Asia’s U.S. assets: Treasuries, MBS, equities and risk

At 30 June 2025, Table A5 of the U.S. Treasury’s Treasury International Capital report assigned $9.834 trillion of U.S. securities to Asian residents. The rounded components were $4.504 trillion of equities and fund shares, $3.284 trillion of long-term Treasuries, $741 billion of agency debt, $848 billion of corporate debt and $458 billion of short-term debt. Those components add to $9.835 trillion. The published total remains $9.834 trillion. The one-billion-dollar gap is rounding. The consequential finding is that the survey’s largest visible category was equities and fund shares, ahead of federal debt.

This is the fifth part of The Asian dollar loop. Taiwan exposed the mismatch between foreign assets and local liabilities. Japan followed the rate shock through life-insurer balance sheets. South Korea examined the cost of near-complete currency protection. The previous instalment opened the machinery of swaps, forwards and collateral. The next question is more basic: what sits inside the label “dollar assets”?

Table A5 separates long-term securities from short-term debt. Exhibit 4 reallocates short-term securities by issuer and shows $3.724tn of Treasury debt, $742bn of agency debt, $865bn of corporate debt and $4.504tn of equities. The two views reconcile within rounding.

The Treasury recycling story is now incomplete

For decades, the standard account was straightforward. Asian trade surpluses and savings were recycled into Treasuries. That remains central to the history of Chinese and Japanese official reserves. Applied to the 2025 regional portfolio, however, it blends central banks with insurers, pension funds, banks, investment vehicles and households.

In the final SHL survey released on 30 April 2026, equities, fund shares and other capital instruments accounted for 45.8% of the summed rounded Asian categories. Long-term Treasuries were 33.4%. Exhibit 4 adds short-term Treasury instruments to the federal-debt line, taking it to roughly $3.724 trillion, still below the $4.504 trillion equity total.

That does not mean Asian reserve managers switched wholesale into the Nasdaq. The regional total combines official and private holders. It also combines economies dominated by public institutions, insurance systems with large foreign-bond books and financial centres that host funds and custodians.

The country profiles make the differences visible.

The bars use Table A5 components. Singapore’s rounded categories add to $1.119tn while its reported total is $1.118tn. The chart preserves the components to show composition and displays the official total on the right.

In Table A5 of the report, Japan’s reported $2.883 trillion spans every major class: $1.169 trillion of equities, $1.024 trillion of long-term Treasuries, $250 billion of agency debt, $315 billion of corporate debt and $125 billion of short-term paper. Mainland China remains much more Treasury-heavy, with $657 billion of long-term Treasuries against $344 billion of equities.

South Korea is almost the mirror image. Equities account for $592 billion of its $804 billion total, or 73.6%. Singapore reports $759 billion of equities against a published total of $1.118 trillion. Taiwan’s profile is bond-dense: $304 billion of long-term Treasuries, $185 billion of agency debt and $188 billion of corporate debt, compared with $172 billion of equities. Hong Kong combines $263 billion of equities, $190 billion of Treasuries, $91 billion of corporate debt and $52 billion of short-term debt.

Together these six jurisdictions account for $7.546 trillion, or 76.7% of the published Asia total. They are a useful map, not a complete definition of the region.

“Equity” includes funds with no look-through

TIC uses a broader equity definition than the everyday idea of directly held listed shares. It includes common stock, preferred and other equity, and shares in investment funds. Fund shares remain equity securities in the survey regardless of what the fund owns underneath.

Asia’s $4.504 trillion equity line consists of $3.156 trillion of common stock, $832 billion of fund shares and $516 billion of preferred and other equity. The fund-share bucket alone is larger than Asia’s entire agency-debt line. Yet Table A8 cannot tell whether those funds ultimately hold technology stocks, bonds, cash, structured credit or a blend.

TIC classifies the legal form of the security held, not necessarily the final economic exposure. A bond fund share still appears in the equity table.

Singapore makes the classification issue especially clear. Of its $759 billion equity total, $405 billion is common stock, $104 billion fund shares and $250 billion preferred or other equity. South Korea reports $372 billion of common stock, $122 billion of funds and $97 billion of preferred and other instruments.

The distinction also matters when reading annual changes. Foreign holdings of U.S. equities rose by $2.98 trillion globally between June 2024 and June 2025. Table 1 of the full report attributes $2.24 trillion to valuation changes and $640 billion to net purchases, with the remainder from stock-swap adjustments and a statistical residual. Most of the increase in the stock was therefore market appreciation, not a near-$3 trillion foreign buying wave. That decomposition is global, not Asia-specific.

A market-value position always contains two stories: what the investor bought and what prices subsequently did. Collapsing them turns an equity rally into an imaginary geopolitical vote.

Inside the agency line: American mortgages

Agency debt sounds close to a conventional government bond. In TIC it is overwhelmingly mortgage-backed. An asset-backed security, or ABS, converts streams of loan payments into tradable securities. Of the $1.344 trillion of long-term agency debt held by all foreign residents, $1.311 trillion was agency ABS, or 97.5%. For Asia, agency ABS was $735 billion against $741 billion of long-term agency debt, roughly 99.2%. The report says these securities are mainly residential mortgage-backed securities, or MBS.

Japan’s $250 billion long-term agency line is entirely classified as agency ABS after rounding. Taiwan reports $184 billion of agency ABS against $185 billion of agency debt, mainland China $186 billion against $186 billion, and South Korea $33 billion against $34 billion. This is not mainly a book of simple agency debentures. It is a claim on mortgage cash flows wrapped in guarantees whose legal form varies by issuer.

Prepayment risk changes the timing of cash flows. It is distinct from homeowner default risk. Agency guarantees reduce some credit exposure but do not freeze duration.

An MBS gives U.S. homeowners partial control over the investor’s cash-flow calendar. When rates fall, borrowers refinance and principal comes back sooner, just when reinvestment yields are lower. When rates rise, refinancing slows, principal returns later and effective duration, a measure of the price and cash-flow schedule’s sensitivity to rates, extends. The New York Fed describes this negative convexity, while the SEC separates prepayment risk from credit risk.

For an Asian insurer, that creates a moving target. MBS can supply yield and duration, but the duration shortens in falling-rate environments and lengthens when yields rise. Rate hedges may need to be adjusted as prepayment behaviour changes. The price is driven not just by the yield curve but by millions of household refinancing decisions.

Corporate debt is still too broad a box

Asian residents held $848 billion of long-term U.S. corporate debt in Table A5. Exhibit 4 adds short-term corporate paper and takes the broad corporate line to $865 billion. The category contains conventional bonds, structured securities and other debt issued by U.S.-resident companies.

ABS detail improves visibility only partially. Global foreign holdings of long-term corporate debt were $4.831 trillion, including $315 billion of corporate ABS. For Asia, Table A9 identifies $50 billion of corporate ABS: $30 billion of private MBS and $20 billion of non-mortgage ABS. Most visible corporate debt is therefore non-ABS.

That finding does not prove Asia has little exposure to U.S. CLOs. A collateralised loan obligation, or CLO, finances a pool of leveraged loans to already indebted companies by issuing several layers of securities. Many CLOs holding loans to American companies are issued by Cayman-domiciled vehicles. TIC classifies securities by issuer residence. A Cayman CLO note is a foreign security in the U.S. system even when the vehicle owns U.S. leveraged loans. The Federal Reserve’s historical note Who Owns U.S. CLO Securities? shows how that domicile breaks the intuitive map.

Two errors follow. It is wrong to call all corporate ABS “CLOs”. It is equally wrong to treat a small corporate-ABS number as Asia’s complete exposure to U.S. leveraged lending. The public tables available on 5 September 2026 do not provide a country-by-tranche reconstruction.

The Federal Reserve’s May 2026 Financial Stability Report and the OFR document vulnerabilities in the leveraged-loan market. They do not deliver a country-by-tranche matrix of Asian holders as of 5 September 2026. That is an unknown, not a licence to manufacture an estimate.

Treasuries remain the foundation

Federal debt still matters enormously: $3.284 trillion of long-term Treasuries in Table A5 and about $440 billion of short-term Treasury debt in Table A7. U.S. sovereign credit risk, interest-rate risk and currency risk are different objects.

A Treasury can repay in full at maturity while showing a large market loss after yields rise. A Japanese or Korean investor can earn a positive dollar return and lose in local-currency terms if the yen or won appreciates. A reserve manager may hold through maturity while a fund facing redemptions may need to sell. The same security, carrying the same CUSIP, the standard identifier used across North American securities markets, has a different risk inside a different balance sheet.

The monthly major-holders table put Japan at $1.1167 trillion and mainland China at $633.4 billion in June 2026. Those data are newer but narrower and still subject to imperfect custody attribution. They do not overwrite the detailed annual survey. They simply show that the order of magnitude had not vanished one year later.

Institutional disclosures use incompatible maps

Moving from country tables to named institutions helps, but it makes aggregation hazardous. The GPIF FY2025 summary reports two totals that should not be merged. Its asset-size series shows ¥293.6437 trillion at 31 March 2026. Its allocation table, which also includes reserves managed in the Pension Special Account before final settlement, totals ¥299.8254 trillion. The ¥73.3990 trillion of foreign bonds and ¥74.3569 trillion of foreign equities belong to this second perimeter. They cover all foreign markets, not just the United States. GPIF also classifies yen-hedged foreign bonds as domestic bonds. Issuer nationality, asset currency and accounting bucket already diverge inside one fund.

South Korea’s National Pension Service reported KRW1,865.6 trillion of assets at end-June 2026. Its global-equity book was KRW661.1 trillion; North America represented 68% of that sleeve at end-2025. Global fixed income stood at KRW110.2 trillion at end-June 2026, with securitized assets making up 13.7% of the end-2025 composition. Its bond benchmark leaves USD-KRW unhedged while partially hedging other currencies into dollars.

These disclosures make the risk tangible. They do not reconcile line by line with TIC. Dates differ, North America is broader than the United States, alternative assets may contain U.S. exposure and classification rules differ.

In Taiwan, the opening article in this series estimated life insurers’ foreign investments at roughly $724 billion at end-2025 using the central bank’s data. That total spans every foreign market, currency and permitted asset. Comparing it with the $854 billion assigned to all Taiwanese residents in TIC offers scale, never an accounting identity.

The visible country can be the custodian’s country

TIC tracks securities at impressive granularity. Geography is less clean. A Japanese investor may keep a security with an intermediary elsewhere. A Singapore fund may manage money for clients in other jurisdictions. A Cayman vehicle may finance American assets. The legal residence displayed in the table need not be the saver’s residence.

Not every investment chain contains every intermediary shown. The diagram explains why a country table is a residence and custody statistic, not a global beneficial-ownership register.

Treasury calls the issue custodial bias and warns explicitly about major financial and custody centres. Singapore and Hong Kong therefore have to be read as both local economies and regional platforms. That does not invalidate their figures. It prevents automatic attribution to domestic households or institutions alone.

Other blind spots follow from scope:

direct investment and controlling stakes belong mainly to BEA statistics;
direct loans are not negotiable securities and may not enter SHL;
rate and FX derivatives are captured elsewhere;
non-negotiable deposits are not bonds;
fund shares provide no asset look-through;
an offshore vehicle can issue a foreign security backed by American economic risk.

The survey is precise inside its contract. Trouble begins when it is asked to answer a different question.

A mark-to-market loss is not yet forced selling

The asset map allows a disciplined transmission story. An equity correction first hits the especially large equity lines in the Korean and Singaporean totals, as well as Japan’s sizeable reported book. Higher yields affect Treasuries through duration and MBS through both duration and convexity. Wider spreads hit corporate bonds. Mortgage-market shifts change MBS prepayments. Funding stress acts on short-term paper and derivative collateral.

The market channel is only the first stage. Forced sales usually require a constraint: redemptions, margin calls, collateral, regulatory capital, insurance payouts or an immediate liquidity need.

This exposure can also be a rational choice. U.S. markets provide depth that many domestic markets cannot. Treasuries provide liquidity and collateral. Agency MBS add yield and reduce part of the credit risk without eliminating duration or prepayment risk. Corporate bonds diversify income. For a long-horizon pension fund, a temporary drawdown can remain a manageable mark-to-market move.

The size of the portfolio is not enough to measure risk. What matters is the intersection of assets and liabilities: what currency must be paid, when policyholders or retirees need cash, what collateral derivatives demand, which solvency threshold binds, and how much can be sold without moving the market?

PORTFOLIO STRESS TEST

Same amount, very different risks

Choose a TIC reporting perimeter and apply a separate price shock to each asset class. The calculation measures a gross market-value move, not a regulatory-capital loss.

local calculation · zero tracker

Published perimeters

Amounts come from Table A5 of the TIC survey at 30 June 2025. Rounded categories can differ from the reported total by $1bn.

Scenario results

Gross change in value$-1,246.5bn-12.67 %

Sum of categories: $9,835bn
Value after shock: $8,588.5bn
Largest contributor: equities and fund shares
Reported total: $9,834bn
Rounding gap versus reported total: +$1bn

Contributions by class

Equities and fund shares: $-900.8bn
Long-term Treasuries: $-197.04bn
Agency debt: $-59.28bn
Corporate debt: $-84.8bn
Short-term debt: $-4.58bn

Visible formulas

Class change = observed amount × price shock. Total change = sum of the five contributions.

ΔV = Σ(encours × choc / 100)

Scope and limitations

The model excludes liabilities, FX hedges, derivatives, convexity, coupons, defaults, forced sales, tax, regulatory capital and indirect holdings inside funds. An entered shock is neither a forecast nor a calibrated crisis scenario.

Source for holdings

U.S. Treasury, Federal Reserve Bank of New York and Federal Reserve Board, SHL survey at 30 June 2025, Table A5.

Model v1.0.0

The tool applies separate shocks to the five A5 categories. It does not estimate sector capital loss. Its purpose is narrower: one assumption such as “U.S. rates rise” cannot describe the response of equities, Treasuries, MBS and corporate credit.

What the data establish, and what remains hidden

The survey establishes three points. Asia’s financial link with the United States extends well beyond federal-debt purchases: at the survey date, the category containing equities, fund shares and other capital instruments was the largest.

Different mechanisms sit beneath the aggregates. Agency debt is almost entirely mortgage-backed. The equity category includes funds whose underlying assets are not visible. U.S. corporate debt misses offshore-issued CLOs backed by American loans.

Finally, geography does not always identify who bears the risk. TIC measures foreign holdings of U.S. securities remarkably well, but custody chains, funds and special vehicles prevent every dollar from being assigned to a particular central bank, insurer or pension beneficiary.

What remains unknown is equally important: net exposure after hedging, collateral timing, structured-credit tranches, sale thresholds and the ultimate beneficiary behind some financial-centre positions. That boundary is not a weakness in the inquiry. It identifies the next step: from countries to balance sheets, then from balance sheets to individual securities.

Main source set

The core data come from the final SHL report at 30 June 2025, released on 30 April 2026. The Treasury liabilities survey page provides the archive. The later but far narrower Treasury check uses the monthly major-holders table and the June 2026 TIC release.

Institutional examples use the GPIF FY2025 summary, the NPS overview, global-equity, global-fixed-income and benchmark pages, and Taiwan’s central-bank financial-stability report.

Risk mechanisms draw on the New York Fed’s MBS convexity analysis, the SEC MBS disclosure study, the Federal Reserve’s May 2026 Financial Stability Report, the OFR 2025 risk review, and the Fed’s historical note Who Owns U.S. CLO Securities?. Treasury’s warning on custodial bias limits country-by-country interpretation.

What Asia really buys in America