shock did not close in a month. Annual inflation kept rising, to 3.8% in April then 4.2% in May, energy remaining the main contributor, joined by a slight firming of housing and tariff effects. The end of the war, sealed by a US-Iran memorandum on 17 June, and the ensuing pullback in oil should lighten the energy component in the coming months. For the source of the shock, see the state of the Strait of Hormuz. --- Primary sources: US Bureau of Labor Statistics, Consumer Price Index (March, April and May 2026 reports); EIA, Short-Term Energy Outlook (April 2026); analyses by the Center for Commercial Agriculture (Purdue) on the transmission of the shock. ============================================================================ ANALYSIS: The Fed's balance sheet, Kevin Warsh's first battlefield URL: https://l0g.fr/en/analysis/warsh-and-the-fed-balance-sheet/ Canonical French source: https://l0g.fr/posts/warsh-bilan-qt/ Date: 2026-07-13 (reviewed 2026-07-13) Topics: fed, monetary policy, macro, us politics ---------------------------------------------------------------------------- On rates, Kevin Warsh is a prisoner: of inflation at 4.2%, of the Iranian energy shock, and of his eighteen FOMC colleagues who leave him only one vote. It is elsewhere that he will leave his mark. The 17th chair of the Federal Reserve, in office since 22 May 2026, has made the central bank's balance sheet the heart of his doctrine. Yet he inherits a paradoxical situation: the committee has just closed the very file he would like to reopen. Where the balance sheet stands A reminder of the orders of magnitude. The Fed's balance sheet went from about $900 billion before 2008 to a peak of $8.97 trillion in April 2022, swollen by the massive pandemic purchases. Since then, the Fed has contracted it through quantitative tightening, by letting a capped amount of securities mature each month without replacing them. As of 3 June 2026, total assets stand at about $6.71 trillion, of which $4.47 trillion of Treasuries and $1.97 trillion of MBS, the mortgage-backed securities. Two facts frame all the rest. First, over the three and a half years of QT, the Fed has erased only about half of the pandemic expansion. Its share of nominal GDP has fallen from 33% to 20%, but the balance sheet remains more than seven times its pre-2008 level. Second, and this is the essential, this QT is over. On 29 October 2025, under Powell, the FOMC announced the halt of securities reduction as of 1 December. Since then, the Fed reinvests the full principal of maturing Treasuries and shifts the principal of MBS toward Treasury bills. The balance sheet therefore no longer contracts in aggregate. The reason for this halt is not ideological, it is plumbing. Bank reserves, around $2.89 trillion at the end of 2025, were approaching the threshold below which funding markets seize up, estimated between $2.5 and $2.7 trillion. The memory of September 2019, when the previous slimming cure triggered a spike in repo rates and forced the Fed to re-inject liquidity on an emergency basis, still haunts the committee. The reverse repo facility, that cushion of excess liquidity that exceeded $2.3 trillion at the end of 2022, is today almost empty. In other words, the cushion has melted, and the committee judged there was no margin left to continue without risk. What Warsh wants Warsh's doctrine runs against this status quo. As early as July 2025, on CNBC, he called for a regime change and denounced a credibility deficit among the incumbent leaders. His vision of the balance sheet rests on three points. First point: substitute active reduction for the mere passive runoff. Where Powell waited for securities to mature, Warsh raised the possibility of selling assets, a faster and more brutal instrument, never used by the Fed in its two QT episodes. Second point: dismantle the Fed's status as buyer of last resort in the bond market. Warsh reproaches the central bank for having become a permanent, distorting presence in the Treasury market. He advocates a durable retreat of this footprint. Third point: explicit coordination with the Treasury. Warsh argued for the Fed chair and Treasury Secretary Scott Bessent to announce a balance-sheet size target to markets together. This idea breaks with the tradition of independence and blurs the border between monetary policy and debt management, which worries some observers. The balance-sheet trap Warsh's problem is that he wants to reopen a door the committee closed for good technical reasons. Relaunching an aggregate balance-sheet contraction, all the more through active sales, would amount to pushing reserves below the stress threshold, with a risk of repeating the 2019 episode. No chair, however hawkish, wishes to inaugurate his term with a liquidity crisis in the most important market in the world. But the balance sheet also offers Warsh a weapon he does not have on rates. Reducing the balance sheet, or even just slowing its recomposition, tightens financial conditions through the long-rate channel, without touching the policy rate. It is a disguised tightening. For a chair caught between an inflation he judges he must fight and a US president who demands rate cuts, it is precious room for manoeuvre: acting on the balance sheet lets him hold a restrictive line without displaying the loathed word of a hike. The real playing field is on MBS. Part of the committee has long held that the Fed has no business holding mortgage securities, which amounts to indirectly subsidising housing credit. The plan set in December already provides for letting the MBS run off and replacing them with short-term Treasury bills. That is where Warsh can accelerate without triggering a liquidity stress, since the total balance sheet stays stable: he changes the composition, not the size. Shrinking the MBS pocket, still close to $2 trillion, and shortening the portfolio's maturity toward T-bills, that is the reform he can lead this very year, with the assent of a fraction of the committee. Why this is the real subject The balance sheet is the silent lever. Markets scrutinise the dot plot and the press conference, but it is in the composition and trajectory of the balance sheet that the reality of the Warsh mandate of the coming months will play out. Three structural tensions will knot there. The first pits doctrine against plumbing. Warsh wants a smaller Fed; reserves say there is almost no room to reduce without breaking something. The second pits the Fed against the Treasury. A state whose financing needs are exploding has an interest in the central bank remaining a stable buyer of its debt; a Fed that retreats by selling its Treasuries pushes up the government's borrowing cost at the worst moment. The third pits displayed independence against advocated coordination. By calling to set the balance-sheet target hand in hand with Bessent, Warsh risks turning a monetary-policy tool into an instrument of public-debt management, exactly the reproach the hawks addressed yesterday to the pandemic-era Fed. The orientation of the coming months therefore reads thus. On rates, an imposed status quo. On the balance sheet, no brutal relaunch of QT, too dangerous, but a targeted offensive on MBS and portfolio maturity, presented as a technical normalisation while it carries a strong doctrinal intention. Warsh will not be able to make the Fed as small as he dreams. He can, on the other hand, make it more discreet, shorter in duration, and less present in the mortgage market. It is less spectacular than a regime change. It is already a change of regime. Sources 1. Federal Reserve, Policy Normalization, end of runoff on 1 December 2025, decline of more than $2.2 trillion since June 2022 ($1.6trn of Treasuries, $600bn of MBS), share of GDP from 33% to 20%: https://www.federalreserve.gov/monetarypolicy/policy-normalization.htm 2. Congressional Research Service (Congress.gov), balance sheet from $8.9 trillion in 2022 to $6.5 trillion in 2025, QT ended in December 2025, about half of the pandemic expansion reversed: https://www.congress.gov/crs-product/IF12147 3. StreetStats, balance-sheet composition on 3 June 2026: $6,711bn of assets, $4,469bn of Treasuries, $1,965bn of MBS: https://streetstats.finance/liquidity/fed-balance-sheet 4. PrimeRates, peak of $8.97 trillion in April 2022, total at $6.66 trillion, reverse repo near-emptied: https://primerates.com/primerate/fed-balance-sheet/ 5. Wolf Street, 4 December 2025, total QT of $2.43 trillion over three years and five months, MBS at $2.05 trillion, plan to replace MBS with T-bills: https://wolfstreet.com/2025/12/04/fed-balance-sheet-qt-37-billion-in-november-2-43-trillion-from-peak-to-6-54-trillion/ 6. Banking Exchange, 26 November 2025, halt of Treasury runoff on 1 December, reserves at $2.89 trillion, stress threshold $2.5-2.7 trillion, MBS runoff continues: https://www.bankingexchange.com/news-feed/item/10480-treasury-market-resilience-and-the-early-end-to-balance-sheet-runoff 7. Federal Reserve / Benzinga, 30 October 2025, end of QT announced, balance sheet at $6.59 trillion, rate lowered to 3.75-4.00%: https://benzinga.com/markets/macro-economic-events/25/10/48506713/federal-reserve-decision-october-30-25-basis-point-fed-funds-rate-balance-sheet-qt-quantitative-tightening 8. CNBC, 17 July 2025, Warsh on regime change, the credibility deficit, coordination with the Treasury on the balance-sheet target: https://www.cnbc.com/2025/07/17/kevin-warsh-touts-regime-change-at-fed-and-calls-for-partnership-with-treasury.html 9. QuantSandbox, peak of $8.95 trillion in April 2022, equilibrium balance-sheet estimates of $6.0-6.5 trillion, debate on holding MBS: https://quantsandbox.orientedplatforms.com/learn/fedbalancesheet 10. Federal Reserve, Warsh's swearing-in as 17th chair, 22 May 2026: https://www.federalreserve.gov/newsevents/pressreleases/other20260522a.htm ============================================================================ ANALYSIS: US regional banks: from the 2023 panic to liquidity reform URL: https://l0g.fr/en/analysis/us-regional-banks-liquidity-lcr/ Canonical French source: https://l0g.fr/posts/banques-regionales-us-liquidite-lcr/ Date: 2026-07-13 (reviewed 2026-07-13) Topics: macro, central banks, liquidity, regulation, markets ---------------------------------------------------------------------------- March 2023: three failures in seven weeks Silicon Valley Bank was closed on 10 March 2023 by its state regulator, which named the FDIC as receiver. The bank showed $209 billion of assets at the end of 2022, making it then the second-largest bank failure in US history. The mechanism was not a borrower default: the rapid rise in rates in 2022 melted the market value of its portfolio of Treasuries and mortgage securities. By selling part of these securities at a loss to meet withdrawals, SVB triggered a crisis of confidence. Its clients, mostly tech companies with largely uninsured deposits, withdrew more than $42 billion in a single day. Speed, not size, was the real novelty. Signature Bank followed two days later. On 1 May, First Republic, $229 billion of assets and about two-thirds uninsured deposits, was seized then sold to JPMorgan, becoming the largest failure since 2008. SVB's cost to the FDIC guarantee fund is estimated at about $20 billion. The response: BTFP and stabilisation On 12 March 2023, the Fed created the BTFP, an emergency facility lending for one year against Treasuries and agency MBS valued at par, thus disregarding unrealised losses. At its peak, the outstanding exceeded $165 billion. Coupled with the exceptional protection of SVB and Signature's uninsured depositors, the measure stopped the contagion. The programme stopped lending on 11 March 2024 and was repaid in full a year later. No significant-sized bank has defaulted since. The underlying fragility did not vanish for all that: in early 2024, New York Community Bancorp wobbled on its commercial real-estate exposure, a reminder that unrealised losses and property risk remain on the regionals' balance sheets. Liquidity today: ample reserves, QT over The quantitative tightening begun in June 2022 has ended: the FOMC stopped shrinking its balance sheet on 1 December 2025. The Fed's balance sheet went from about $8.9 trillion in 2022 to nearly $6.5 trillion at the end of 2025, its securities falling by more than $2.2 trillion. In December 2025, the Fed judges reserves to have returned to an "ample" level and resumes Treasury-bill purchases, so-called reserve-management purchases, to keep them there; the standing repo facility moves to full allotment on 10 December, and the RRP has fallen back near zero. The system is therefore officially in ample liquidity, not scarce. This plumbing is detailed in our guide on Treasury liquidity, our read of the Fed balance sheet and our analysis of the repo market. Residual strains have not vanished: Jerome Powell warned there would be further failures among small and mid-sized banks, against the backdrop of commercial real estate and growing competition from non-bank actors. Liquidity reform, the 2026 agenda On 10 February 2026, the Fed, the OCC and the FDIC repealed their FAQs on the LCR and announced they would put regulatory changes out to consultation. On 3 March, at a roundtable in Washington, the Fed's vice-chair for supervision, Michelle Bowman, and Treasury Secretary Scott Bessent, made the case for easing. Their argument: the current framework encourages liquidity hoarding. Banks hold high-quality liquid assets, HQLA, well above the minimum, because the LCR cushion is in practice unusable: a bank refuses to go below 100% and internally aims for 115 to 120%. As a result, per Bessent, about 25% of big banks' balance sheet is tied up in safe assets, against nearly 10% before 2008, so much credit forgone. The path proposed: recognise in the LCR the borrowing capacity at the discount window against pre-positioned collateral, up to a cap (the industry mentions about 20%), and reduce the stigma attached to that window. FDIC chair Travis Hill defends the same idea, as well as a revision of the NSFR. The counterpoint is substantial. The 2023 failures showed that a run can play out in hours, well within the LCR's 30-day window, and that monetising assets on an emergency basis has its limits. Some therefore argue for a complementary five-day ratio. At this stage, nothing is set in stone: the work is at the speeches-and-roundtable stage, with no formal rule proposal published. The direction, though, is clear: make the cushions usable and depend less on a static stock of HQLA. Reading 2023 came down to a poorly hedged duration risk, volatile uninsured deposits and supervision gaps. 2026 flips the question: the regulator now worries that the framework forces banks to tie up liquidity and lend less, while runs have become faster. Fed staff estimate that lowering internal LCR targets by ten points would free up about $350 billion of HQLA. The whole trade-off is there: make liquidity usable without reopening the door to flash panics. --- Primary sources: FDIC, "Lessons Learned from the U.S. Regional Bank Failures of 2023" and SVB material-loss review (Office of Inspector General, September 2023); Federal Reserve, page and FEDS note on the Bank Term Funding Program, policy-normalisation statements (end of tightening, announcement of 29 October 2025), Michelle Bowman speech "Liquidity resiliency, financial stability, and the role of the Federal Reserve" (3 March 2026), repeal of the LCR FAQs (10 February 2026), FEDS Note "The Central Bank Balance-Sheet Trilemma" (14 January 2026) and St. Louis Fed; Department of the Treasury, Scott Bessent's remarks at the liquidity roundtable (3 March 2026); FDIC, Travis Hill's remarks on reforming the regulatory toolkit (2026); Bank Policy Institute for BTFP statistics (peak above $165 billion); Reuters for the programme closure timeline. Figures and dates verified one by one. ============================================================================ ANALYSIS: The Treasury basis trade: the leveraged arbitrage at the heart of US debt URL: https://l0g.fr/en/analysis/the-treasury-basis-trade/ Canonical French source: https://l0g.fr/posts/basis-trade-treasuries-levier/ Date: 2026-07-13 (reviewed 2026-07-13) Topics: macro, markets, central banks, regulation ---------------------------------------------------------------------------- There is an arbitrage that, on its own, links the repo market, Treasury futures and the stability of the US debt market. It is called the basis trade. In normal times, it brings cash and futures prices together, and supplies liquidity to a $29 trillion market. Under stress, the same mechanism reverses: its high leverage forces unwinds, which amplify the fall. It contributed to the Treasury debacle of March 2020, and its size has kept growing since. A direct sequel to our piece on repo and collateral, here is the anatomy of a trade regulators watch like a hawk. The basis trade is an arbitrage on the basis, the price gap between a cash Treasury bond and the corresponding futures contract. Futures generally trade a little rich to cash, because asset managers buy futures massively to get duration exposure without holding the securities. A fund captures this gap by putting on three simultaneous legs: it buys the bond in cash, sells the futures contract, and finances the purchase in repo by pledging the bond as collateral. At expiry, the basis converges to zero and the gap is pocketed. The mechanics, and the leverage The captured gap is tiny, on the order of a few basis points. For the trade to be profitable, it therefore needs enormous size and high leverage. The reference studies, including that of the Treasury Borrowing Advisory Committee, use leverage of around 20 times. The initial margin on futures is only 2 to 3%, and much of the repo funding is done at near-zero haircut. According to the OFR, about 74% of hedge funds' repo borrowing was done at zero or negative haircut, which multiplies the leverage and the liquidation risk. Concretely, a $100 million position may tie up only $7 to $8 million of its own capital. What it really does This trade is not just speculation. It exists because asset managers have a structural demand for long futures, to manage the duration of their portfolios, and someone must take the other side. Leveraged hedge funds fill this role, and by hedging through the basis trade, they link cash and futures prices and provide liquidity, including on older, less-traded issues. In short, the Treasury market has come to depend on heavily indebted actors to run smoothly. It is the same logic as in repo: liquidity is manufactured on constrained balance sheets. The size, and why it is poorly measured No one knows the exact figure, because the trade is not reported as such. It is approached through two proxies: the net short positions of leveraged funds in Treasury futures, published by the CFTC, and the sponsored-repo volumes tracked by the OFR. The IMF's April 2026 Global Financial Stability Report thus estimates the size of the trade at around $1 trillion, after rapid growth. The short positions of leveraged funds in the 2-, 5- and 10-year contracts exceeded $1 trillion as early as March 2025. Relative to a $29 trillion market, the figure looks modest, but the risk does not lie in the volume, it lies in the leverage and the forced nature of the exits. The breaking point: the margin spiral The fragility is known and documented. When volatility rises, two things happen at once: clearing houses raise futures margins, and repo lenders raise their haircuts. The fund must then post cash on an emergency basis. If it cannot, it unwinds, so it sells its cash bonds, which weighs on prices and further feeds volatility. It is the margin spiral described by the BIS as early as 2020. In March 2020, the unwinding of the basis trade represented nearly half of hedge funds' Treasury sales per the OFR's work, and the Fed had to intervene massively to stabilise the market. What mandatory clearing will change Regulators have not stood by. Fed governor Lisa Cook again described, on 20 November 2025, hedge funds' Treasury positions as a systemic vulnerability, liable to make the market more vulnerable to stress. But the main lever is elsewhere: mandatory central clearing. The SEC rule, adopted in late 2023 then pushed back a year, now requires clearing of cash Treasury transactions by 31 December 2026, and of repo by 30 June 2027. The scope of clearing houses has been widened, with the approval of CME in December 2025 and ICE in early 2026, and a new FICC service to limit double margining. The effect is double-edged. By interposing a clearing house, you reduce counterparty risk and make exits more orderly. But by imposing margins where repo was done at zero haircut, you make the trade more expensive and reduce its profitability, which could shrink its size, or push it toward less-regulated corners. The lasting lesson matches that of repo: the deepest market in the world leans on a heavily indebted arbitrage that lubricates it in calm times and drains it in a crisis. As long as this trade exists at this scale, the stability of US debt depends, in part, on the ability of a handful of funds to hold their positions when volatility runs away. --- Primary sources: IMF, Global Financial Stability Report (April 2026, size of the trade estimated at around $1 trillion via CFTC and OFR proxies); CFTC, Market Risk Advisory Committee, "The Treasury Cash-Futures Basis Trade and Effective Risk Management" (10 December 2024); Office of Financial Research, Hedge Fund Monitor and cleared-repo collection; Federal Reserve, FEDS Notes on hedge-fund positions and the Barth and Kahn note; BIS, Schrimpf, Shin and Sushko, "Leverage and Margin Spirals in Fixed Income Markets during the COVID-19 Crisis" (2020); remarks by governor Lisa Cook (20 November 2025); SEC, Treasury clearing rule and compliance timeline (extensions of 25 February 2025, deadlines of 31 December 2026 and 30 June 2027); Federal Reserve Bank of Chicago on the effect of the clearing mandate. Figures and dates verified one by one. ============================================================================ ANALYSIS: US debt: the awakening of the term premium URL: https://l0g.fr/en/analysis/the-return-of-the-term-premium/ Canonical French source: https://l0g.fr/posts/dette-americaine-le-reveil-de-la-prime-de-terme/ Date: 2026-07-13 (reviewed 2026-07-13) Topics: macro, debt, rates, us treasury, risk ---------------------------------------------------------------------------- import InfographicBars from '../../components/infographic/InfographicBars.astro'; Some risks do not explode, they settle in. The US fiscal trajectory is one of them: slow, predictable, and for that very reason regularly ignored. Yet three developments converge in 2026 and deserve to be looked at together rather than separately. Debt crosses a historic threshold, the interest bill becomes a major spending item, and the market begins again to demand a premium to lend long term. This last point, technical in appearance, is the most important, because it touches the very price of the debt. The weight of the numbers Let us start with the base, relying on the projections of the Congressional Budget Office, Congress's independent budget body. The federal deficit for fiscal year 2026 is expected at $1.9 trillion, or 5.8% of gross domestic product, an unusual deficit level outside recession or war. Debt held by the public would go from 101% of GDP in 2026 to 120% in 2036, thus exceeding its previous record of 106%, reached in 1946 at the end of the Second World War. The real regime change is elsewhere, in the interest bill. The CBO puts net interest at $1 trillion in 2026, or 3.3% of GDP, and projects it will reach $2.1 trillion, or 4.6% of GDP, in 2036. This interest now exceeds all defence spending, and would surpass total discretionary spending in 2038. In other words, a growing share of taxation serves not to fund services, but to pay creditors. It is the first gear: the higher debt and rates rise, the more the interest bill swells the deficit, which in turn feeds the debt. The term premium, that forgotten price of risk The yield on a ten-year bond breaks into two bricks. The first is the average of expected short rates over the security's life, what the Federal Reserve will do. The second is the term premium, the extra yield an investor demands to tie up their money for a long time and bear the risk that rates, inflation or debt supply move unfavourably. The reference model, that of Adrian, Crump and Moench at the New York Federal Reserve, allows this premium to be estimated. And it has just woken up. Negative for the first time in the series in 2014, then staying zero or negative for a decade, the ten-year term premium has moved back into positive territory, around half a point in mid-2026, for the first time since 2023. At the end of April 2026, the model decomposed a ten-year yield of 4.45% into 3.72% of rate expectation and 0.73% of term premium. This point is decisive for two reasons. First, this level of 0.73 point stays modest against history, below the 1.41-point median over sixty-five years. The premium is therefore not at an extreme, it has simply turned positive again, which leaves room to rise if debt supply keeps swelling. Second, in a regime of positive term premium, Fed rate cuts no longer translate mechanically into an easing of long rates, because the premium can rise when rate expectation falls. The central bank's leverage over the real cost of the debt is thereby weakened. Who buys, now that the Fed and foreigners are retreating A price rises when demand weakens against supply. Yet the supply of Treasuries is abundant and the structure of demand is degrading. The Federal Reserve has stopped being a buyer with the end of quantitative tightening, a subject we treated in our coverage of the Fed's balance sheet under Warsh. Foreign holders, for their part, hold about $8.5 trillion of Treasuries, or 28 to 30% of the marketable debt, with Japan in the lead at $1.13 trillion, ahead of the United Kingdom and China. But while their holdings rise in dollars, their share is falling, because the debt grows faster than they do, a dynamic legible in the TIC data and consistent with the gradual move of de-dollarisation. The clearest signal came from the agencies. In May 2025, Moody's stripped the United States of its last AAA rating, downgrading it to Aa1 and joining S&P and Fitch, citing gross debt of $36 trillion and an interest bill absorbing 18% of federal revenue. The thirty-year yield had then briefly exceeded 5%. Who fills the void? Increasingly, marginal and fragile buyers. Stablecoins, framed by the GENIUS Act, must back their reserves with very short-term Treasury bills, of 93 days at most, and repos of less than seven days. They therefore support the short end of the curve, not the long end, the one where the premium forms. Hedge funds, for their part, carry enormous positions through the basis trade, a highly leveraged arbitrage between the cash bond and the futures contract. The Financial Stability Board recalls that the precipitous unwind of $90 billion of these positions contributed to the Treasury market crisis of March 2020, and notes that in the first quarter of 2025 hedge-fund leverage reached a historic high. Demand increasingly provided by leveraged actors is less stable demand. The spectre of fiscal dominance All these threads converge toward the same worry, fiscal dominance: the situation where the weight of the debt constrains monetary policy, the central bank hesitating to raise or hold high rates for fear of rendering the debt unsustainable. The debate is not theoretical. It runs through Kevin Warsh's Federal Reserve, caught between an inflation it judges it must fight and an executive demanding rate cuts. A rising term premium is precisely the symptom that a market is starting to doubt the state's ability, or willingness, to stabilise its debt without resorting to inflation. Why the worst is not written Rigour requires weighing the objections, and they are serious. The first is the exorbitant privilege of the dollar. The United States borrows in its own currency, which it issues, and therefore cannot default in the strict sense on debt denominated in dollars. The Treasury remains the safe-haven asset par excellence, bedrock of the global financial system, and the structural demand for safe dollar assets is immense. As long as this status holds, the tolerance threshold for US debt is higher than that of any other state. The second objection is that high debt does not mechanically entail a crisis. Japan has proven it for decades, with debt well above 200% of GDP without a major episode of distrust, because it is financed by abundant domestic savings. The third rests on the arithmetic of sustainability: as long as nominal growth exceeds the average interest rate paid on the debt, the debt-to-GDP ratio can stabilise without violent fiscal effort. Finally, the term premium, as we have seen, stays moderate, and auctions keep finding takers, with even a cautious return of foreign buyers in early 2026. The market has often wrongly announced the imminent revenge of the creditors. These arguments bound the risk, they do not erase it. The dollar's privilege reduces the probability of a brutal crisis, it does not abolish the cost of an interest bill that crowds out other spending. The Japanese counter-example recalls that domestic financing changes everything, which conversely underlines US vulnerability to foreign demand. And the sustainability condition, growth above the rate, is nothing guaranteed in a regime of a rising term premium. Following the risk, indicator by indicator The subject cannot be read from a single figure, but from a dashboard. The New York Fed's ACM term premium says whether the market demands more for duration. Auction quality, the bid-to-cover ratio and the share of indirect buyers, signals the depth of demand. The thirty-year yield and the MOVE index, which measures rate volatility, capture stress. TIC data traces the behaviour of foreign holders. The interest bill relative to revenue measures the crowding-out effect. And debt-ceiling episodes, with the rebuilding of the Treasury's account at the Fed, punctuate the calendar with liquidity jolts, a theme developed in our guide on Treasury liquidity. The honest conclusion is measured. There is no programmed imminent crisis, and betting on its trigger at a given date would be as imprudent as denying the problem. But the direction is clear: debt supply swells, demand grows fragile, and the price of duration risk, long anaesthetised, is waking up. The real danger is not a sudden crash, it is the durable installation of a higher cost of capital, which weighs on everything, from the federal budget to asset valuations. A slow risk, then, but one that deserves for precisely that reason a sustained watch. Sources 1. Congressional Budget Office, "The Budget and Economic Outlook: 2026 to 2036": 2026 deficit of $1.9 trillion (5.8% of GDP), net interest of $1 trillion in 2026 (3.3% of GDP) to $2.1 trillion in 2036 (4.6%), debt held by the public from 101% of GDP in 2026 to 120% in 2036, exceeding the 106% record of 1946, interest exceeding defence: https://www.cbo.gov/publication/62105 2. Federal Reserve Bank of New York, Treasury term premia (Adrian, Crump, Moench model): yield decomposition, premium back in positive territory: https://www.newyorkfed.org/research/dataindicators/term-premia-tabs 3. FRED (Federal Reserve Bank of St. Louis), series THREEFYTP10, ten-year term premium: historical levels, turn negative in 2014, return to positive: https://fred.stlouisfed.org/series/THREEFYTP10 4. Moody's Ratings, downgrade of the US sovereign rating from Aaa to Aa1, 16 May 2025: debt and interest bill cited, alignment with S&P and Fitch: https://ratings.moodys.com/ratings-news/443154 5. CNBC, 19 May 2025, thirty-year yield briefly exceeding 5% after the Moody's downgrade: https://www.cnbc.com/2025/05/19/us-treasury-yields-moodys-downgrades-us-credit-rating.html 6. U.S. Department of the Treasury, Treasury International Capital (TIC) system, foreign holdings of Treasuries (about $8.5 trillion, Japan, United Kingdom, China) and falling share: https://home.treasury.gov/data/treasury-international-capital-tic-system 7. Congress.gov, GENIUS Act of 2025, reserve obligations of stablecoin issuers in short-maturity Treasury bills and repos: https://www.congress.gov/bill/119th-congress/senate-bill/1582/text 8. Financial Stability Board, vulnerabilities of repo markets and non-bank leverage: unwind of $90 billion of basis trade in March 2020, call to limit hedge-fund leverage: https://www.fsb.org/uploads/P040226.pdf 9. Federal Reserve, FEDS Notes, sizing of hedge-fund Treasury positions and the basis trade: https://www.federalreserve.gov/econres/notes/feds-notes/recent-developments-in-hedge-funds-treasury-futures-and-repo-positions-20230830.html 10. Peter G. Peterson Foundation, tracking of the federal debt interest bill: https://www.pgpf.org/programs-and-projects/fiscal-policy/monthly-interest-tracker-national-debt/ ============================================================================ ANALYSIS: The yen carry trade: the fuse is now in Japanese bonds URL: https://l0g.fr/en/analysis/the-yen-carry-trade/ Canonical French source: https://l0g.fr/posts/carry-trade-yen-la-meche-dans-les-obligations-japonaises/ Date: 2026-07-13 (reviewed 2026-07-13) Topics: yen, boj, carry trade, jgb, rates, macro, markets ---------------------------------------------------------------------------- The yield on the Japanese 10-year government bond reached 2.88% on 9 July 2026, its highest level since 1996, at the end of nine straight sessions of gains. A bond move in Tokyo may seem distant. Yet it touches one of the most discreet cogs of global finance: the yen carry, the cheap funding that irrigates risk assets across the planet. The danger has changed address. Until now, Japanese risk read off the exchange rate: a USD/JPY glued to 162, a threat of intervention from Tokyo, the fear that a forced repurchase of yen would trigger a disorderly exit from the carry. That grid stays valid, but a second front has opened, and it comes from the bond market. Japanese long rates are rising, fast, and this rise attacks the carry through a channel other than the exchange rate. Here is why it matters. The yen carry, how it works The yen carry trade rests on a simple idea. You borrow in the world's cheapest currency, the yen, whose rates stayed close to zero for twenty years, and place the proceeds in better-remunerated assets: US Treasuries, equities, credit, emerging currencies, sometimes crypto. The gain is the yield gap, the carry. As long as the yen stays weak and markets are calm, the position banks a steady income with comfortable leverage. Its size is hard to quantify because it mixes bank, speculative and corporate positions. Estimates range from a few hundred billion dollars for the speculative core alone to an order of magnitude of $4 trillion for the broader web, depending on the perimeter used. This imprecision is itself information: no one knows the exact size of the short yen position, which makes any unwind impossible to calibrate in advance. The carry works like an implicit sale of volatility: it earns bit by bit, then can cost dearly all at once. Two conditions support it, a yen that does not rebound sharply and a Japanese yield that stays negligible. The second has just cracked. The new trigger comes from the bond market On 9 July 2026, the 10-year JGB yield reached 2.880%, its highest since September 1996, its ninth consecutive session of gains, the longest run in nineteen years. The move is sharper still at the long end of the curve: the 30-year rose to 4.030% and the 20-year to 3.85%, an unprecedented peak since 1996 too. For a market used to floor-glued rates, this is a regime change. The immediate cause is fiscal. The government unveiled a long-term economic strategy aiming to mobilise more than 370 trillion yen, about $2.29 trillion, of public and private investment by fiscal year 2040 to strengthen strategic industries. The market translated this programme into a single question: who will finance this extra debt, and at what price? The answer is being written in the auctions, where investors now demand a higher yield to absorb Japanese paper. The local press even coined a word for the episode, the "Honebuto shock", after the government's fiscal framework. Three channels of contagion Why do rising Japanese rates threaten assets on the other side of the world? Through three distinct channels, which can play together. The first is the compression of the carry. The carry lives off the gap between a near-zero yen funding cost and a high foreign yield. If the Japanese yield climbs, the opportunity cost of the strategy rises: keeping your savings in Japan becomes less penalising, borrowing in yen more so. The relative advantage of the trade narrows from below, even before the yen moves. The second is repatriation. Japan is a major creditor to the rest of the world and, per US Treasury TIC data, the largest foreign holder of Treasuries. Japanese life insurers, pension funds and banks hold mountains of foreign bonds. The day the 10-year JGB offers a decent yield in local currency, with no exchange risk, part of these savings can flow back to Tokyo. This repatriation pushes selling of foreign assets, including Treasuries, which tightens their yields and squeezes global liquidity. It is the same interconnected world described in our guide to the Treasury market: exchange rate, sovereign debt and market funding are not separate subjects. The third is the central bank's trap. The BoJ remains the largest holder of JGBs, with a share only just back below 50% of the market in early 2026, against a peak of 53% in 2023. It is now shrinking its balance sheet through quantitative tightening, preferring, per Wolf Street, this route to aggressive rate hikes to support the yen. But this withdrawal has a mechanical effect: the less the BoJ buys, the fewer captive buyers there are, and the more long rates rise. The central bank is thus caught between two dangers, letting rates run at the risk of a bond accident, or buying back at the risk of reviving yen weakness. Each option feeds a facet of the carry risk. August 2024, the quantified precedent A brutal unwind is no textbook hypothesis. The market saw it on 5 August 2024, when a first BoJ tightening and poor US data triggered a flash rebound in the yen and a chain liquidation. The Bank for International Settlements made it the subject of a dedicated bulletin. The lesson of this episode is twofold. First, the transmission is mechanical: to meet margin calls on FX losses, investors sell what they can, that is, the most liquid assets, including those with nothing to do with the yen. Second, and this is the essential, the 2024 tremor settled only an estimated 10% or so of a position core on the order of $500 billion. The rest stayed in place. The carry did not disappear in 2024; it was reloaded, and today it lives under the threat of a new trigger. The possible paths Three outcomes emerge. These are analyst scenarios, not forecasts, and the order in which I present them does not prejudge their probability. The first path is an orderly retreat. The BoJ calibrates its quantitative tightening, long rates rise in digestible steps, and the carry shrinks gradually as the yield gap closes. Investors have time to unwind without trampling each other. It is the scenario the central bank seeks, and its prudent management so far, a yen defended by the balance sheet rather than by abrupt hikes, points that way. The second is the trap the BoJ would like to avoid. If long rates run away, the temptation will be strong to slow the withdrawal, or even to buy back JGBs to calm the curve, a de facto return toward yield curve control. But each such move weakens the yen, revives the appeal of the carry and pushes the problem back while enlarging it. The central bank would buy time against an even heavier short position. The third is the accident. An auction that goes badly, a fiscal slip or a confidence shock, and long rates tighten too fast. Repatriation kicks in, the yen firms all at once, volatility explodes and the unwind becomes forced, as in August 2024, but on a broader position base. This scenario is not the most probable, it is the most costly, and it is the one to guard against. Why the worst is not written It would be dishonest to present only the alarmist thesis. Several factors argue for a landing without drama. Japanese debt is held more than 90% by residents, which makes a buyers' strike far less likely than in a country dependent on foreign capital: an Italian- or British-style failed auction remains improbable in Japan. The BoJ retains, with yield curve control, an instrument able to cap long rates overnight if it deems it necessary. And the 2024 episode itself showed a capacity for rapid rebound: once the shock passed, markets and the yen had stabilised within a few weeks, without a lasting systemic crisis. Above all, a rise in Japanese rates can reflect good news, the finally successful exit from three decades of deflation, rather than a distress signal. A Japan normalising its rates because its economy holds up is a healthier Japan, even if the transition roughs up a speculative trade. The nuance has its flip side: at 2.88%, the Japanese yield stays far below the 4% and more of Treasuries, so the carry keeps a margin and its disappearance is nothing imminent. The risk is not that the carry collapses tomorrow. It is that a poorly measured stock of positions unwinds one day in disorder, and that the fuse, this time, was lit in Tokyo, on the bond market, where few people were looking. Sources 1. Business Recorder / Reuters, 10-year JGB yield at 2.880% on 9 July 2026, highest since September 1996, ninth session of gains (longest run in 19 years), 30-year at 4.030%, 20-year at 3.85%: https://www.brecorder.com/news/40429220/japan-benchmark-bond-yield-extends-rise-after-hitting-30-year-high 2. Yahoo Finance / Reuters, benchmark JGB at a 30-year high amid fiscal concerns: https://finance.yahoo.com/economy/policy/articles/japan-benchmark-bond-yield-hits-004445377.html 3. StoneX, Japanese investment plan of more than 370 trillion yen ($2.29 trillion) by fiscal 2040, pressure on the curve and carry trades: https://www.stonex.com/en-us/insights/japan-yield-curve-pressure-threatens-global-carry-trades/ 4. Trading Economics, 10-year JGB yield series: https://tradingeconomics.com/japan/government-bond-yield 5. Bank of Japan, monetary policy decision of 16 June 2026, policy rate raised to 1.00%: https://www.boj.or.jp/en/mopo/mpmdeci/index.htm/ 6. Nippon.com, BoJ share of the JGB stock (peak of 53.34% in March 2023): https://www.nippon.com/en/japan-data/h01720/ 7. Nikkei Asia, BoJ share of JGB holdings back below 50% in early 2026: https://asia.nikkei.com/business/markets/bonds/bank-of-japan-s-share-of-jgb-holdings-dips-below-50 8. Wolf Street, the BoJ favours quantitative tightening over rate hikes to support the yen, 3 July 2026: https://wolfstreet.com/2026/07/03/qt-instead-of-rate-hikes-to-put-a-floor-under-plunging-yen-bank-of-japan-sheds-15-6-of-its-massive-assets/ 9. Bank for International Settlements, Bulletin no 90, "The market turbulence and carry trade unwind of August 2024" (Nikkei -12.4%, VIX beyond 65, unwind of about 10% of positions estimated at $500 billion): https://www.bis.org/publ/bisbull90.pdf 10. U.S. Treasury, TIC data, major foreign holders of Treasuries: https://ticdata.treasury.gov/resource-center/data-chart-center/tic/Documents/slttable5.html ============================================================================ ANALYSIS: The cross-currency basis: the hidden price of the dollar, when the law of international finance breaks URL: https://l0g.fr/en/analysis/the-cross-currency-basis/ Canonical French source: https://l0g.fr/posts/cross-currency-basis-prix-cache-dollar/ Date: 2026-07-13 (reviewed 2026-07-13) Topics: macro, central banks, liquidity, markets ---------------------------------------------------------------------------- There is in international finance a rule so solid it is compared to a physical law: covered interest parity. It says that borrowing dollars directly or manufacturing them through another currency should cost the same, otherwise a riskless arbitrage would erase the gap. Since 2008, this law no longer holds. Manufacturing dollars via an FX swap costs a premium, sometimes a few basis points, sometimes much more under stress. This premium has a name, the cross-currency basis, and it has become the most reliable thermometer of strains in dollar funding. A sequel to our piece on eurodollars, here is the mechanics of this gap and its reach. Covered interest parity, or CIP, rests on a simple idea. An investor holding a dollar can invest it for a year at the dollar rate. Or they can convert it into euros spot, invest those euros at the euro rate, and lock in today the reconversion rate into dollars a year out through a forward contract. In theory, both paths must return exactly the same. If they did not, anyone could borrow via the cheaper path, lend via the other, and pocket a riskless margin. This arbitrage should bring the gap to zero instantly. The law that no longer holds Before 2008, that was the case, and the basis hovered around zero for the major currencies. Since the financial crisis, the gap has settled durably, oscillating between 20 and 100 basis points, and much more during stress episodes. Concretely, a negative basis on the euro or the yen means that manufacturing dollars through an FX swap costs more than borrowing them directly. The party that wants dollars pays a premium, and the party that supplies them gets a discount. The near-physical law of international finance has become a pricing of the dollar by its scarcity. Why arbitrage no longer closes the gap If a riskless margin exists, why do banks not erase it? Because the arbitrage is no longer either riskless or free. The work of the BIS and the Fed converges on two causes. First, balance-sheet constraints: since 2008, the leverage ratio and risk-weighted capital requirements make the mere act of expanding a balance sheet to capture the gap costly. The arbitrageur must tie up capital, and demands to be paid for it. Second, the imbalance in hedging demand: Japanese insurers and European pension funds structurally need dollars they obtain through swaps, which pushes the gap one way, with no symmetric counterparty to fill it. These two forces explain why the basis widens especially on balance-sheet dates. At quarter- and year-ends, banks cut their market-making to lighten their balance sheet, liquidity grows scarce, and the gap widens brutally on small demand shocks. It is the same balance-sheet-constraint mechanism described for repo, applied this time to the FX market. The dollar's thermometer This gap has become the indicator central banks and treasurers watch, because it sums up in one number the strain on dollar funding outside the United States. When it widens, it is because the offshore dollar is short, intermediaries' balance sheets are saturated, and hedging demand overwhelms supply. When the Federal Reserve activated its swap lines with the other central banks, in 2008 then in March 2020, the gap narrowed almost at once, a sign that the shortage was indeed of dollars, not of solvency. The cross-currency basis is therefore the displayed price of the plumbing described in our earlier pieces. It puts a number on the cost, at a given moment, of access to the dollar for the rest of the world. A rule reputed unbreakable has turned into a permanent barometer, and as long as dollar demand stays structurally imbalanced and bank balance sheets constrained, this gap will remain, in normal times as in crisis, one of the most honest signals of the global financial system. --- Primary sources: Bank for International Settlements, Borio, McCauley, McGuire and Sushko, "Covered interest parity lost: understanding the cross-currency basis" (Quarterly Review, September 2016) and BIS Working Papers no 590, "The failure of covered interest parity: FX hedging demand and costly balance sheets"; International Monetary Fund, Dao and Gourinchas, "Covered Interest Parity in Emerging Markets" (Working Paper, 2025) and WP/19/169 on CIP deviations in Asia (Korean won up to about 597 basis points in 2008); Federal Reserve, FEDS, "Quantities and Covered-Interest Parity" (2024); Du, Tepper and Verdelhan on post-crisis deviations. Figures and markers verified one by one. ============================================================================ ANALYSIS: Eurodollars: the offshore dollar, the debt no one sees in full URL: https://l0g.fr/en/analysis/eurodollars-the-offshore-dollar/ Canonical French source: https://l0g.fr/posts/eurodollars-dollar-offshore-dette-cachee/ Date: 2026-07-13 (reviewed 2026-07-13) Topics: macro, central banks, liquidity, markets ---------------------------------------------------------------------------- The most important dollar for global financial stability is not the one that circulates in the United States, but the one that lives outside it. Japanese banks, European insurers, emerging-market companies, pension funds: all borrow, lend and fund themselves in dollars without ever touching American soil. This system, the eurodollar, weighs tens of trillions, and a large part escapes both balance sheets and statistics. When it seizes up, the Federal Reserve must play firefighter to a blaze it can barely see. After repo and shadow banking, here is the offshore floor of the dollar's plumbing. A eurodollar is a dollar held outside the US banking system, with no link to the euro currency. The term dates from the 1950s and 1960s, when dollar deposits accumulated in London, both to escape American regulation and because some actors preferred to keep their dollars out of Washington's reach. The prefix simply designates an external dollar. Since then, this market has become the international layer of the dollar: non-American banks accept dollar deposits and lend in dollars, creating dollar credit entirely outside the borders and beyond the Fed's direct reach. The visible part: offshore dollar credit The BIS, the only institution to map this system, tracks the dollar credit granted to non-bank borrowers located outside the United States. At the end of 2025, this outstanding reached $14.3 trillion, up 8.5% year on year, the strongest increase since 2014, carried by a weak dollar. For comparison, euro credit outside the euro zone stood at €4.9 trillion, up 11%, while yen credit outside Japan fell 4.9% after the unwinding of the carry trade. Dollar credit to emerging economies alone reached $4.3 trillion, against $3.2 trillion a decade earlier. Taken together, international credit, cross-border and in foreign currency, represents about 38% of world GDP. It is the measurable part, the one that appears in bank balance sheets and recorded bond issuance. It is already considerable. Yet it is only the visible tip. The missing debt: FX swaps Beneath this visible part hides a far larger mass, lodged in foreign-exchange swaps. In a swap, a Dutch pension fund or a Japanese insurer borrows dollars and lends euros or yen on the way out, then does the reverse on the way back. The operation resembles a repo, but with a currency for collateral. The crucial difference: these dollar payment obligations are recorded off balance sheet, in an accounting blind spot. They do not appear in classic debt statistics. The BIS has tried to size this missing debt. In its reference estimate, on 2022 data, non-banks located outside the United States owed close to $26 trillion through these instruments, double their on-balance-sheet dollar debt, and up sharply from the $17 trillion of 2016. For non-American banks, this off-balance-sheet amount exceeded $39 trillion, more than ten times their equity. Most of this debt is very short term, which creates permanent refinancing needs and, therefore, a squeeze risk at every strain. Why it is a systemic risk The problem is not size in itself, but the combination of short-term funding, an opaque off-balance-sheet, and total dependence on a dollar issued elsewhere. When conditions tighten, holders of dollars outside the United States all seek to refinance at the same time, and the offshore dollar becomes brutally scarce. That is what happened in 2008, then in March 2020: the FX swap market froze, and the Federal Reserve had to open swap lines with the major central banks to re-inject dollars into the global system. Yet the Fed then acts as firefighter to a blaze it can barely see. As the BIS stresses, the authorities intervened in 2008 and 2020 with little information on who owed what and where. The lender of last resort of the global dollar steers partly blind, on a debt that appears nowhere in the balance sheets it supervises. It is precisely the zone of opacity this journal seeks to illuminate. The offshore floor of the same machine The eurodollar does not live apart. It funds itself on the same markets as repo, it houses part of non-bank intermediation, and offshore centres like the Cayman Islands concentrate a growing share of cross-border credit. It is the international layer of one and the same system, where liquidity is manufactured on constrained, leveraged balance sheets, and where an ever-larger part escapes measurement. The question is not whether this offshore dollar is systemic, it obviously is, but how much longer we will accept supervising it with statistics that ignore half of it. --- Primary sources: Bank for International Settlements, global liquidity indicators, data at end-December 2025 (dollar credit to non-banks ex-US at $14.3 trillion, strongest growth since 2014; euro and yen credit outside their zone); BIS, Quarterly Review of December 2022, Borio, McCauley and McGuire, "Dollar debt in FX swaps and forwards: huge, missing and growing" (off-balance-sheet dollar debt estimated at about $26 trillion for non-banks ex-US and more than $39 trillion for non-American banks, on 2022 data); BIS, "FX swaps and forwards: missing global debt?" (Quarterly Review, September 2017); McGuire and von Peter, "The US dollar shortage in global banking and the international policy response" (BIS Working Papers, 2009). Figures and dates verified one by one. ============================================================================ ANALYSIS: The silent contagion: how private credit weaves an invisible web across banks, insurers, equities, crypto and stablecoins URL: https://l0g.fr/en/analysis/the-silent-contagion-of-private-credit/ Canonical French source: https://l0g.fr/posts/la-contagion-silencieuse-credit-prive/ Date: 2026-07-13 (reviewed 2026-07-13) Topics: finance, private credit, macro, crypto ---------------------------------------------------------------------------- Methodological note: this article is the result of an analytical conversation. The figures cited were verified against several independent sources (institutions, financial media, market publications) wherever possible. The orders of magnitude are robust. The scenario probability estimates are by nature subjective and do not constitute investment advice. A market that has never lived through a full cycle The dominant reassuring narrative, carried among others by Jamie Dimon, CEO of JPMorgan, and David Solomon, CEO of Goldman Sachs, during the October 2025 quarterly releases, boils down to one argument: private credit is too small to be systemic. The argument is lazy. It is lazy for three reasons this article proposes to dismantle methodically. First, raw size has never been the right indicator of a financial risk. Subprime weighed about $1.3 trillion in 2007 and triggered the worst financial crisis since 1929. Second, private credit has long ceased to be an isolated parallel universe: it is financed, guaranteed, refinanced and redistributed through channels that pass through regulated banks, private-equity-controlled life insurers, pension funds, listed BDCs and, the latest bridge, the crypto ecosystem via yield-bearing stablecoins and tokenised private credit. Third, and this is probably the most important point, private credit at its current size and scope has never been tested during a severe economic downturn, as the Financial Stability Board explicitly recalls in its May 2026 report. This article proposes as exhaustive a map as possible of the potential contagion channels. It draws on reports from the IMF, the US Federal Reserve, the Financial Stability Board, on the quarterly releases of the big banks, on the Tricolor and First Brands episodes of late 2025, on the mechanics of the USDe depeg in October 2025, and on recent figures from the Japanese bond market. The question is not to predict a Lehman-type collapse, which is unlikely in the short term. The question is to understand the transmission chain that makes this market far less isolated than the official line suggests. --- Part 1: The question of size, a battle of definitions When the French press reports that "private credit weighs $2 trillion", that figure is defensible but incomplete. The real range, according to the institutions that publish serious estimates, stretches from $1.5 trillion to more than $30 trillion depending on the definition used. | Source | Estimate | Scope used | | ------------------------- | -------------------------------------------- | -------------------------------------------------------------- | | IMF (April 2024) | $2.1 trillion | AUM + committed capital, of which ~75% in the United States | | Fed (FEDS Note, May 2025) | $1.34 trillion US; ~$2 trillion globally | Mainly direct lending, up fivefold since 2009 | | FSB (May 2026) | $1.5 to $2 trillion | Attempted harmonised definition across jurisdictions | | AIMA / ACC (Dec. 2024) | $3 trillion | Includes asset-backed, real-estate debt, infrastructure debt | | Wellington (Dec. 2025) | $30 trillion | Potential addressable market, not the existing market | The gap is not a battle of figures between experts but a definition problem. The Financial Stability Board, in its report published on 6 May 2026, acknowledges it explicitly: the absence of a harmonised definition of private credit across actors and authorities prevents a proper assessment of the market at the global level. The low range (IMF, Fed) counts essentially corporate direct lending. The high range (AIMA, Alternative Credit Council) adds asset-backed loans, real-estate debt and infrastructure debt, which today represent about 40% of the private-credit market according to AIMA. The growth is without recent equal. Over the last five years, the US private-credit market has gone from about $500 billion to $1.3 trillion, a 2.6-fold multiplication, while Moody's forecasts it will exceed $3 trillion of assets under management by 2028. Wellington values the potential addressable market at more than $30 trillion, mainly by moving beyond the perimeter of traditional leveraged corporate credit to encompass asset-based lending, commercial real estate and AI-linked infrastructure financing. This dizzying growth since the 2008 crisis is explained by a simple mechanism: Basel III made direct loans prohibitively costly in capital for banks, which gradually delegated that risk to specialised actors. Apollo, KKR, Blackstone, Blue Owl, Ares or Carlyle have built a colossal alternative market. The five largest listed managers now manage about $1.5 trillion in perpetual capital, or about 40% of their combined AUM, against 35% in 2021. If growth is maintained at the pace observed since 2021, they will manage nearly $5 trillion in permanent capital by the end of the decade. This concentration is not trivial. The larger the market grows, the further down the credit spectrum funds go to find borrowers. The IMF estimates that more than a third of borrowers today carry interest charges higher than their current earnings. In other words, more than a third of the market no longer generates enough cash flow to cover its debt service in the current rate environment. That is colossal and historically unprecedented. --- Part 2: The five bridges to the regulated banking system The myth of "isolated" private credit rests on a superficial reading of bank balance sheets. In reality, there are at least five documented transmission channels between private-credit funds and regulated banks. 1. Subscription credit lines and NAV loans This is the most direct channel. Banks lend massively to private-credit funds through subscription lines (backed by LPs' uncalled commitments), NAV loans (backed by the fund's net value), and various leverage facilities. According to a Fed note published in May 2025, bank commitments to BDCs (business development companies) rose about 186% over five years, by far the strongest increase among all NBFIs (Non-Bank Financial Institutions). US commercial banks' loans to non-depository financial institutions (NDFIs) now represent about 13% of total loans and leases. On the disclosed figures: three of the six largest US banks reported, in their Q4 2025 results, financial exposure of about $108 billion to private credit or related loans. Three banks only. Extrapolated to the entire US banking system, aggregate exposure is probably on the order of $300 to $400 billion. 2. Syndicated and structured asset-backed loans On structured financing (asset-backed loans, securitisation vehicles, CLOs), banks and private funds coexist in the same structure, often without full transparency on who holds which tranche. It is the very mechanism that trapped JPMorgan and Fifth Third on Tricolor in September 2025. Tricolor, a subprime lender specialised in auto for Texan Latino communities, filed for bankruptcy on 10 September 2025 in a liquidation proceeding. JPMorgan booked $170 million of charge-offs in the third quarter of 2025 linked to wholesale loans granted to Tricolor. Fifth Third announced a write-down of between $170 and $200 million on an asset-backed loan to Tricolor. Jamie Dimon called the episode "not our finest moment" and produced the now-famous line: "when you see one cockroach, there's probably more". 3. Private-equity-controlled life insurers This is probably the most dangerous and least discussed channel in the mainstream press. Apollo owns Athene, KKR owns Global Atlantic, Brookfield controls American Equity Life. These large life insurers have migrated a substantial share of their bond portfolio into private credit, often originated by their parent. The IMF stresses in its April 2024 Global Financial Stability Report that a selected group of pension funds and insurers are wading deeper into private credit, significantly increasing their share of these less liquid assets. This includes private-equity-influenced life insurers. The transmission mechanism under stress is known: a life insurer taking losses on its private-credit portfolio must, to meet its solvency ratios, sell other, more liquid assets, typically Treasuries and investment-grade corporate bonds. This immediately pollutes public bond markets. 4. Semi-liquid and evergreen funds sold to retail Since 2022, the large private-credit managers have massively developed so-called "semi-liquid" vehicles aimed at retail and wealth management, Blackstone BCRED, Blue Owl OBDC II, Apollo ATCRED, KKR KCRED. These funds offer a quarterly redemption window, generally capped at 5% of NAV per quarter. Beyond that, the gates close. In the first quarter of 2026, Blackstone BCRED faced $3.7 billion of redemption requests and had to raise its redemption cap from 5% to 7% to manage the flow. Blue Owl Capital's OBDC II fund, with $1.6 billion, permanently suspended redemptions at the end of 2025. These are not isolated incidents. Wellington estimates that US retail allocation to private credit has gone from about $100 billion currently to a projected target of $2.4 trillion by 2030, an annualised growth of nearly 80%. This retailisation creates a psychological risk that did not exist in the institutional market. Institutional LPs accept long lockups and understand illiquidity. Retail, by contrast, panics fast, and as soon as the gates close, distrust spreads to other liquid asset classes, such as listed-BDC shares or crypto. 5. Listed BDCs as an instant barometer Apollo, Ares, Blackstone, Blue Owl, KKR all have their own listed BDCs. When sentiment deteriorates, their shares plunge 15 to 25% in a few weeks, as seen in October-November 2025. This fall immediately contaminates equity sentiment, triggers short-seller analyses (Burry, Ackman, Einhorn have all taken positions or commented), and feeds global distrust of the asset class. The conclusion of this first analysis is clear: the "isolated private credit" thesis is untenable. The bridges to the regulated banking system are numerous, deep and growing fast. The real question is not binary (isolated versus contagious), but rather: how much can these channels self-reinforce in a synchronised recession? --- Part 3: Crypto / private-credit porosity, the invisible bridge The great blind spot of mainstream analysis concerns the bridge between private credit and the crypto / DeFi ecosystem. This contagion channel now exists, and it works both ways. It was mostly set up between 2024 and 2026, and it is recent enough to have escaped the vigilance of traditional regulators. Tokenised private credit has become massive Three platforms structure this market: - Maple Finance: more than $4 billion of AUM in 2026, specialised in structured fixed-rate facilities for crypto-native trading firms. The syrupUSDC product, which distributes private-credit yield as a stablecoin, saw its transfer volume double to $4.98 billion at the end of January 2026. - Centrifuge: pools of loans backed by real-world assets, more than $1.1 billion of active loans, yields between 8% and 12%. Institutional partnerships with Janus Henderson, S&P and BlackRock. - Ondo Finance, Goldfinch, Credix, Huma Finance: a smaller but fast-growing segment, totalling several billion more. In all, per RWA.xyz data of November 2025, active on-chain private credit stands at $18.91 billion, with cumulative originations of $33.66 billion since these protocols launched. The BeInCrypto report of September 2025 indicates that the total tokenised RWA market reached $30.26 billion in 2025, of which $7.3 billion in Treasuries and $15.9 billion in private credit, meaning tokenised private credit already represents more than twice the volume of tokenised Treasuries. The underlying move is explicit: capital is "climbing the yield curve", moving from safe Treasuries (4 to 5% yield) toward riskier private credit (10 to 16%). It is the classic end-of-cycle pattern: the search for yield pushes up the risk spectrum, until the tide goes out and reveals who was swimming naked. The Ethena / USDe case, the most fragile link To grasp the porosity concretely, one must stop at Ethena Labs and its synthetic stablecoin USDe. It is, to my mind, the most important and most misunderstood link in the whole chain. USDe is not a classic stablecoin backed by dollars in a bank or by Treasuries (like USDT or USDC). It is a "synthetic dollar" that replicates the dollar's stability through a delta-neutral strategy: for each dollar of USDe issued, the protocol holds a dollar of crypto (ETH, BTC, stETH) and simultaneously opens a short position of equivalent notional on perpetual futures. When the collateral price rises, the short loses; when the collateral falls, the short gains. The net value in synthetic dollars stays stable. The yield distributed (the famous 9 to 15% offered to holders of sUSDe, the staked version) comes from three sources: the funding rates of the perpetuals (about 92% of the backing, the essential), the rewards of ETH staking (stETH), and the interest on the liquid stablecoins held in reserve (USDC at Coinbase, T-bill exposure via BlackRock's BUIDL fund). Here is the critical angle: Ethena is structurally long crypto bullish sentiment, disguised as a stablecoin. The funding rates of the perpetuals are positive when traders are net-long and pay for that leverage. They turn negative in a deep bear market. Ethena's reserve fund, which serves to absorb periods of negative funding, stands at about $61 million against a supply of $5.6 billion of USDe in the first quarter of 2026, or 1.18% coverage. That is very little. The 10 October 2025 episode On 10 October 2025, Donald Trump's announcement of additional 100% tariffs on Chinese imports triggered a wave of generalised risk-off. In crypto, it became the biggest liquidation event in history: more than $19 billion of leveraged positions liquidated in 24 hours. During the storm, USDe briefly fell to $0.65 on Binance, an apparent depeg of 35%. On Curve, Bybit, Kraken and the DeFi pools, the price stayed around $0.98-0.99. The problem was not a failure of the protocol: Ethena confirmed that its mint and redemption mechanisms continued to work normally, that collateralisation stayed above 100%, and that more than $2 billion of redemptions were processed during the day. The problem was one of infrastructure fragility: Binance's internal oracle read its own order book, only $8 million deep, instead of aggregating external prices. With a USDe supply of $9 billion, that gives a depth ratio of 0.09%. Compared with USDC: $40 billion of supply, $2 billion+ of Binance depth, ratio 5%. In other words, USDe is about 55 times more fragile than USDC in terms of liquidity ratio on the dominant exchange. A single sale of $90 million was enough to move the price by 35%. Lasting consequence: Ethena processed more than $3 billion of redemptions in 8 hours, USDe supply falling from $14.8 billion in October to $7.6 billion at the end of November, a fall of more than 50%. The governance token ENA, for its part, went from an all-time high of $1.52 to about $0.11 on 18 May 2026, a fall of more than 92%. The protocol's proclaimed resilience is real at the level of the central mechanism, but the leverage loops around it (Pendle PT, Aave, Morpho) were massively unwound. The six structural risks of USDe The Q1 2026 report from Stablecoin Insider lists six structural risks of USDe that all materialised at least once in 2024 or 2025: - Durably negative funding rates: if the crypto bear market drags on, Ethena must pay instead of receive. At -10% annualised over 6 months, that represents about $280 million of losses against a $61 million reserve fund, a deficit of 4.6 times the absorption capacity. - Oracle and localised liquidity risk (materialised in October 2025): the dependence on CEX price feeds creates brutal dislocations in case of infrastructure failure. - Exchange counterparty risk: if Binance or Bybit (where Ethena hosts its short positions) experienced operational difficulties, the delta-neutral can break temporarily. - stETH depeg risk: during the Terra crisis in May 2022, stETH depegged to 0.93 ETH. A similar dislocation would widen the gap between the value of the spot leg and the short notional. - Short-squeeze risk on a violent rally: if BTC or ETH pump too fast, the shorts can be forcibly liquidated before Ethena can rebalance. The delta-neutral then becomes directional. - Scaling limit of the perp market itself: USDe cannot grow beyond the capacity of the perpetual markets to absorb massive shorts. At $14 billion at the peak, USDe already represented a significant share of BTC/ETH open interest. The direct channel to TradFi: iUSDe This is where the bridge thickens. In Q1 2026, Ethena launched iUSDe, a wrapped version of sUSDe aimed specifically at regulated TradFi capital, with Kraken Institutional custody, weekly Proof of Reserves, and backing from Franklin Templeton and F-Prime Capital. Concretely, insurers, family offices and pension funds can now indirectly hold a synthetic product backed by crypto funding rates, a "yield-bearing" exposure marketed as prudent. And in the other direction: USDe is now accepted as collateral on Aave V4 ($4.91 billion of TVL on Ethereum/Base on 1 April 2026), on Morpho, on Spark, and on the Pendle PT markets. This allows the creation of leverage loops: deposit USDe, then borrow USDC, then buy more USDe, then re-deposit. In October, these loops accounted for most of the massive outflows observed. --- Part 4: The bond channel, stablecoins, Treasuries and Japanese yields The probably most systemic, and most paradoxical, channel passes through the US Treasury market. It connects private credit, stablecoins, US sovereign yields and the Japanese bond market into a single loop of interdependence. The presence of stablecoins in the Treasury market Stablecoins are no longer peripheral actors. USDT (Tether) is the largest stablecoin in circulation at $186 billion in January 2026, of which 63% of reserves in T-bills according to the BIS. USDC (Circle) holds about 32% of its reserves in T-bills, and about 43% in reverse repos per Circle's disclosures. In Q3 2025, all stablecoins together held about $170 billion in US Treasuries. The Federal Reserve Bank of Kansas City projects that this figure could reach $450 billion by 2028, partly under the effect of the GENIUS Act adopted in July 2025, which mandates payment-stablecoin issuers to hold their reserves in HQLA (High-Quality Liquid Assets), typically T-bills with a maturity below 93 days, cash, and reverse repos. The BIS asymmetry: why an outflow hurts 2 to 3 times more than an inflow BIS Working Paper No 1270 published in 2026 documents a crucial asymmetry in the effect of stablecoin flows on yields: an inflow of 2% of stablecoin capital could lower 3-month yields by 2 to 2.5 basis points, whereas an equivalent outflow could raise them by 6 to 7 basis points. In other words, when stablecoins absorb Treasuries, the effect on yields is moderate. When they sell them, the effect is two to three times more violent. This asymmetry is explained by the non-linear liquidity of the T-bill market: in normal conditions, money market funds absorb flows easily. In stress conditions (bill scarcity, panic), depth vanishes and each marginal transaction impacts the price far more strongly. It is precisely the phenomenon the BIS calls "tail impact non-linearity". The Japanese bond spike, the lit fuse Since November 2025, the Japanese bond market has seen a historic move. The Japanese 40-year yield reached 3.697%, its highest since the instrument launched in 2007. The 30-year touched 3.334%, the 20-year 2.80%, the 10-year 2.80% on 18 May 2026 per Trading Economics. The trigger is twofold: the election of Sanae Takaichi as Prime Minister of Japan in October 2025, followed by the announcement of a fiscal stimulus plan of 17 to 21 trillion yen (about $110 to $135 billion). Markets interpreted this as a renunciation of fiscal discipline by a country indebted at more than 230% of GDP. Goldman Sachs now speaks of the return of a fiscal risk premium on the Japanese bond market. The global implications are multiple and all negative for risk assets: - Unwinding of the yen carry trade: estimated at about $20 trillion per the Kobeissi Letter, this carry trade consisted of borrowing yen at near-zero rates to buy higher-yielding assets everywhere in the world, US equities, corporate credit, real estate, crypto. At 2.8% on the 10-year JGB, the arbitrage becomes marginal. - Repatriation flows: Japan holds about $1.20 trillion in US Treasuries as of 31 October 2025 per US Treasury TIC data, making it the largest foreign creditor of the US government ahead of China. If Japanese institutions (life insurers, pension funds like GPIF) repatriate even 10% of this exposure to capture domestic yields, that is $120 billion of selling pressure on the US market. - Mechanical rise in US yields: fewer foreign buyers means higher yields on Treasuries, which immediately translates into a tightening of global financial conditions, unfavourable to private credit, growth stocks and crypto. The stablecoin / yields loop Here is how the three channels self-reinforce under stress: Imagine a scenario where, following a serious private-credit event (five simultaneous Tricolors, for example), BDCs suffer massive redemptions. Holders seek to exit adjacent positions too, including yield-bearing stablecoins like sUSDe or syrupUSDC. Run on yield-bearing stablecoins. To honour the redemptions, Ethena must unwind its crypto short positions (which temporarily pushes crypto prices up via forced short covering, paradoxically). Maple must sell its underlying private-credit positions, but the secondary market is thin. Ondo and BUIDL must sell Treasuries. At that moment, the BIS asymmetry effect kicks in: what was a moderate upward demand for Treasuries becomes a brutal sale that spikes yields. The yield spike re-marks the entire classic private-credit portfolio lower (borrowers become even more stressed to pay their indexed interest). BDCs suffer new markdowns. Loop. It is this loop that makes the current system particularly hard to model. Traditional regulators (ECB, Fed, FSB) are aware of the problem, but their stress-test tools are essentially banking, they do not capture the stablecoin / crypto channel. --- Part 5: Contagion scenarios, three trajectories On the basis of the preceding map, I propose three distinct scenarios for the next 12 to 24 months, in increasing order of severity. Scenario 1, Bumpy normalisation (high probability) Defaults keep appearing in isolated pockets (two to five "cockroaches" per quarter in US private credit). Losses stay contained at the bank scale: no G-SIB bank is put in difficulty. BDCs keep managing their redemptions through high caps. Ethena goes through a period of weak to occasionally negative funding rates, but the reserve fund holds. The yen does not break brutally, the BoJ intervenes if needed to stabilise the 10-year below 3%. Credit spreads widen moderately, US High Yield goes from 268 bps to 400-450 bps. In this scenario, the private-credit asset class comes out weakened but not broken. Consolidation accelerates: small managers without access to premium deals disappear or are bought. Apollo, Ares, Blackstone, Blue Owl and KKR consolidate their oligopolistic position. The probability that this scenario materialises seems to me the highest, say 55-65%. Scenario 2, Concentrated liquidity stress (medium probability) An unfavourable combination materialises over 3 to 6 months: a mild US recession, simultaneous defa [...] ============================================================================ ANALYSIS: Private credit: one asset, two prices URL: https://l0g.fr/en/analysis/private-credit-one-asset-two-prices/ Canonical French source: https://l0g.fr/posts/credit-prive-un-actif-deux-prix/ Date: 2026-07-13 (reviewed 2026-07-13) Topics: private credit, valuation, BDC, NAV, price discovery, systemic risk ---------------------------------------------------------------------------- A non-traded private-credit fund redeems its units at par while the listed funds that lend to the same companies trade nearly a fifth below their declared value. Two prices coexist for a near-identical credit risk. In 2026, the question is no longer whether they diverge, but which will eventually win. The price you declare, the price you quote Private credit is distributed through two families of vehicle. Non-traded funds, interval funds and non-traded BDCs, redeem their units at net asset value, the NAV. This NAV is not quoted: it is estimated at level 3 fair value, from unobservable inputs, then struck periodically. The Fidelity Private Credit Fund, in its redemption notice filed with the SEC for the second quarter of 2026, describes the mechanism plainly: its NAV is "validated each month by a third-party valuation process", and its direct-lending portfolio showed an average valuation of 98.7% of par in the first quarter. The two families nonetheless lend to the same universe: senior secured loans to mid-sized companies, most often at floating rates. The price gap therefore does not stem first from the nature of the assets, but from the way they are valued. Listed BDCs carry this dual display permanently. On one side you read the NAV published by the manager, on the other the price at which the market trades the share. At the end of February 2026, the listed-BDC index tracked by VanEck traded at about 0.83 times book value, nearly 17% below declared NAV, and about 14% below its historical average of 0.97 times. In its note of 29 June, PIMCO notes that this ratio has hit a local low, but that discounts remain wide and their dispersion is widening. For the listed BDC, the market renders a second opinion every day; for the non-traded one, only the model speaks, and it is at its price that redemption settles. Dispersion betrays the absence of an anchor These marks are not homogeneous. Analysing the documents filed with the SEC for a sample of 32 BDCs, With Intelligence notes that 27 of them saw their NAV per share fall in 2025. On average, the decline reaches 3.8% for the listed ones (median 2.5%) and 1.7% for the non-traded ones (median 2%). But the range is gaping: on the listed side, from Prospect Capital at minus 20.8% to Main Street Capital at plus 5%; on the non-traded side, from Monroe's fund at minus 4.8% to Golub's non-traded fund, GCRED, the only one to rise, by 0.04%. This is the point PIMCO stresses: valuation dispersion is high from one manager to another and, counter-intuitively, stronger still among the non-traded ones, despite smoother declared performance. Low volatility over time combined with high cross-sectional dispersion sits poorly with a common reference price. It suggests that NAVs reflect each manager's own assumptions rather than a market clearing price. The secondary market puts a number on this doubt. The valuation firm Mercer Capital reports that Saba Capital, run by Boaz Weinstein, offered to buy units of several private-credit funds at 20 to 35% below declared NAV; and the aborted merger, at Blue Owl, between a non-traded fund and its listed counterpart exposed the same gap between private marks and public price for comparable assets. The mechanism closes on itself: a holder who can exit at NAV while judging the real value far lower has an interest in requesting redemption before the write-down. The International Monetary Fund named this effect as early as 2024: a "first-mover advantage", when stale valuations let some leave before losses are recognised, at the expense of those who stay. Three paths to a single price According to PIMCO, the gap can close in three ways: through NAV write-downs, through wider discounts on the secondary market, or through realised losses. The first path is already visible. In the first quarter of 2026, the listed BDC FS KKR Capital saw its NAV per share go from $20.89 to $18.83, a 9.9% fall in one quarter, with non-accrual loans at 4.2% of the portfolio at fair value. To stabilise the vehicle, KKR injected $150 million in preferred shares and launched a buyback offer, but at $11 a share, nearly 40% below declared NAV. Public price and model price converge, from below. The bond market ruled earlier: many BDC bonds trade at yield spreads close to the BB segment of the Bloomberg US Corporate index, PIMCO notes. Shareholders doubt the credibility of NAVs; creditors, for their part, separate valuation uncertainty from default risk. What lets the model price hold owes to its plumbing. The Financial Stability Board, in its report of 6 May, flags the opacity of valuations and the reliance on private ratings, sometimes issued by little-known providers. Level 3 fair value leaves a margin of judgement to the manager and the board: smoothed marks support the NAV in the short term, at the cost of transparency. What still holds back the panic The picture calls for a counterweight. Neuberger Berman recalls that the BDC universe, listed and non-traded combined, weighs only about $500 billion, and that about $600 billion of committed but undeployed institutional capital, half of it in direct lending, can absorb part of the outflows. Above all, redemptions do not first reflect poor performance: several large vehicles delivered high single-digit total returns in 2025. The July 2026 filings confirm it: at Goldman Sachs Private Credit, second-quarter redemption requests reached 3.24% of units, below the 5% cap, and were served in full; at Fidelity Private Credit, about 2.9%, also served in full, with positive net inflows. PIMCO insists on a final point: private credit is not one block. The strain concentrates on corporate direct lending, while asset-backed financing, with flows less tied to the earnings cycle, behaves differently; and the discount applied to the listed ones is no longer a uniform macro rebate, the market now differentiating managers by asset quality and the credibility of their marks. The useful question is therefore not the displayed level of NAV, but how the gap between the two prices will close: through an orderly write-down, or through materialised losses. For the mechanics of contagion, see the silent contagion of private credit; for the wider read on private credit, our guide on reading private credit. Sources - PIMCO, The Credit Market Lens: What BDC Redemptions and NAV Pressures Mean for Investors, 29 June 2026 (confidence gap, mark dispersion, paths to convergence, BDC bonds near BB), - VanEck, What is Driving BDC Valuations? (listed BDC index P/B ≈ 0.83x on 27 February 2026, against ≈ 0.97x historical average), - With Intelligence, What is actually going on in BDC portfolios?, 30 April 2026 (NAV analysis on SEC filings at 31 December 2025; PSEC / MAIN / Monroe / GCRED dispersion), - International Monetary Fund, Global Financial Stability Report, April 2024 (chapter on private credit; first-mover advantage tied to stale valuations), - Mercer Capital, Public Prices, Private Marks: What BDC Discounts Are Signaling, 9 April 2026 (Saba Capital offers at 20 to 35% below NAV; aborted Blue Owl merger), - FS KKR Capital Corp., first-quarter 2026 earnings release, 11 May 2026 (NAV of $18.83 against $20.89; non-accruals 4.2%; tender at $11), ; Form 8-K, SEC EDGAR, - Financial Stability Board, Report on Vulnerabilities in Private Credit, 6 May 2026 (valuation opacity, private ratings, PIK), - Neuberger Berman, Private Credit and BDCs: Why the Sell-Off Tells an Incomplete Story, 6 May 2026 (BDC universe ≈ $500bn; ≈ $600bn of dry powder; redemptions unrelated to performance), - Goldman Sachs Private Credit Corp., Form SC TO-I/A, SEC EDGAR, July 2026 (Q2 redemptions at 3.24% of units, below the cap, served in full), - Fidelity Private Credit Fund, Form SC TO-I/A, SEC EDGAR (Q2 redemptions ≈ 2.9% served in full, positive net inflows, average mark 98.7%, NAV validated monthly by third-party valuation), ============================================================================ ANALYSIS: Repo and collateral: where liquidity is made, and where it breaks URL: https://l0g.fr/en/analysis/repo-the-liquidity-factory/ Canonical French source: https://l0g.fr/posts/repo-collateral-fabrique-liquidite/ Date: 2026-07-13 (reviewed 2026-07-13) Topics: macro, central banks, liquidity, markets ---------------------------------------------------------------------------- Beneath the surface of markets, trillions of dollars change hands every day against collateral, for a few hours. This is the repo market, the plumbing that funds Treasuries, dealers and leveraged positions. When it works, no one talks about it. When it seizes up, as in September 2019, short rates break loose within hours and the central bank must inject cash on an emergency basis. In late 2025, after the end of quantitative tightening, the first tremors reappeared. A read of a mechanism where liquidity does not fall from the sky: it is manufactured on the balance sheet of intermediaries, against collateral. A repurchase agreement, or repo, is a sale coupled with a repurchase. A borrower hands over a security, most often a US Treasury bond, and receives cash, with a commitment to buy the security back the next day at a slightly higher price. The difference is an interest rate. The loan is therefore secured by the security, the collateral, which makes repo far safer than an unsecured loan. The benchmark rate for the secured segment is SOFR, published each day by the New York Federal Reserve. The essential point holds in one idea: in repo, cash and collateral are two faces of the same coin, and liquidity is born from their circulation. How repo manufactures liquidity The market connects three families of actors. On one side, cash lenders looking for a short, safe placement: mostly money market funds, but also banks and companies. On the other, cash borrowers who hold securities: hedge funds, asset managers, and indirectly the Treasury, whose debt is partly carried by leveraged investors. Between the two, the dealers, primary investment banks, who intermediate the flow by borrowing on one side to lend on the other. Liquidity is made at this intermediary floor. The same security can serve as collateral several times, and the dealer transforms maturities and counterparties on its own balance sheet. The system's capacity to produce funding therefore depends directly on the space available on dealers' balance sheets, a resource constrained by post-2008 regulation. This is the first fault line: when this capacity shrinks, funding grows scarce even if cash exists elsewhere. The Fed's floor and ceiling The Fed does not act on repo by setting a price, but by bracketing a corridor. The interest on reserve balances, IORB, steers the cost of cash for banks. At the bottom, the reverse repo facility, RRP, offers a floor placement: its use passed $2 trillion in 2023, before falling back near zero. At the top, the standing repo facility, SRF, lets eligible counterparties borrow cash against Treasuries, agency debt and agency MBS: it acts as a ceiling on short rates. All of it hinges on the Treasury's account at the Fed, the TGA, and on the overall level of bank reserves. This machinery, and the net-liquidity proxy that follows from it, is detailed in our guide on liquidity. Where it breaks The breaking points are known, and they owe less to a lack of cash than to its distribution and to balance-sheet constraints. At quarter-ends and reporting dates, dealers cut their intermediation to lighten their balance sheet, and secured rates can climb above the corridor. A sudden rise in the TGA, when the Treasury rebuilds its cash, drains reserves out of the system. A heavy Treasury issuance swells the collateral to be financed. And when reserves become scarce, the slightest of these tremors transmits to rates. The reference lesson remains September 2019. Reserves had fallen to about $1.4 trillion, and a collision between tax payments and auction settlements sent repo rates jumping well above target, up to double-digit transactions, forcing the Fed to inject up to $100 billion a day. The regulators' conclusion: in a world of regulatory liquidity floors, so-called abundant reserves can prove illusory if they are poorly distributed. 2025-2026: the tremors are back The setting changed in late 2025. The Fed closed its quantitative tightening on 1 December 2025, bringing its balance sheet from $8.9 trillion in 2022 to about $6.5 trillion, nearly $2.4 trillion of liquidity withdrawn. Bank reserves came back to around $3 trillion. And stress signals reappeared even before the end of QT: in mid-September 2025, the standing facility was tapped for about $18.5 billion in a single day, the largest draw since its creation, with SOFR around 4.42% and secured rates durably above the fed funds rate. The Fed reacted. On 10 December 2025, it removed the facility's aggregate cap, which was $500 billion a day, switched it to full allotment, renamed it to fight the stigma, and restarted reserve purchases of about $40 billion of Treasury bills a month until mid-April 2026. The test came fast: on 31 December 2025, at the year turn, SOFR jumped to 3.87%, some transactions reaching 4.0%, and banks borrowed $75 billion from the facility before unwinding it all on 2 January. The ceiling worked, but it had to be triggered. Collateral, the hidden variable There remains the piece we look at least: collateral itself. The whole system rests on the quality and availability of the Treasuries that back it. Yet the supply of Treasuries swells with deficits, and part of this debt is carried by leveraged investors who finance their purchases in repo, notably through the cash-futures basis trade, whose size several market observers estimate at more than $1 trillion. This funding demand is near-structural, which makes repo dependent on dealers' ability to intermediate without limit. The Fed is, moreover, preparing central clearing of its standing operations to free up this balance-sheet capacity. The Fed staff note on the "balance-sheet trilemma", published on 14 January 2026, formalises the tension: the lower reserves fall relative to the stock of Treasuries, the greater the sensitivity of secured rates to liquidity shocks, and the higher volatility climbs absent intervention. The March 2026 survey of bank chief financial officers, cited by the New York Fed, shows a very steep reserve-demand curve: even a modest fall in reserves could trigger a marked rise in short rates. In other words, the cushion is thinner than it looks. The lasting lesson is here. Liquidity is not a stock parked somewhere, it is a flow manufactured continuously on the balance sheet of constrained intermediaries, against a collateral whose supply keeps growing. As long as reserves are clearly abundant, the mechanism runs quietly. When they approach the ample threshold, balance-sheet dates, TGA spikes and auctions become so many tipping points, and the central bank has no choice but to keep its ceiling armed. --- Primary sources: New York Federal Reserve, Roberto Perli speeches "Money Market Conditions and the Federal Reserve's Balance Sheet" (12 November 2025), "Reflections on the Early Days of Reserve Management Purchases" (26 March 2026) and remarks to the Atlanta Fed (19 May 2026); FOMC, implementation note and minutes of 9-10 December 2025, Open Market Trading Desk statements on the standing facility (10 December 2025); Board of Governors, FEDS Note "The Central Bank Balance-Sheet Trilemma" (14 January 2026) and H.4.1 release; Congressional Research Service, "The Federal Reserve's Balance Sheet"; Treasury, Quarterly Refunding Statement (4 February 2026); Wolf Street for the year-end draws. Figures and dates verified one by one. For the conceptual framework, reference work by the New York Fed (Liberty Street Economics), the BIS, the OFR and the Zoltan Pozsar archive on repo and collateral. ============================================================================ ANALYSIS: Collateral and rehypothecation: one security, several owners, and the keystone of the whole plumbing URL: https://l0g.fr/en/analysis/collateral-and-rehypothecation/ Canonical French source: https://l0g.fr/posts/collateral-rehypothecation-cle-de-voute/ Date: 2026-07-13 (reviewed 2026-07-13) Topics: macro, markets, liquidity, central banks ---------------------------------------------------------------------------- At the close of this series on the plumbing of the dollar, one mechanism remains to be exposed, the one hiding beneath all the others. Repo, the basis trade, shadow banking, eurodollars and the cross-currency basis all rest on one idea: a security pledged as collateral does not sleep, it immediately goes off to back another loan. This reuse, rehypothecation, turns a finite stock of Treasury bonds into a far larger volume of funding. Collateral is the oil of the financial engine, and its circulation is the keystone of the whole edifice. Rehypothecation is the right, for whoever receives a security as collateral, to pledge it themselves as collateral for another obligation. A hedge fund hands a Treasury bond to its broker to borrow, the broker reuses that same bond to fund itself with a money market fund, which can in turn mobilise it elsewhere. The security has not changed economic owner, but it now backs several superimposed claims. This is how collateral chains form, through the securities-financing transactions that are repo and securities lending. Collateral velocity The reference work is Manmohan Singh's, at the IMF, who proposed measuring this phenomenon through a velocity, on the model of the velocity of money. Collateral velocity is the ratio of the total volume of collateral received by the major intermediaries to the original collateral supplied by primary holders, hedge funds, pension funds, insurers, official accounts. At the end of 2007, ten to fifteen banks at the core of the global plumbing received close to $10 trillion of collateral, for a velocity on the order of 3. In other words, each unit of source collateral backed on average three obligations. The Fed formalised a close measure, the collateral multiplier, analogous to the money multiplier. When the lubrication seizes This circulation is not neutral for overall liquidity. After Lehman's collapse, two things happened at once: available source collateral fell, and velocity dropped. The combined effect, per the IMF, amounts to a contraction of collateral in circulation on the order of $4 to $5 trillion. Post-crisis regulations, by requiring the big banks to cut leverage and strengthen capital, shrank the balance-sheet space dealers devoted to circulating collateral. Less balance-sheet space, less reuse, less lubrication. The move is not one-way. As central banks shrank their balance sheets, freeing space at dealers, collateral reuse turned back up. It is the same parameter, intermediaries' balance-sheet space, that governs repo, the basis trade and the cross-currency basis. Collateral and the bank balance sheet are the two scarce resources around which the whole system turns. The hidden risk: chains that freeze Rehypothecation creates liquidity, but it also creates a particular fragility. Along a chain, the same security appears as an asset and as collateral at several points at once. If one link defaults, everyone downstream discovers that their collateral is immobilised or contested. That is what happened in 2008, when broker clients saw their rehypothecated collateral frozen in the bankruptcy, and again in 2011 in the collapse of a broker that had reused client assets. Long chains maximise liquidity in calm times, and destroy it at a stroke in a crisis. Add a measurement problem, central for this journal. Because the same security is counted in several places, the system's real leverage is higher than it looks, and hard to reconstruct. Financial-stability statistics often include neither pledged collateral nor its reuse, so the supervisor sees a stock, not the cascade of claims it backs. The keystone Everything meets here. Repo is the operation by which collateral circulates. The basis trade stacks several floors of it with leverage. Shadow banking makes it its fuel outside the banks. Eurodollars extend the mechanism to the offshore layer of the dollar. And the cross-currency basis displays its price when access tightens. Beneath each of these markets is the same elementary gesture: a security pledged as collateral, then reused, again and again. The stability of the whole therefore depends on a variable that few dashboards really track, the speed at which a finite stock of Treasury bonds turns into a far larger volume of promises. Making this circulation visible is making measurable the very opacity of the system. --- Primary sources: International Monetary Fund, Manmohan Singh, "Velocity of Pledged Collateral: Analysis and Implications" (Working Paper 11/256, 2011), "Collateral Reuse and Balance Sheet Space" (WP 17/113, 2017), "Collateral and Financial Plumbing" (Risk Books) and, with Goel, "The Pledged Collateral Market's Role in Transmission to Short-Term Market Rates" (2019); Singh and Aitken, "The Sizable Role of Rehypothecation in the Shadow Banking System" (2010); Federal Reserve, FEDS Notes, "The Ins and Outs of Collateral Re-use" (2018, Infante, Press, Strauss); Pozsar and Singh, "The Nonbank-Bank Nexus and the Shadow Banking System". Figures and markers verified one by one. ============================================================================ ANALYSIS: Shadow banking: non-bank intermediation has overtaken the banks URL: https://l0g.fr/en/analysis/shadow-banking-nonbank-intermediation/ Canonical French source: https://l0g.fr/posts/shadow-banking-intermediation-non-bancaire/ Date: 2026-07-13 (reviewed 2026-07-13) Topics: macro, private credit, regulation, markets ---------------------------------------------------------------------------- A crossover passed almost unnoticed. For the first time since the pandemic, non-bank financial institutions hold more than half of global financial assets. Hedge funds, money market funds, insurers, asset managers, private credit, securitisation vehicles: the old shadow banking now weighs $256.8 trillion and is growing twice as fast as the banks. Credit has migrated off the bank balance sheet, to where oversight is looser and data scarcer. Extending our pieces on repo, the basis trade and private credit, here is the overall map. The term shadow banking, coined in 2007, originally described an opaque system that did credit like a bank, without its status or its safeguards. Regulators today prefer a less loaded and more precise expression, non-bank financial intermediation, or NBFI. The shift is not cosmetic: it marks the move from a hunt for suspect entities to a surveillance of activities, those that transform liquid savings into long-term credit, with leverage, outside the banking perimeter. The measure: broad and narrow The FSB keeps the reference count, across 29 jurisdictions covering more than 90% of world GDP. Its broad measure aggregates all financial institutions that are neither central banks, nor banks, nor public actors. In 2024, this perimeter grew by 9.4%, against 4.7% for the banking sector, reaching 51% of global financial assets. The most dynamic category, other financial intermediaries, which groups money market funds, hedge funds, investment funds and securitisation vehicles, jumped 11% to $169.4 trillion. Alongside, the FSB tracks a narrow measure, more relevant for financial stability: the subset of activities that truly mimic bank credit and carry risks of the same nature, fragility to investor withdrawals or use of leverage. This narrow measure rose 12% in 2024, to $76.3 trillion, or 15.4% of global financial assets. It is this core that best matches the original idea of shadow banking. Where the vulnerabilities are Three fragilities recur in every analysis. Liquidity and maturity transformation first: an open-ended fund promises daily withdrawal while holding illiquid assets, which exposes it to forced sales in case of mass redemptions. Vehicles liable to suffer panic withdrawals weighed on their own $58 trillion in 2024, up 15%. Leverage next, concentrated in hedge funds, finance companies and securitisation vehicles, which amplifies shocks, as repo and the basis trade show. Interconnection finally, because non-banks and banks are tied by a thousand threads. It is this last point, the bank and non-bank nexus, that the FSB has documented particularly this year. The links take three forms: non-banks place deposits with banks, banks grant them credit, repo and other exposures, and funds, insurers and pension funds hold securities issued by banks. In calm times, these links widen access to funding. In a crisis, they become channels of contagion. The blind spot: private credit The FSB report flags a major limit: the scarcity of data on private credit, in official statistics as in regulatory filings. This opacity is precisely the subject of our guide on private credit and our analysis of its silent contagion. When credit leaves the bank balance sheet for closed-end funds, finance companies and structured vehicles, it escapes detailed prudential reporting. The supervisor then sees an aggregate, not the detail of exposures or of nested leverage. The growth of hedge funds, up 19% in 2024 and concentrated mostly in the Cayman Islands, illustrates this migration toward less legible jurisdictions. Why it matters The non-bank system is not an evil in itself. It has widened access to credit, diversified funding sources and supported market liquidity. But it has shifted risk toward actors less capitalised, less watched, and tied to the banks by channels that tighten in a crisis. March 2020 and the unwinding of the basis trade, the 2022 crisis of leveraged UK pension funds, the episodes of stress on money market funds: every recent tremor came from this zone. With a narrow measure at $76.3 trillion and a growing dependence of credit on non-bank actors, the question is no longer whether this system is systemic, but whether it is sufficiently mapped to be supervised. On private credit, the FSB answers clearly that it is not. --- Primary sources: Financial Stability Board, Global Monitoring Report on Non-Bank Financial Intermediation 2025, published in December 2025 on 2024 data (broad measure at $256.8 trillion and 51 percent of global assets, other financial intermediaries at $169.4 trillion, narrow measure at $76.3 trillion, hedge-fund growth and the bank and non-bank nexus); European Central Bank and ESRB, EU Non-bank Financial Intermediation Risk Monitor 2025; Congressional Research Service, "Nonbank Financial Intermediation (NBFI or Shadow Banking) Policy Issues"; Bank for International Settlements, Aramonte, Schrimpf and Shin, "Non-bank financial intermediaries and financial stability". Figures and dates verified one by one. ============================================================================ ANALYSIS: The migration of credit risk: out of the banks, out of sight URL: https://l0g.fr/en/analysis/the-migration-of-credit-risk/ Canonical French source: https://l0g.fr/posts/migration-risque-credit-hors-du-regard-reglementaire/ Date: 2026-07-13 (reviewed 2026-07-13) Topics: risk, credit, banks, private credit, shadow banking, regulation ---------------------------------------------------------------------------- Here is the great paradox of post-2008 finance. Fifteen years were spent making banks safer: more capital, stress tests, greater transparency. And it worked, banks today absorb shocks that would have carried them off yesterday. Yet credit risk itself has not shrunk. It has moved. It has left the most closely watched compartment of the financial system to settle where the regulator sees less, in private-credit funds, insurers, securitisation vehicles. Non-bank finance now weighs more than half of the world's financial assets. This piece synthesises that migration, traces where the risk has lodged and how it hides there, shows that it comes back to the banks through a back door, and weighs the question that divides the regulators themselves: has the system become safer, or merely less legible? The great shift Let us start with the measure of the phenomenon, because it is spectacular. Non-bank finance, the non-bank financial intermediation (NBFI) that the Financial Stability Board tracks, has grown from about $67 trillion of assets in 2004 to some $238 trillion in 2023. It now represents more than half of global financial assets, around $257 trillion. The centre of gravity of financing the economy has tipped outside the banks. This shift is no accident, it is a direct and largely intended consequence of the response to the 2008 crisis. By imposing heavier capital requirements on banks for risky credit, regulation made that credit less profitable for them to hold. The market did what it always does against a constraint: it went around it. This is regulatory arbitrage, the shift of activity toward the least-regulated compartment. Corporate credit migrated to private credit funds, mortgage credit to non-bank lenders, all off the bank balance sheet and beyond the supervisor's watch. The risk was not removed, it was relocated. The hiding places of risk Where, exactly, has the risk lodged, and how does it make itself less visible there? The answer maps our recent coverage, because each compartment has its technique of opacity. In private credit, the value of loans is not quoted but model-estimated, which lets the same asset carry two prices depending on who holds it, the subject of our piece on private credit and its two prices. In leveraged credit and CLOs, negotiated restructurings push out default without recording it, making the default rate look lower than it is. In commercial real estate, extend-and-pretend prolongs troubled loans to avoid booking the loss. The common thread of these techniques is not fraud, it is reduced visibility. A loss not materialised, a value one estimates oneself, a maturity one pushes out: in each case the risk exists but does not show up in the public figures. Add a discreet and massive actor, the life insurer, often owned by a private-equity firm, which has loaded its balance sheet with private credit and illiquid securitisation tranches. Credit risk has not only moved; it has dissolved into structures designed, knowingly or not, to be poorly measured. The return through the back door Here is where the story of risk "leaving the banks" cracks, and it is the most important point. Banks have not left credit, they have changed roles. Rather than lending directly to the risky firm, they lend to the private-credit fund that in turn lends to the firm. They finance the non-banks through credit lines, portfolio-backed loans, warehouse facilities. The private-credit regulator says it plainly: banks remain at the core of non-bank finance, structuring and financing a large part of it. One analyst has named this move the "Great Retranching": the bank has moved up the capital structure, from direct lender to senior creditor of the non-banks, but credit risk has not left the banking system, it has been transformed. The apparent disintermediation is partly a disguised re-intermediation. That is why reading bank results now requires hunting down their exposure to non-bank actors, as we stressed in our second-quarter bank-earnings preview. The reassuring thesis: patient capital We must now give full force to the opposite reading, because it is defended by serious authorities and is nothing absurd. In it, this migration has made the system safer, not more dangerous. The central argument is patient capital. A private-credit fund finances its loans with money locked up for years, from pension funds and insurers, not with deposits repayable on demand. It therefore cannot suffer a run like a bank, nor be forced to dump its assets overnight. In case of loss, it strikes long-term investors who accepted it, not the payment system nor the taxpayer. This reading has the support of the regulators themselves. The chair of the SEC judged in 2026 that private credit "is not a systemic risk", and the head of the IMF's capital-markets division that this risk "is certainly contained". Academic work goes the same way: one study estimates that a shock adding $36 billion to banks' private-credit exposure would give up only about two basis points of their hard capital ratio, a trifle. And the most awkward argument for the alarmists remains that the 2023 panic hit regional banks, that is, the most regulated place in the system, not private credit. To regulate is not to make safe, and shifting risk toward patient capital could well be progress. Why the migration still worries The antithesis is solid, but it has flaws the analysis cannot ignore. The first is the re-coupling just described: if banks finance the non-banks, risk is not walled off, it circulates between the two worlds through the wholesale-funding channel. The IMF said it explicitly, a stress in the non-bank sector can propagate to the banks, and contagion would hit leveraged credit, regional banks, insurers and pension funds simultaneously. The second flaw is hidden leverage, stacked on several floors, that of the fund, that of its assets, that of the insurer that holds them, and that no aggregate statistic captures well. The third flaw is the deepest, and it is the heart of the l0g thesis: opacity is itself a risk. You cannot manage what you do not measure, and private credit is valued on models, not on market prices. "What we still don't know about private credit is troubling", sums up a headline in the trade press. Add that this market has never lived through a real default cycle: it grew during fifteen years of easy money, and its promise of resilience remains an untested hypothesis. Finally, the border of patient capital blurs with the rise of semi-liquid vehicles sold to the retail public, which reintroduce run risk where it was sworn there was none, as the gating episode at a semi-liquid private-credit fund showed. The Financial Stability Board was not fooled, publishing in 2026 a report dedicated to the vulnerabilities of private credit. The real stake: seeing Our reading, measured, does not choose between the two theses, because the data to do so does not yet exist. It points instead to the problem that transcends them. The risk most surely created by this migration is neither an excess of leverage nor a wave of imminent defaults, two things that will be debated for a long time. It is the loss of legibility. Regulation optimised itself for the last crisis, the banks', with its ratios and its stress tests, while risk settled where data is scarce, late and estimated. The supervisor fights the last war, weapon in hand, in an empty room, while the game is played in the next room, without light. Yet illegibility is in itself a form of risk, regardless of whether the system is objectively more fragile. A system you cannot measure is a system whose true fragility you will discover at the worst moment, when a shock forces the figures out. It is the same lesson as that of pre-2008 shadow banking: the danger was not only in the subprimes, it was in the fact that no one knew who held them. Finance has moved credit risk from a place it watched to a place it barely looks at. Whether that is safer or not stays open; that it is less visible does not. In sum Credit risk has not vanished from the financial system, it has changed address, and its new address is less well lit than the old. Regulators assure it is better placed, in patient hands that can carry it; the facts of re-coupling, stacked leverage and valuation opacity invite us not to take them at their word. Both camps are right about part of reality, and neither can prove its thesis until a real default cycle has occurred. What must be followed, then, is not a single number but a set of signals: banks' exposure to the non-banks, the truth of valuations when it filters through, the behaviour of semi-liquid vehicles under strain, and the first big default that will force everyone to look. Risk has left the light. The analyst's job is to keep following it in the shadow. Sources 1. Financial Stability Board, "Global Monitoring Report on Non-Bank Financial Intermediation 2025": NBFI from $67trn (2004) to $238trn (2023), more than half of global financial assets: https://www.fsb.org/uploads/P161225.pdf 2. Finance Watch, "Shadow banking no more? Banks are at the core of Non-bank financial intermediation": the central role of banks in funding the non-bank sector: https://www.finance-watch.org/press/shadow-banking-no-more-banks-are-at-the-core-of-non-bank-financial-intermediation-nbfi/ 3. Financial Stability Board, "Report on Vulnerabilities in Private Credit", 6 May 2026: https://www.fsb.org/uploads/P060526.pdf 4. American Investment Council, "Regulators Affirm Private Credit Does Not Pose Systemic Risk": statements by Paul Atkins (SEC) and Tobias Adrian (IMF): https://www.investmentcouncil.org/what-they-are-saying-regulators-affirm-private-credit-does-not-pose-systemic-risk/ 5. Perspective on Risk, "Private Credit" (April 2026): the "Great Retranching" and the transformation of risk into senior exposure to the non-banks: https://perspectiveonrisk.substack.com/p/perspective-on-risk-april-11-2026 6. Wealth Management, "What we still don't know about private credit is troubling": the opacity and the first test of private credit: https://www.wealthmanagement.com/alternative-investments/what-we-still-don-t-know-about-private-credit-is-troubling 7. l0g, Private credit, one asset two prices, The silent contagion of private credit and the guide CLOs and leveraged loans. ============================================================================ ANALYSIS: AI and productivity: between measured gains and assumed effects URL: https://l0g.fr/en/analysis/ai-and-productivity/ Canonical French source: https://l0g.fr/posts/ia-et-productivite-entre-gains-mesures-et-effets-supposes/ Date: 2026-07-13 (reviewed 2026-07-13) Topics: ai, productivity, labour, macro, risk ---------------------------------------------------------------------------- import InfographicBars from '../../components/infographic/InfographicBars.astro'; Few subjects concentrate as many promises and as little hindsight as the productivity of generative artificial intelligence. On one side, a narrative of imminent rupture, quantified in points of GDP. On the other, a body of empirical studies that, read without filter, tells a more nuanced story: real gains on specific tasks, a very uneven frontier of competence, and a macroeconomic signal that is for now nowhere to be found. Sorting it out means distinguishing three scales, the task, the firm and the economy, which do not say the same thing. In the lab, real gains Let us start with what research establishes most solidly: on isolated, measurable tasks, generative-AI assistance improves productivity, often markedly. Three controlled experiments stand as references. On software development, an experiment with GitHub Copilot saw the assisted group implement an HTTP server in JavaScript 55.8% faster than the control group. On customer support, the study by Brynjolfsson, Li and Raymond, covering 5,172 agents at a Fortune 500 firm, measures a 14% rise in the number of cases resolved per hour, with a far stronger effect, on the order of 34%, among the least experienced agents. On professional writing, the experiment by Noy and Zhang, published in Science, exposed 453 professionals to ChatGPT: assisted writers worked 0.8 standard deviations faster and produced texts rated 0.4 standard deviations better by blind evaluators. One result recurs in this work and deserves emphasis: it is the lowest-performing workers who gain the most. AI compresses the distribution of performance rather than widening it, bringing beginners closer to experts on standardised tasks. An uneven frontier, and a mirage of speed These gains, however, have a deceptive geometry. The study by Dell'Acqua, Mollick, Lakhani and co-authors, run with 758 Boston Consulting Group consultants, introduced an image that has become central: the "jagged technological frontier". Inside AI's zone of competence, across 18 realistic tasks, assisted consultants did 12.2% more tasks, 25.1% faster, with higher quality. But on a complex task chosen outside that zone, assisted consultants were 19% less likely to produce a correct answer. The same tool helps or harms depending on whether the task falls on the right side of a frontier the user cannot see. To this unevenness is added a trap of perception. In 2025, the evaluation body METR ran a controlled experiment on 16 experienced open-source developers, handling 246 real tasks on large code repositories. A counter-intuitive result: with AI tools, they took 19% longer. More striking still, afterwards these same developers estimated they had been 20% faster. The gap between felt speed and real speed is a warning sign for any firm steering its gains by guesswork. The result's reach stays narrow, seasoned developers on mature codebases, but it reminds us that lab gains do not transfer mechanically to every context. The productivity paradox, 2026 edition If task gains are real, they should eventually show up in the aggregate figures. That is where the shoe pinches. The reference estimate, Daron Acemoglu's in "The Simple Macroeconomics of AI", caps AI's effect at a rise in total factor productivity of at most 0.53 to 0.66% over ten years, on the order of 0.05 to 0.07 point a year. A modest order of magnitude, far from the promised growth leap. Acemoglu adds that this figure could even be overstated, because the early evidence comes from easy-to-learn tasks, while hard tasks, context-rich and with no objective measure of success, resist more. This gap between visible micro gains and an invisible macro effect has a name: the Solow paradox, from the economist's 1987 line, "you can see the computer age everywhere but in the productivity statistics". The history of electrification and computing offers two opposing readings of this lag, which we return to in conclusion. For now, keep the raw fact: in 2026, no aggregate productivity boom attributable to AI is visible in the data. The corporate chasm Between the task and the economy there is the organisation, and that is where the promise most often gets lost. The MIT report, "The GenAI Divide: State of AI in Business 2025", built from more than 300 initiatives, 52 interviews and 153 executive responses, reaches a severe finding: about 95% of generative-AI pilot projects produce no measurable impact on the income statement, and only about 5% actually accelerate revenue. The report also documents a wide usage gap. While only 40% of firms have an official subscription to a large language model, 90% of surveyed employees say they use personal tools like ChatGPT or Claude for their work every day. This "shadow AI" signals massive but disorganised adoption, where individual gains do not rise to the firm level for lack of integration into processes. The lesson matches that of the great technological waves: value comes not from the tool, but from the reorganisation around it. The report notes, moreover, that buying specialised solutions succeeds about twice as often as in-house development, and that the best return is found in automating support functions, not in the marketing uses where budgets nonetheless concentrate. This gap between capital invested and value created echoes the risk we described in AI circular financing and the financial fragility flagged by the BIS. Employment effects: early signals, cautious causality There remains the question that worries most, employment, and it is the one where we must be most rigorous about the distinction between correlation and causation. Work by the Stanford Digital Economy Lab, run by Brynjolfsson, Chandar and Chen on payroll data, uncovers a clear signal: since late 2022, employment of workers aged 22 to 25 in the occupations most exposed to AI, such as software development and customer support, has fallen by about 16%. Over the same period, employment of workers aged 30 and over in these same occupations has risen by 6 to 12%. The proposed explanation is plausible and instructive: AI mainly substitutes for codified knowledge, that of manuals and curricula, which makes up the bulk of a beginner's value added, while it struggles to replace the tacit knowledge accumulated through experience. The young graduate therefore finds themselves in more direct competition with the machine than the seasoned professional. Caution remains in order, however: isolating AI's own effect from a broader sector slowdown is hard, and we observe in parallel a wage premium, with pay rising for those who do enter AI occupations. The signal is real and persistent, but it documents a recomposition, not yet a massive net destruction. The other reading: historical lag or oversell? How to reconcile undeniable task gains with a listless macro? Two readings clash, and honesty commands laying out both. The first is optimistic and historical. It recalls that between the invention of a general-purpose technology and its imprint on aggregate productivity, decades pass. Electricity took nearly forty years to transform industrial productivity, the time it took to redesign factories around the electric motor. Computing had its own Solow paradox in the 1980s before the gains of the 1990s. In this reading, AI follows the same J-curve: the gains are ahead of us, as organisations reinvent themselves. The second is soberer. It stresses that the easiest gains, on standardised tasks, are perhaps already largely captured, and that the remaining tasks are precisely those where AI runs into its jagged frontier. In this reading, Acemoglu's cautious estimate is not a floor awaiting upward revision, but a realistic order of magnitude, and the gap with the dominant narrative mostly measures an oversell. It is not possible to settle this today, and pretending otherwise would be dishonest. What the evidence permits saying is more modest, but solid. AI's productivity gains are real at the task level, heterogeneous along the frontier of competence, largely uncaptured at the firm level, and invisible at the macro level. Employment effects are starting to show, but as a recomposition between generations more than a bloodletting. The risk, for the analyst as for the decision-maker, is not so much that AI fails as that a promise is mistaken for a proof, and that investments or policies are sized on the former while awaiting the latter. Sources 1. Peng et al., "The Impact of AI on Developer Productivity: Evidence from GitHub Copilot": HTTP-server task done 55.8% faster by the assisted group: https://arxiv.org/pdf/2302.06590 2. Brynjolfsson, Li and Raymond, "Generative AI at Work": 5,172 customer-support agents, +14% cases resolved per hour on average, +34% for the least experienced: https://www.nber.org/papers/w31161 3. Noy and Zhang, "Experimental evidence on the productivity effects of generative artificial intelligence", Science: 453 professionals, time cut by 0.8 standard deviations, quality 0.4 standard deviations higher, largest gains for the lowest performers: https://www.science.org/doi/10.1126/science.adh2586 4. Dell'Acqua, McFowland, Mollick, Lakhani et al., "Navigating the Jagged Technological Frontier", Organization Science 2025: 758 BCG consultants, +12.2% tasks and 25.1% faster inside the frontier, 19% fewer correct answers outside it: https://papers.ssrn.com/sol3/papers.cfm?abstractid=4573321 5. METR, "Measuring the Impact of Early-2025 AI on Experienced Open-Source Developer Productivity": 16 experienced developers, 246 tasks, 19% more time with AI while estimating they were 20% faster: https://metr.org/blog/2025-07-10-early-2025-ai-experienced-os-dev-study/ 6. Acemoglu, "The Simple Macroeconomics of AI", NBER Working Paper 32487: effect on total factor productivity of at most 0.53 to 0.66% over ten years, potentially overstated: https://www.nber.org/papers/w32487 7. MIT NANDA, "The GenAI Divide: State of AI in Business 2025": about 95% of pilots with no measurable impact on the bottom line, 90% of employees in informal use against 40% official subscriptions, buying more effective than in-house development: https://fortune.com/2025/08/18/mit-report-95-percent-generative-ai-pilots-at-companies-failing-cfo/ 8. Brynjolfsson, Chandar and Chen (Stanford Digital Economy Lab), employment effects on the young: fall of about 16% for the 22-25s in exposed occupations since late 2022, rise of 6 to 12% for the 30-and-overs: https://digitaleconomy.stanford.edu/news/ai-and-labor-markets-what-we-know-and-dont-know/ 9. Fortune, follow-up on the Stanford study of entry-level employment, persistent and non-reversible effect: https://fortune.com/2026/06/27/what-is-ai-impact-entry-level-jobs-stanford-adp-canaries-brynjolfsson-richardson/ ============================================================================ ANALYSIS: The bubble within the bubble: valuations, AI and inflated earnings URL: https://l0g.fr/en/analysis/the-bubble-within-the-bubble/ Canonical French source: https://l0g.fr/posts/la-bulle-dans-la-bulle-valorisations-ia-et-benefices-gonfles/ Date: 2026-07-13 (reviewed 2026-07-13) Topics: markets, ai, valuations, risk, macro ---------------------------------------------------------------------------- import InfographicBars from '../../components/infographic/InfographicBars.astro'; The debate over a possible stock-market bubble, fuelled by artificial intelligence, plays out on two registers that answer each other poorly. The first is qualitative: judging whether an industry's real potential justifies its prices. The second is quantitative: comparing valuations with their own history. A market note relayed by the Financial Times in early July raised eyebrows precisely because it pushes the second register to its logical conclusion, with a deliberately provocative number. Let us take both registers, methodically, before drawing a reading. Two ways to sense a bubble This dual reading grid, framed by the entrepreneur and digital thinker Gilles Babinet in commenting on the episode, sets two ways of judging a bubble against each other. The first method assumes fine knowledge of the sector: estimating that its speed of development justifies very high prices. Applied to AI, this reading rests on an imaginary of rupture, summed up in a series of acronyms that have become totems in the industry: AGI for artificial general intelligence, ASI for superintelligence, RSI for an AI able to improve itself. These horizons are championed by the field's leading players. In his essay "Machines of Loving Grace", Anthropic's CEO Dario Amodei places the arrival of a "powerful AI", with capabilities matching or exceeding the best human specialists across most disciplines, as early as late 2026 or early 2027. Elon Musk announces his Optimus humanoid robot going on sale for late 2027, between $20,000 and $30,000, promised for domestic uses. The conviction that these milestones will be met, and on those dates, is the bedrock of the bullish scenario. The second method dispenses with forecasting the future: it takes the big valuation ratios. The P/E, the ratio of a stock's price to its earnings, and above all the CAPE, popularised by the economist Robert Shiller, which relates price to the ten-year average of inflation-adjusted earnings, so as to smooth the cycle. It is on this terrain that the number circulating this week sits. The number that gave the FT pause In FT Alphaville, on 3 July 2026, Bryce Elder relays a monthly note from strategists Joachim Klement and Francisca Reis, of Panmure Liberum. Their reasoning comes in two steps. First, the classic observation: per Shiller's data, the S&P 500 CAPE was 32.6 in 1929, 1.8 standard deviations above its trend, and 44.2 in 2000, 3.3 standard deviations, a clear bubble signal. Today it is at 41.0, 2.9 standard deviations above trend. Bubble territory, then, but nothing unprecedented at this stage. The second step is the more original. In 1929 as in 2000, earnings were within their normal range, less than one standard deviation from their trend. Today they are themselves 1.8 standard deviations above. In other words, the valuation is high at a moment when the denominator, earnings, is already abnormally inflated. Corrected for this anomaly, the CAPE would come out not at 41 but at 67.6, 4.6 standard deviations above trend, a level that exceeds anything US history has known. Klement draws a striking image from it: under the assumption, false and acknowledged as such, of a normal distribution of valuations, such a level would occur in 0.00019% of months, or once every 43,432 years. A point of method is in order, and the FT author stresses it first: this "once every 43,432 years" must not be taken literally, because valuations do not follow a normal law. It is a way of expressing the size of the gap, not a real probability. Elder adds the usual caveat: supra-normal profits always end up normalising, but trying to time the turn is a mug's game. Rigour commands keeping the order of magnitude without lending it false precision. The bubble within the bubble: earnings themselves The heart of the thesis is therefore not the CAPE, but the denominator. The idea that current earnings are abnormally high holds up beyond the Panmure note alone. US corporate margins run around 14%, a record high in the available data, supported by factors that are not all durable: sector concentration, favourable taxation, massive share buybacks that flatter earnings per share, and the direct effect of AI spending on a few champions. Yet the profit margin is, in the phrase of investor Jeremy Grantham, "probably the most mean-reverting series in finance, and if profit margins do not mean-revert, then something has gone badly wrong with capitalism". The asset manager GMO devoted a study with a telling title to this anomaly, "The Curious Incident of the Elevated Profit Margins". If this logic holds, using inflated earnings in the denominator of a valuation ratio makes the market look cheaper than it is. That is exactly the mechanism of the "bubble within the bubble": an overvaluation laid on profits that are themselves in excess. The rising doubt over AI profitability What makes the question burning in 2026 is that the bad news is piling up on the real profitability of AI, both for those who sell it and for those who buy it. Infrastructure spending is exploding far faster than revenue. The five largest hyperscalers plan between $700 and $900 billion of investment in 2026, up about 36% year on year, while the AI ecosystem would generate far lower revenue, leaving a shortfall estimated at around $600 billion a year. According to Allianz Research, the divergence between AI investment and revenue growth reaches about 46%, beyond the 32% seen during the 2001 telecom excess, which preceded a brutal and durable correction. OpenAI's case illustrates the tension: about $25 billion of annualised revenue in early 2026, but past losses of $540 million in 2022, $1.5 billion in 2023 then $5 billion in 2024, with no profitability expected before 2029. On top of that sits a partly circular financing architecture, where a fraction of declared revenue is capital recycled among interconnected players rather than independent organic demand, a risk we described in AI circular financing and the fragility flagged by the BIS. Finally, the value actually created stays uncertain: as our review of the evidence on AI productivity shows, the gains are real at the task level but struggle to diffuse to the economy, which limits the pool of revenue capable of justifying the investment. The antithesis, taken seriously Rigour forbids stopping at the bearish thesis, however well supported. Several solid objections deserve to be raised. First, the CAPE has its limits, widely documented. It is criticised for ignoring changes in accounting standards, the rise of buybacks, the forty-year secular decline in interest rates, and the shift in the market's sector composition, lighter in capital than it once was. Some argue that part of the rise in margins is structural, driven by dominant, capital-light firms, and not a mere cyclical excess bound to correct. If this reading is right, the denominator is not so inflated, and the "bubble within the bubble" deflates on its own. Next, even accepting the diagnosis, the timing stays unpredictable. An extreme valuation says an asset is expensive, not that it will fall tomorrow. Bubbles can inflate for years, and betting on their bursting has ruined plenty of sceptics: over the past decade, recurrent warnings of a "tech bubble" caused many to miss a considerable rise in the indices. A high CAPE is an indicator of mediocre ten-year forward returns, not a timing signal. Finally, the bullish scenario has its internal coherence. If the industry's totems materialise, even in part, future earnings could grow fast enough to catch up with prices, and today's valuation would then be nothing aberrant. A truly transformative AI would justify high and durable margins. That is the bet, perfectly defensible in theory, of those who buy the promise rather than the proof. Reality's verdict The weak point of the bullish scenario is not its logic, it is its dependence on dated predictions. And dated predictions have a merciless judge: the calendar. Financial history is a graveyard of expert convictions belied at the worst moment. In October 1929, days before the crash, the economist Irving Fisher, a luminary of his discipline, assured that prices had reached "what looks like a permanently high plateau", an episode John Kenneth Galbraith recounts in "The Great Crash 1929". The lesson is not that optimists are always wrong, but that a player's closeness to an industry does not protect it from excess enthusiasm, it exposes it to it. What to conclude, without yielding to either catastrophism or denial? Three things, that the evidence permits. US valuations are, on historical measures, at rarely reached peaks, and the margin of safety is thin. These valuations rest on earnings whose durability is contested, which doubly weakens the structure. And the scenario that would justify them depends on technological promises still unkept, whose deadline is approaching. None of this says when, or even whether, a correction will come: the timing stays out of reach, as the classics of the genre recall, from Robert Shiller's "Irrational Exuberance" to Charles Kindleberger's "Manias, Panics, and Crashes". But the honest analyst must distinguish a promise from a proof, and acknowledge that today, an important share of stock prices rests on the former. Sources 1. Bryce Elder, FT Alphaville, 3 July 2026, relaying the Panmure Liberum note (Joachim Klement, Francisca Reis): CAPE of 32.6 in 1929, 44.2 in 2000, 41.0 in 2026; earnings 1.8 standard deviations above trend; corrected CAPE at 67.6 (4.6 standard deviations); 0.00019% of months, or once every 43,432 years, under a normality assumption explicitly presented as false: https://www.ft.com/content/8e9337f8-9191-48e9-9289-a8defda89431 2. Robert Shiller, historical CAPE data and "Irrational Exuberance" (Princeton University Press, 2000): construction and reading of the cyclically adjusted valuation ratio: http://www.econ.yale.edu/~shiller/data.htm 3. Dario Amodei, "Machines of Loving Grace", personal essay: arrival of a "powerful AI" possible as early as late 2026 or early 2027: https://www.darioamodei.com/essay/machines-of-loving-grace 4. Entrepreneur / statements by Elon Musk: Optimus going on sale targeted for late 2027, between $20,000 and $30,000, domestic uses: https://www.entrepreneur.com/business-news/elon-musk-tesla-sell-optimus-humanoid-robots 5. Jeremy Grantham (GMO), quote on the mean reversion of margins, and GMO, "The Curious Incident of the Elevated Profit Margins": https://www.gmo.com/americas/research-library/the-curious-incident-of-the-elevated-profit-margins-part-1whitepaper/ 6. Fortune, 19 May 2026, Jeremy Grantham on the "AI war" and coming pressure on margins: https://fortune.com/2026/05/19/blood-in-the-streets-jeremy-grantham-ai-monopoly-brutal-competitive-world-recession/ 7. Allianz Research, 2026: AI capex-revenue divergence of about 46%, beyond the 32% of the 2001 telecom cycle: https://www.allianz.com/content/dam/onemarketing/azcom/Allianzcom/economic-research/publications/specials/en/2026/march/20260325AI.pdf 8. Forbes, 2 June 2026, widening gap between AI spending and revenue, hyperscalers between $700 and $900 billion of capex in 2026: https://www.forbes.com/sites/jasonkirsch/2026/06/02/the-ai-capex-to-revenue-gap-is-widening---and-markets-are-starting-to-notice/ 9. FutureSearch, OpenAI financials: about $25 billion of annualised revenue in early 2026, losses of $540 million (2022), $1.5 billion (2023) and $5 billion (2024), profitability expected around 2029: https://futuresearch.ai/openai-revenue-forecast/ 10. John Kenneth Galbraith, "The Great Crash 1929" (1955): Irving Fisher's statement on a "permanently high plateau" on the eve of the 1929 crash. 11. Charles P. Kindleberger, "Manias, Panics, and Crashes: A History of Financial Crises" (1978): the recurring anatomy of bubbles and their turn. 12. Gilles Babinet, comment on X reacting to the FT article (secondary source, analysis rather than primary source): proposes the two-method reading grid taken up here, compiles the AI industry's dated predictions and recalls the precedent of experts belied in 2000: https://x.com/babgi/status/2073731501045248405 ============================================================================ ANALYSIS: The AI boom under the BIS lens: real revolution, opaque financing, possible overcapacity URL: https://l0g.fr/en/analysis/ai-boom-bis-financial-fragility/ Canonical French source: https://l0g.fr/posts/boom-ia-bis-fragilite-financiere/ Date: 2026-07-13 (reviewed 2026-07-13) Topics: macro, private credit, tech, central banks ---------------------------------------------------------------------------- The Bank for International Settlements does not deal in sensation. When it files investment in artificial intelligence among the pressure points that "warrant attention", on a par with inflation and fiscal stress, you should read the source document rather than the headlines. In a thread published on 29 June 2026, Finneko (@finnekoprgrm) relays and comments on chart 13 of the BIS annual report, whose caption says the essential: corporate credit is vulnerable to a repricing should AI disappoint. Here we go back to the primary source, and test that outlook against the state of economic research on technological revolutions. Finneko's thread draws five ideas from the BIS that deserve to be taken seriously: a productivity gain on one task does not make a durable rise for the whole economy, the investment race is partly defensive and risks overcapacity, a bottleneck may appear on the demand side, the ecosystem's financing is opaque, and the market's error would be to turn a true story into financial certainty. Let us go back to the record. What the BIS actually says The diagnosis rests on data. BIS bulletin no. 120, by Aldasoro, Doerr and Rees, sizes the phenomenon: by mid-2025, spending on data centers and semiconductor plants equalled 1% of US GDP, and total computing-related investment reached 5% of GDP, above its 2000 dot-com peak. AI contributed 0.4 point to US growth over three years, and computing as a whole nearly half of recent growth. The trajectory is not weakening: the BIS projects annual data-center spending rising to 0.8 to 1.3% of GDP, against 0.5% today. The hard point is financial. The five big hyperscalers plan more than $1,000 billion of AI-related capex across 2025 and 2026, a pace that exceeds their earnings and free cash flow and pushes them toward borrowing. Chart 13 shows three converging moves: investment increasingly financed by debt, rising credit risk, circular financing become commonplace. Since January 2025, the CDS premiums of AI issuers rated BBB or better have risen while the comparable broad quality index has eased: credit is starting to single these issuers out. Circular financing, the blind spot This is the passage Finneko foregrounds most, and he is right to press it. The BIS defines circular financing as an arrangement where hyperscalers take stakes in AI labs in exchange for purchase commitments, which sends capital back to investors as revenue. Added to this are leases on data centers built by third parties, take-or-pay capacity clauses, and off-balance-sheet commitments, where leverage does not vanish for all that. Our piece on AI circular financing details these loops. Private credit is at the heart of the shift. According to the BIS, direct lending by private-credit funds to AI-linked firms has gone from close to zero to more than $200 billion, its share rising from under 1% to nearly 8% of the stock, with a projection of $300 to $600 billion by 2030. Yet the spreads demanded on these loans, 6.2 points against 6.1 for other sectors, are almost identical. Lenders are therefore treating AI risk as average risk, while equity valuations assume off-the-charts returns. One of two things: either credit underestimates the risk, or equities overestimate future profits. This dissonance between debt and equity is the report's real signal. The backdrop is a private-credit market grown colossal: assets managed by these funds have risen from about $100 billion in 2010 to more than $2,200 billion today. The BIS March 2026 Quarterly Review speaks of "shadow leverage" for commitments economically akin to debt but largely held off balance sheet, and notes gross hyperscaler bond issuance topping $100 billion in 2025, at long maturities matched to the assets' lifespan. For the mechanics of private credit, see our guide on reading private credit. Productivity cannot be decreed Finneko's first point, the gap between a gain on one task and a gain for the whole economy, is one of the best-established results in economic history. In 1990, Paul David showed, with the example of the dynamo, why electrification only lifted productivity decades later: factories had to be reorganised, the central drive shaft abandoned for distributed motors. The technology was there long before its gains showed up in the statistics. Contemporary research confirms this lag. Brynjolfsson, Rock and Syverson formalised in 2021 a productivity J-curve: general-purpose technologies demand complementary intangible investment, in reorganisation and skills, so that measured productivity first stagnates before accelerating. And Daron Acemoglu's cautious 2024 estimate puts the total-factor-productivity gain attributable to AI at no more than 0.66% cumulatively over ten years, below the 1% threshold, far from the promises. His reasoning starts from a task-based model: as long as AI's effect passes through cost savings at the level of each task, its macroeconomic effect follows from the share of tasks actually touched and the average saving per task. Nothing rules out a deep transformation, but it will be measured in years, not quarters. Defensive over-investment and the lessons of bubbles The second point, the defensive race for capital, echoes a known dynamic. Carlota Perez described in 2002 how each technological revolution passes through an installation phase where financial capital runs hot, over-invests and inflates a bubble, before a break and then a soberer deployment. Railways, electricity and the Internet bubble all combined a genuine breakthrough and an excess of capital invested too fast. Recent work by Rahil Solanki, in 2026, likens the AI circular economy to three precedents: telecom vendor financing in 2000, the off-balance-sheet opacity of 2008, and shale-oil over-investment. The BIS itself calibrates what comes next. The end of past investment booms came with a growth slowdown of more than 1 point on average, and nothing indicates that a boom, even carried by a genuine technological advance like the Internet, leads to durably stronger growth. The sharpest setback followed the Internet bubble, modest though it was relative to GDP. The size of the shock does not depend only on the size of the boom. The demand bottleneck The third point, more forward-looking, stays open. If AI shifts income from labour to capital, whose propensity to consume is lower, an economy more productive in theory can run into insufficient demand. The BIS raises this risk without quantifying it; it echoes the literature on automation and the labour share, in Acemoglu and Restrepo, who documented how automation weighed on the wage share of US national income since the 1980s. A conditional, medium-term risk, that rests as much on the sharing of value as on the technology. What it is worth The synthesis holds in a balance. The BIS does not say AI is a bubble, and its bulletin calls the macro-financial risks moderate at this stage. But the annual report goes up a notch: a repricing, through higher rates or AI disappointment, could be as disruptive to credit as the 2008 crisis, all the more so as US households are heavily exposed to equities and non-bank institutions have become the largest holders of advanced-economy sovereign debt. The real frictions are not abstract. McKinsey puts at close to $6,700 billion the global data-center capex needed by 2030 to keep up with compute demand, and the IEA reminds us that electricity, the grid and cooling are becoming a physical constraint as serious as financing. Add semiconductor shortages and competition that can compress margins, so many obstacles between the promise of use and actual profit. Finneko's contribution is to name the cognitive error: taking a true story and treating it as financial certainty. The revolution is probably real, but that compels acceptance of neither any level of capex nor any structure. For anyone who wants to judge on the record, three dials beat a thousand narratives: the gap between capital spending and free cash flow, the share of circular financing in announced revenue, and the credit premium of AI issuers. Credit often sees before equities do. This data is public, and the BIS has just published its map. --- Primary sources: - BIS, Annual Economic Report 2026, 28 June 2026: chart 13, debt financing, rising CDS of AI issuers, circular financing, repricing compared with 2008. - BIS, press release on the 2026 annual report, 28 June 2026. - BIS, bulletin no. 120, "Financing the AI boom: from cash flows to debt" (Aldasoro, Doerr, Rees), 7 January 2026: investment figures (1% and 5% of GDP, 0.4 point of growth), shift to debt, private credit, historical perspective on booms. - BIS, Quarterly Review, box on "shadow leverage", March 2026. - IMF, Global Financial Stability Report, chapter 2 "The Rise and Risks of Private Credit", April 2024. - IEA, Energy and AI, April 2025. - McKinsey, "The cost of compute: A $7 trillion race to scale data centers", 28 April 2025. - Daron Acemoglu, "The Simple Macroeconomics of AI", NBER Working Paper 32487, 2024. - Erik Brynjolfsson, Daniel Rock, Chad Syverson, "The Productivity J-Curve", NBER Working Paper 25148, 2021. - Rahil Solanki, "The AI Circular Economy: Systemic Risk, Vendor Financing, and the Keystone Problem", SSRN, April 2026. - Paul A. David, "The Dynamo and the Computer", American Economic Review, 1990. - Carlota Perez, Technological Revolutions and Financial Capital, 2002. - Starting point: Finneko thread (@finnekoprgrm), 29 June 2026. ============================================================================ ANALYSIS: AI circular financing: when the same dollar goes round and comes back as revenue URL: https://l0g.fr/en/analysis/ai-circular-financing/ Canonical French source: https://l0g.fr/posts/financement-circulaire-ia/ Date: 2026-07-13 (reviewed 2026-07-13) Topics: markets, valuations, tech, regulation ---------------------------------------------------------------------------- A question comes back every quarter: is the demand carrying AI valuations real, or partly manufactured by the players themselves? The suspicion has a name, circular financing. A chipmaker invests in an AI lab, which commits to leasing compute from cloud providers, which buy the maker's chips. The same dollar goes round and comes back as revenue, which can make demand look organic. Without settling the bubble debate, here is the loop laid flat, quantified, and the argument of those who consider it healthy. The core of the arrangement reads in three moves. Nvidia has committed to invest up to $100 billion in OpenAI as part of a capacity-deployment partnership. OpenAI, for its part, has piled up colossal compute commitments with cloud providers: around $300 billion over five years with Oracle, $250 billion with Microsoft, $22.4 billion with CoreWeave, $38 billion with Amazon Web Services. Yet these providers fit out their data centers with Nvidia chips. The capital injected by the maker at the top of the chain therefore comes back to it as orders at the bottom. The loop, in plain terms The clearest case is CoreWeave. Nvidia owns more than 5% of its equity, and agreed in September 2025 to buy $6.3 billion of cloud services from it, committing to pay for the compute time CoreWeave failed to sell to others. With that backstop, CoreWeave can order more Nvidia chips with confidence. The same pattern recurs, to varying degrees, between the maker, the lab and the compute landlords. None of these deals is illegal or abnormal in itself, vendor financing has existed for a long time, but their accumulation among a very small number of interconnected players blurs any read on real demand. The numbers, and the gap It is the scale that stops you. Per the publicly disclosed commitments, OpenAI has amassed close to $1.4 trillion of compute commitments over the decade, spread across a handful of providers, including roughly $350 billion with Broadcom and $90 billion with AMD on top of the amounts already cited. Against that, its revenue was on the order of $13 billion in early 2026, growing fast but on no common scale with its commitments, and the company was reportedly losing around $14 billion across 2026 per press estimates. This gap between spending promises worthy of a state and still-modest revenue is the knot of the debate. The echo of the dot-com bubble Industry veterans see a whiff of déjà vu. In the late 1990s, telecom equipment makers financed their own customers, the carriers, through loans and facilities, to support the build-out of fiber networks. Some carriers even swapped capacity rights between themselves, booking them as sales, while the transactions largely cancelled out. When demand disappointed, the model broke, over-leveraged carriers went bankrupt, and much of the capacity sat unused for years. The fear is that massive cross-commitments now play the same amplifier role, on the way up as on the way down. The counter-argument Against this charge there is a serious defence, which must be laid out honestly. Vendor financing is not fraud: it helped build the railways, the telecoms and the first waves of computing, by bootstrapping real markets. Underlying AI demand is not only circular, businesses, developers and individuals pay for genuinely real uses, outside the loop. And the scale of the cross-commitments, though considerable, stays measured relative to the businesses: according to UBS, the OpenAI-Nvidia deal would represent up to 13% of Nvidia's expected 2026 revenue, around $272 billion, far from making up the bulk of its income. The risk is real, but equating it straight away with accounting fraud would be excessive. What to watch Strain, if it builds, will read first in the balance sheets of the infrastructure providers: rising debt, swelling lease commitments, widening credit-default-swap spreads. A warning light already flashed in early 2026, when the press reported that Nvidia's investment in OpenAI was stalling, triggering a bout of nervousness across three giant market caps before a denial. That is the structural vulnerability of any loop: one link only has to hesitate for confidence to wobble across the whole. The role of a data journal is not to proclaim the bubble, but to map precisely who funds whom, and how much, and to make legible the share of demand that runs in a circle. The rest is a matter of judgement, and judgement needs numbers. For the wider picture, see also the debt behind AI and the BIS warning on the boom's financial fragility. --- Primary sources: company statements and disclosures (OpenAI, Nvidia, Oracle, Microsoft, CoreWeave, Amazon Web Services, AMD, Broadcom); Reuters and Bloomberg for the amounts and timeline of the deals (Oracle roughly $300 billion over five years, Nvidia up to $100 billion of investment, CoreWeave $22.4 billion, AWS $38 billion, the $6.3 billion Nvidia-CoreWeave backstop); UBS Chief Investment Office (share of the OpenAI-Nvidia deal in Nvidia's 2026 revenue, estimated around $272 billion); press estimates for the roughly $13 billion of revenue and OpenAI's expected 2026 loss. Commitments and dates verified one by one; the amounts are announced commitments, not recorded spending. ============================================================================ ANALYSIS: The debt behind AI: off-balance-sheet SPVs, bonds, private credit URL: https://l0g.fr/en/analysis/the-debt-behind-ai/ Canonical French source: https://l0g.fr/posts/la-dette-derriere-l-ia-spv-obligations-credit-prive/ Date: 2026-07-13 (reviewed 2026-07-13) Topics: ai, debt, private credit, bonds, risk, data centers, shadow banking ---------------------------------------------------------------------------- The AI debate has so far played out on two registers: stock-market valuations and the circular revenue looping between a handful of players. A third, quieter register actually decides how sound the structure is: how the build-out gets paid for. As long as the tech giants funded their data centers out of their own cash, the risk stayed contained on their balance sheets. That is no longer the case. In 2026, hyperscaler capex swallows almost all of their operating cash flow, and the shortfall goes looking for funding in the debt markets. Morgan Stanley expects close to $570 billion of AI-related issuance this year; by late 2025 this debt was already the single largest slice of the investment-grade bond market. This piece follows the plumbing of that debt-financed boom: the off-balance-sheet Meta-Blue Owl structures, a bond market beginning to choke, private credit and insurers at the end of the chain. And the counter-argument, which is not a weak one. When capex outgrows cash flow For a decade, the big cloud players funded their data centers the way any highly profitable company funds its growth: out of its own earnings. That is what long defused the comparison with the 2000 telecom bubble, where the infrastructure was built on credit. That dam has just broken. According to Morgan Stanley, hyperscaler investment is on track in 2026 to consume close to 100% of their operating cash flow, against a ten-year average of 40%. The rest has to be borrowed. The mechanism is arithmetic before it is speculative: when capital spending exceeds what operations generate, the difference is financed by issuing debt or equity. The hyperscalers have chosen debt, and at scale. Again per Morgan Stanley, global AI-related debt issuance should reach close to $570 billion in 2026, more than double 2025. By late October 2025, the outstanding stock of this debt had already passed $1.2 trillion, becoming the largest segment of the investment-grade market and overtaking US banks as the biggest sector in the JPMorgan US Liquid index. A regime change: AI is no longer only a story of expensive stocks, it has become a story of credit. Off-balance-sheet: the lesson of the Meta-Blue Owl deal The most discreet form of this debt shows up on no tech balance sheet. On 21 October 2025, Meta announced a joint venture with funds managed by Blue Owl Capital to build its Hyperion megacampus in Louisiana, for a development cost of roughly $27 billion. The split is the heart of the design: Blue Owl's funds own 80% of the structure, Meta only 20%. Meta contributes the land and construction-in-progress, collects a one-time distribution of about $3 billion, keeps operational control through a lease, but leaves the debt outside its own accounts. That debt is enormous. According to trade press reporting, PIMCO anchored a bond tranche of roughly $26 billion, rated at the top of the quality spectrum and amortising over a very long maturity, described as the largest private-credit deal ever closed. All of it through a special-purpose vehicle that carries the leverage in Meta's place. What Meta keeps, on the other hand, is a residual-value guarantee for the first sixteen years: if the lease is not renewed, Meta commits to a capped cash payment. In other words, the economic risk has not entirely left the house; it has merely changed accounting line. As the real-estate press put it, these structures let companies fund tens of billions of infrastructure with debt that will never officially appear on the balance sheet. The appeal for the issuer is twofold: preserve the group's credit rating and deconsolidate a colossal amount of borrowing. The drawback for the observer is symmetric: it makes the sector's true leverage harder to measure, exactly the kind of opacity that runs through the wider migration of credit risk beyond the regulatory gaze. The bond market starts to choke Alongside the off-balance-sheet channel there is the public market, and it is running flat out. In the first seven months of 2026 alone, AI-related bond issuance topped $250 billion, of which $218 billion in investment-grade debt and $31.9 billion in high yield, the latter almost entirely ($27.9 billion) earmarked for data centers. The acceleration is brutal: across the whole of 2025, AI-linked high yield weighed only $14.1 billion. Amazon placed $25 billion in July, Oracle turned the bond market into its main funding lever, with fiscal-year capex above $55 billion and borrowing plans that sent its share price sliding once investors took their measure. The point of stress shows up at the bottom of the credit stack. CoreWeave, a cloud operator specialised in AI compute, issued six-year notes in June 2026 at around 9.6%, a cost that betrays how the risk is perceived, and its paper then traded below par, around 96.5. The word used by credit analysts is telling: buyside "indigestion". This is not yet a market freeze, but the first sign that investor appetite, however voracious, has a limit. That is the crux: investment-grade AI debt places without difficulty, but the most fragile segment, backed by fast-depreciating equipment, is already testing the edges of demand. Private credit and insurers at the end of the chain Following the debt all the way through leads to whoever holds it. A growing share sits not in liquid bond funds but in private credit, whose corporate lending assets are expected to exceed $2 trillion in 2026 according to Moody's. This is the compartment that absorbed the Meta deal, and the one funding a share of the riskiest build-out. The problem the agency flags is not the volume but the legibility: light covenant documentation, interest paid in kind (PIK) rather than in cash, loans secured against fund net asset value, layered leverage at the vehicle level. All structures that obscure real leverage rather than remove it. Behind private credit there is often the life insurer, the big final buyer of these long-dated assets. And that is exactly where Moody's locates one of the contagion channels in a shock. In a January 2026 analysis, the agency maps what would happen if AI-related valuations fell 40%: private-credit managers forced to renegotiate to avoid defaults, exposed insurers, a wealth effect on consumption through falling equities. On top of that sits a risk specific to data centers: the performance of securitisations backed by these assets depends on tenant demand for compute capacity. If AI adoption plateaus or shifts direction, over-investment is paid for in half-leased campuses. The full chain, from the debt-financed chip to the insurance policy, is longer and more opaque than the equity surge alone suggests. Vendor financing closes the loop One last link connects this debt-financed boom to the other big debate of the moment. Part of the demand that makes this debt sustainable is itself manufactured by the suppliers. Nvidia has committed more than $40 billion of equity stakes in 2026, including around $30 billion in OpenAI, a customer that uses those funds to buy, directly or not, Nvidia chips. This vendor financing echoes the Nortel-Lucent episode of the telecom bubble, when equipment makers lent to their own customers to prop up sales. We laid out this mechanism in the anatomy of AI circular financing and through the BIS warning on the boom's financial fragility. The point that matters here is the joint: if a fraction of demand is recycled capital rather than organic need, then the debt issued to serve that demand rests on a narrower base than it appears. The counter-argument The concern is real, but the catastrophist reading runs into several solid facts. First, most of this debt is investment grade. Of the $250 billion issued in 2026, $218 billion is investment grade, carried by companies whose cash flows are, themselves, very real and among the highest in the world. Reaching straight for the 2000 telecom comparison ignores that Meta, Microsoft, Amazon or Alphabet generate profits the over-leveraged operators of that era never came close to. Second, borrowing while it is cheap is a rational balance-sheet decision, not necessarily a headlong rush. Diversifying funding sources, issuing long-dated fixed-rate debt, deconsolidating infrastructure through a long-term partner: these are humdrum financial-engineering practices for firms of this size. The Meta-Blue Owl deal, precisely, rests on amortising, well-rated debt backed by a top-tier tenant, a long way from fragile credit. Third, the dividing line is sharp. It separates debt backed by established cash flows, which places without difficulty, from bottom-of-the-range debt backed by fast-depreciating equipment and still-hypothetical revenue. It is this second compartment, a minority by volume, that shows the first signs of strain. Systemic risk does not arise from debt financing as such, but from its concentration on a handful of issuers and from the difficulty of measuring real leverage once it migrates off balance sheet and into private credit. What to watch Three signals beat one forecast. The first is the behaviour of high yield backed by data centers and the level of credit spreads: a widening would say that investor demand, voracious today, is closing up. The second is the performance of data-center securitisations, a thermometer of real demand for compute capacity. The third is the marking of positions in private credit, the least liquid and most opaque link. For the reading tools, our guides on private credit, CLOs and leveraged loans and credit spreads give the grid. The real question is not whether AI will deliver on its technological promises, but whether the debt building it will hold long enough for it to. --- Sources - Morgan Stanley, forecast for AI-related debt issuance (June 2026), via Yahoo Finance: capex ≈ 100% of operating cash flow, ≈ $570bn of issuance in 2026, $1.2trn outstanding by late 2025. - AI-related bond issuance in 2026, Yahoo Finance: investment-grade / high-yield split, CoreWeave, buyside "indigestion". - Meta Platforms, press release on the Hyperion joint venture with Blue Owl Capital (21 October 2025). - Data Center Dynamics: the debt tranche anchored by PIMCO. - Bisnow: the off-balance-sheet structure. - Yahoo Finance / Reuters: Oracle's capex and borrowing plans. - Moody's, 2026 private credit outlook and data-center credit risk: assets, opaque structures, contagion channels. - Bloomberg: cross-commitments between Nvidia, OpenAI and cloud providers (circular financing).