Capital goods imports into the United States hit a record in September, and the AI data-center buildout is a major reason. The U.S. goods trade deficit widened 2.1% to $134.9 billion, its largest level since March 2025 and the third straight monthly increase. Economists had expected a narrowing to about $130.2 billion. The surge in imports of computers, accessories and chips is one of the clearest signs that AI spending is reaching the physical economy, not just corporate earnings.
The Numbers
Goods imports rose 2.4% in September, while goods exports rose 2.6%. Capital goods imports jumped 7.5% to a record, driven by equipment for AI data centers. That category includes computers, accessories, semiconductors and telecom equipment. The capital goods deficit now stands $54.3 billion larger than a year earlier, which is more than the entire widening of the overall goods deficit over that period.
Consumer goods imports rose 3.1% in September but were still 12% lower than a year earlier. In other words, the pressure is coming almost entirely from business equipment rather than household spending. Before the release, the Atlanta Fed's GDPNow model estimated that net exports would subtract 2.86 percentage points from third-quarter GDP growth.
A separate tally of computer and chip imports, cited by 24/7 Wall St. from Kobeissi Letter data, puts the category at $86.9 billion in September. That is up $12.1 billion from August and 151% from a year earlier. The monthly gain was the second-largest on record, behind July's $14.7 billion increase. Imports in these categories have more than quadrupled since the start of 2024.
Where the Pressure Comes From
The imports reflect a buildout of unusual scale. The same 24/7 Wall St. report cites NVIDIA's view that the top five hyperscalers will spend nearly $800 billion on capital projects in 2026 and $1.3 trillion in 2027. Its examples of suppliers facing strong demand include Broadcom, whose AI semiconductor revenue grew 221% to $16.70 billion, and Dell, which ended its quarter with a record $95 billion AI backlog.
The link to AI is an inference rather than a line item. The trade release describes capital goods imports as driven by equipment for AI data centers, but it does not break out how much of the computer and chip surge goes to AI facilities. Readers should treat the connection as strong circumstantial evidence, not a precise measurement.
Caveats
Two numbers in this story need careful reading. The first is the "$1 trillion" figure that circulated in coverage of the data. It is an annualized run rate, about $1.04 trillion, calculated by multiplying a single month of computer and chip imports by twelve. It is not a cumulative total, and it holds only if the pace continues for a full year. One strong month does not establish a trend.
The second caveat is that a single month of data can mislead. The trade deficit has widened for three straight months, but the September gain in capital goods imports is one reading. The next release will show whether it marks a new level or a temporary peak. Consumer goods imports remaining 12% below last year is a useful reminder that the current pressure comes from business investment rather than a broad surge in household buying.
Why It Matters
For business leaders, the trade data is a useful cross-check on AI investment claims. Earnings from chipmakers and cloud providers show revenue, but import data shows the physical equipment flowing into the country to support that buildout. When capital goods imports reach a record and the goods deficit widens for three months running, the AI spending story is visible in shipping data, not just in corporate announcements.
The open question is whether that spending pays off. Investors, policymakers and enterprise buyers will be watching for productivity gains and revenue from AI products large enough to justify the equipment now arriving at ports. Until those returns show up in the numbers, the trade data will keep serving as a running measure of how much bet the economy is placing on AI infrastructure.
