The US trade deficit, from the port to the factory

A machine arrives at a US port. Its value enters the import statistics. A few weeks later, it may be installed in a factory or a data centre. The purchase now tells two stories: money spent with a foreign supplier, and productive equipment acquired in the United States. Understanding the trade deficit means following the shipment beyond customs.

On September 30, 2026, the Census Bureau released its advance estimate for August goods trade: a $132.6 billion deficit, up from $118.9 billion in July. Purchases from abroad rose faster than sales abroad. These are monthly, seasonally adjusted, current-dollar figures on the customs-based Census basis. The August estimate is subject to revision. 1

The detail shows a large role for capital goods, an increase in industrial supplies, and falling imports in the consumer goods category. That mix deserves a closer reading than the headline balance. It reveals something about US demand while leaving questions about quantities, final uses and investment returns open.

Both sides of the border

From July to August, goods imports rose about $17.4 billion and exports rose $3.7 billion. The gap between those movements accounts for the $13.7 billion widening of the deficit. Finding its source means opening both columns. 1

The most striking import increase came in industrial supplies and materials, up $8.90 billion. But exports in the same category rose $5.95 billion. Once both flows are included, the category’s contribution to the widening falls to $2.95 billion. The United States still had an August goods surplus in this category; that surplus became smaller. l0g calculations using Census end-use categories. 1

Capital goods excluding automotive contributed $4.24 billion to the wider deficit. Imports rose $5.58 billion, outpacing exports. Among the six published categories, capital goods made the largest contribution to the monthly deterioration. 1

Consumer goods excluding automotive tell a less intuitive story. Imports fell, but exports fell more. The category’s deficit therefore widened even as purchases from foreign suppliers declined. Watching only the containers coming into the country would miss half the mechanism. 1

Following the equipment

Capital goods accounted for 43.4% of the value of August goods imports, on our calculation. Their import value was 57.2% higher than in August 2025. Those proportions indicate substantial purchases of equipment in a broad sense. Understanding them requires opening up the category. 1

The official classification includes industrial machinery, generators, computers and accessories, telecommunications equipment and semiconductors. Products are grouped by principal end use. A component in this category might still be incorporated into another machine, held in inventory or re-exported. When translating product flows into national accounts, the BEA distinguishes intermediate uses from final investment purchases. 4 5

There is concrete evidence for the computing connection, with an important date attached. In its full report on July, released on September 3, the BEA recorded substantial increases in imports of computers and computer accessories. Those findings concern the previous month and an earlier statistical vintage. August’s advance report stops at broad product groups. 3

Research by Michael Waugh, published by the Federal Reserve Bank of Minneapolis on April 8, 2026, provides structural context. His classification of products associated with AI infrastructure, built from customs codes with assistance from a large language model, estimates their share at 23% of US imports in 2025. This is a research estimate of product categories potentially used for AI, rather than a direct observation of each shipment’s destination. Applying that share mechanically to August 2026 would manufacture precision. 9

The strongest supported reading is sustained demand for imported equipment in an economy where digital infrastructure matters. Measuring the exact AI share would require more detailed products, purchasers and uses.

Oil tells a different story over a month and a year

Because petroleum is included in industrial supplies, an energy explanation for the increase is tempting. The category also includes metals, chemicals and nonmonetary gold. Those components need to be separated before assigning a specific share of the movement to each. 4

Price data make the time horizon particularly important. In its September 16 release, the Bureau of Labor Statistics reported that the price index for imported petroleum and petroleum products rose 0.1% from July to August, but 27.3% over twelve months. Imported capital goods prices rose 0.9% over the month. Energy can therefore remain expensive relative to the previous year while barely moving from one month to the next. 2

An import bill depends on prices, quantities and the mix of goods purchased. More expensive cargo, a larger number of shipments and a shift towards higher-value products can all increase the dollar total. Identifying quantity growth requires appropriate prices for the products concerned and consistent statistical coverage.

BLS price indexes are not seasonally adjusted; the Census value changes used here are adjusted. Simply subtracting headline price growth from import-value growth would mix measures. The price indexes inform the diagnosis, while a volume decomposition requires separate work. 1 2

The balance changes with its boundaries

The $132.6 billion figure covers goods. Services are collected separately and can reduce the overall deficit. July provides a useful example. In its September 3 release, the BEA reported a $119.6 billion goods deficit, a $31.0 billion services surplus and an overall deficit of $88.6 billion. All three figures belong to the same publication and balance-of-payments basis. 3

That report’s goods figure differs from the customs-based figure used as the comparator for August. The two bases have different coverage conventions and adjustments, and revision schedules can also diverge. Moving between them is a statistical reconciliation task, rather than a gap to erase with an improvised subtraction. 3 7

August’s full balance requires August services and adjustments. The complete release is scheduled for October 6, 2026. Carrying July’s services surplus forward would insert an assumption in place of the very observation that is still awaited. 11

Customs duties introduce another distinction. The import value in the Census report excludes duties, freight, insurance and other charges incurred in bringing goods to the United States. A higher tariff can increase the importer’s bill without adding the same amount directly to the reported customs value. Subsequent economic effects operate through negotiated prices, margins, quantities and supplier choices. 1

A server appears on two lines of GDP

Gross domestic product (GDP) measures production within the country. The expenditure calculation includes consumption and investment, then removes the imported content already included in those purchases. That offset separates spending by Americans from production in America. 7

Consider a hypothetical purchase. A company acquires imported servers for $1 million and pays $200,000 for US installation services. The whole transaction occurs in the same period. Assume the installation services are domestic value added with no additional intermediate inputs, and that there are no taxes or inventory changes.

The company acquires installed equipment worth $1.2 million. In the expenditure measure of GDP, that investment is matched by $1 million of imports. The transaction therefore adds $200,000 of immediate US production in this simplified model. The hardware retains its full value to the purchaser; manufacturing it was foreign production. l0g illustration of the BEA accounting mechanism. 5 7

Later, those servers may support services produced in the United States. The effect will depend on labour, other inputs and demand. The accounts distinguish acquiring the tool, manufacturing it and producing with it. The same reasoning applies to a machine tool or medical equipment.

This distinction helps with growth commentary. Rising imports may contribute negatively to the net-exports line in a GDP decomposition, while the associated expenditure appears in investment or consumption. The overall effect requires both entries, followed by an assessment of any displacement of domestic production. As the BEA explains, imports may complement domestic production or substitute for it. 7

Gold adds a special case. In the national accounts, the BEA removes international-account flows of nonmonetary gold and replaces them with an adjustment based on domestic production and industrial use. This is another reason to reconcile customs trade with GDP rather than transfer the headline straight across. August’s advance release leaves the detailed contribution from gold unresolved. 8

The warehouse adds another clock

Goods may wait between the port and their final buyer. Inventories then determine how sales and production are allocated across periods.

A second hypothetical example makes the timing visible. A retailer imports $100 of goods in the first period and holds them. The addition to inventory contributes 100 to inventory investment, matched by the 100 import offset. Domestic production associated with this transaction is zero at that point in the simplified example.

In the second period, the retailer sells the goods to a household for $120. Consumption rises by 120 and the withdrawal from inventory reduces inventory investment by 100. The remaining 20 is distribution value added. Assume unchanged acquisition prices, no other inputs and no taxes. The local contribution is recorded once, when the distribution service is supplied. 6

Hypothetical period	Consumption	Inventory change	Imports to subtract	GDP
1: import and hold	0	+100	100	0
2: sell the lot	120	−100	0	20

USD, constant acquisition prices, no taxes or other inputs. BEA chapter 7; CSV model.

The observed inventory figures measure something different: the value of stocks at the end of the month. In the August report, wholesale and retail inventories increased by about $6.25 billion and $2.62 billion, calculated from the published levels. They contain goods of different origins and remain value measures. The changes could reflect earlier purchasing, unsold goods, higher prices or a combination. The advance tables do not distinguish those explanations. 1

The BEA measures inventory investment after removing price-related holding gains and losses. Growth contributions also depend on the pace of accumulation: adding 10 to inventories in one period and 5 in the next leaves stocks rising, but reduces the flow of inventory investment by 5. Reading higher stock levels as an automatically positive growth contribution would lose that second difference. 6 12

The equipment still has to earn its cost

More equipment can prepare the ground for future production. Whether it will be used enough and generate sufficient returns to reward the capital committed remains a separate question. Monthly trade records the arrival of the goods; their economic results will unfold over time.

Return to the $1.2 million installation, using a financial model separate from GDP accounting. It operates for five years with no resale value. Annual cash flows arrive at year-end, after running costs but before financing. The hypothetical nominal discount rate is 8%, with no inflation, tax or additional working-capital requirements. Discounting brings future dollars back to today’s value to reflect the waiting time and required return.

The difference between discounted receipts and the initial outlay is the net present value (NPV). Its sign depends on expected cash flows and the chosen discount rate. At $240,000 of annual cash, the present value of five years of receipts falls about $242,000 short of the initial outlay. At $360,000 a year, it exceeds the investment by roughly $237,000. The break-even annual cash flow is about $300,548. 13 All three cases in the chart involve exactly the same initial import and installation. These are teaching calculations, not forecasts for data centres.

The financial detour clarifies the underlying issue: the quality of the spending and the income it can ultimately generate. Underused machines can leave debt to service. Equipment matched to lasting demand can expand productive capacity. The composition of the deficit helps frame the questions; the label “capital goods” still leaves the analyst substantial work.

At the national level, the current account adds cross-border income and current transfers to goods and services trade. It also reflects the gap between national saving and investment. Financing can combine liabilities to foreigners with a reduction in foreign assets; debt, equity and different maturities create different exposures. A monthly goods deficit reveals neither the form of that financing nor its cost or maturity. 10

The next statistical releases

The advance estimate already locates the wider deficit across broad product groups. The full release on October 6 is scheduled to add services, product detail and real measures of August trade. The first estimate of third-quarter GDP is due on October 29. These are distinct stages of the same statistical investigation. 11

The diagnosis will become clearer by following equipment into service, comparing inventories with sales and tracing the domestic value added associated with imported purchases. A rising bill also requires a separation of prices from quantities. Those results can support an account of investment, consumption or input dependence, with the appropriate weight assigned to each.

For now, the figures primarily show an economy buying substantial amounts of equipment, with trade moving differently across products. The balance measures the gap at the border. Understanding what that gap is preparing requires following the goods to the factory, warehouse or customer, and then examining the income they generate.

Further reading

• US GDP and private demand separates investment, external trade and growth.
• Tariffs and importer cash flow follows the bill beyond customs.
• AI: debt, contracts and guarantees examines commitments that survive underused equipment.

Sources

1. U.S. Census Bureau : Advance Economic Indicators Report, August 2026 (CB26-159). Table 1, PDF page 6: current-dollar SA goods, Census basis; tables 2 and 3, page 7: nominal inventories.
2. U.S. Bureau of Labor Statistics : U.S. Import and Export Price Indexes, August 2026 (USDL-26-1514). September 16 release; table 1: petroleum and petroleum products, code 100; capital goods, code 2. NSA price indexes.
3. U.S. Census Bureau / BEA : U.S. International Trade in Goods and Services, July 2026. September 3 vintage, July data: goods and services on a balance-of-payments basis, separate from the August advance estimate.
4. U.S. Census Bureau : Import END-USE codes, 1, 2, 3 and 5 digit. Classification published April 17, 2012; category reference, without observed actual use.
5. BEA : NIPA Handbook, Chapter 6: Private Fixed Investment. Methodological reference.
6. BEA : NIPA Handbook, Chapter 7: Change in Private Inventories. Methodological reference.
7. BEA : NIPA Handbook, Chapter 8: Net Exports of Goods and Services. Methodological reference.
8. BEA : How are exports and imports of nonmonetary gold treated in BEA’s National Economic Accounts?. Methodological reference.
9. Michael E. Waugh / Federal Reserve Bank of Minneapolis : Trade in AI-Related Products, Staff Report 684. Staff Report 684, Michael E. Waugh, published April 8, 2026. Official abstract: a 23% research estimate for 2025, using an LLM-assisted HS10 classification.
10. Atish Rex Ghosh and Uma Ramakrishnan / IMF Finance & Development : Current Account Deficits. Methodological reference.
11. BEA : Release Schedule. Schedule checked October 1, 2026; dates may change.
12. BEA : Change in private inventories (CIPI), glossary. Methodological reference.
13. OpenStax / Rice University : Net Present Value (NPV) Method. Educational reference for the discounting formula; article cash flows are hypothetical.

Scope and method

Checked October 1, 2026. August flows are revisable advance estimates. The decomposition uses only SA levels from the same Census table and vintage. Detail can differ from totals because of rounding and seasonal adjustment. NSA BLS indexes inform the price discussion without directly decomposing SA values into quantities. August goods and services totals remain pending.

The investment, inventory and return examples are separate hypothetical models. They measure neither US growth nor data-centre returns. NPV is reproducible as Σ annual cash / (1 + 0.08)^year − 1,200,000, for years 1 to 5. Break-even cash equals 1,200,000 / Σ(1.08^−year), or $300,547.745 annually, rounded to $300,548. The linked CSVs specify amounts, periods, units and assumptions.