Pulse24 Original
AI Spending Is Now Behind Half of America's Economic Growth Since Last Year. Software Budgets Alone Rose Almost 90% Last Quarter.
November 10, 2026
AI spending is now behind roughly half of all US economic growth since early 2025, and the third quarter's investment numbers show why. Software budgets alone rose almost 90 percent, even as the imports behind the buildout dragged on the same report.
Software spending at American businesses rose 89.6% in the third quarter. Equipment tied to information processing climbed 37%. Both figures sit inside the same government release that put headline GDP growth at 3.4% annualized, the fastest pace in over a year and well above the second quarter's 2.2%.
Strip out the AI-related portion of that number and the picture changes. Gregory Daco, chief economist at EY-Parthenon, has been tracking the contribution on a running basis. His tally puts AI-linked private investment, spanning data centers, power facilities, information-processing equipment, and the software and research built to run on top of them, at an average of 1.1 percentage points of annualized quarterly growth since the first quarter of 2025. Given that the broader economy has expanded by roughly 2.2 points a quarter over that same stretch, a single category of spending accounts for close to half of it.
[[IMG1]]
What Changed
The acceleration shows up most clearly in the investment data itself. Business investment overall rose 12.7% in the third quarter, the sharpest advance outside the pandemic rebound since 2011. Equipment spending climbed 20.9%. Information-processing equipment, the servers, networking gear, and related hardware that fill a data center, rose 37%. Software spending rose 89.6%. Intellectual property investment, which includes research and development, climbed 8.4%.
None of those are small, incremental moves. A software budget that grows by nearly 90% in a single quarter looks less like routine maintenance and more like a company standing up something close to new.
The contribution hasn't moved in a straight line. By Daco's figures, AI-linked spending added about 1.3 percentage points to growth in each of the first two quarters of 2025, fell to roughly 0.6 points in the third quarter of that year, climbed back to just under a point by year-end, then peaked near 1.6 points in the first quarter of 2026. It eased in the second quarter before rebounding again in the quarter just reported.
One detail complicates the celebration. A large share of the hardware behind this boom is imported. Purchases of computers, peripherals, and related parts rose 62% year over year in the third quarter, and that import flow, combined with a drop in exports, pulled 3.3 percentage points off headline GDP. The AI investment cycle added to growth and subtracted from it within the same report, depending on which side of the ledger a given dollar landed on.
Much of the capital behind that import bill isn't coming from retained earnings. The US Treasury borrowed a record $839 billion in a single week last month, and data-center operators have been just as active in credit markets, selling bonds and taking out loans to build capacity ahead of demand rather than behind it.
Why It Matters
Two things lift this past the status of an academic footnote. Concentration is one of them: when half of growth traces back to a single category of spending, built and leased by a relatively small number of companies, the overall growth rate becomes more sensitive to decisions made in a handful of boardrooms than it would be if growth were spread across households, small businesses, and a wider range of industries. The stock market has already split along a similar line this earnings season: strip out AI-linked names and the S&P 500's own performance tells a noticeably different story.
Timing is the other. This growth print lands in the middle of the Fed's December calculus. EY-Parthenon's forecast still assumes policymakers raise the benchmark rate by 25 basis points next month, pushing the range to 4.00% to 4.25%. A 3.4% growth rate, even one built mostly on a single category of investment, doesn't make a comfortable case for holding rates steady while prices keep running hot. October's jobs report, which more than doubled what economists expected, already pointed the same direction.
Daco's own language suggests he isn't arguing against the investment itself. He described an AI-driven cycle that "continues to exceed expectations in both scale and speed," adding that it "remains a powerful tailwind and could provide a meaningful boost to productivity over time." His caution was narrower: the economy, in his words, is becoming "increasingly reliant on a narrow and capital-intensive source of growth."
What to Watch Next
A few signals will say more about how durable this is than any single GDP print: whether the AI contribution keeps rebounding or the second quarter's dip turns into something longer, what happens to the debt and credit spreads financing the data-center buildout, and how Fed officials talk about the composition of growth, not just its pace, in the weeks before the December meeting.
The Pulse24 Take
A headline of "3.4% growth" reads as unambiguous good news, and taken on its own, it is. What's worth sitting with is how unevenly that growth is distributed. A single category of investment, tied to a relative handful of companies and increasingly funded by debt rather than cash on hand, accounts for something close to half of everything the US economy has added since early 2025.
That isn't a reason to bet against the AI buildout. Productivity gains from this kind of infrastructure tend to show up with a lag, and the historical case for transformative technology investment eventually paying off has real precedent behind it. But narrow growth is fragile growth by definition. If AI capital spending merely plateaus rather than reverses, the headline growth rate has further to fall than most investors seem to be pricing in right now. Watching the AI contribution line itself, not just the GDP total sitting on top of it, is the more useful habit going forward.
How we read the data
Curious how we get from raw data to a take like this? Our Trader's Toolkit walks through the tools we lean on.
Explore the Toolkit