A growing body of research is pointing to the same uncomfortable number: somewhere around a third of US economic growth right now traces back to AI. Not AI products people use daily, but the spending behind them.
Think of the chips, data centers, software, and stock market gains that all that spending has produced. The Wall Street Journal reported the figure last week, citing Oxford Economics, and several other research entities have landed on similar estimates using different methods. Why can that be disturbing? As they say, one shouldn’t put all eggs in the same basket. Nevertheless, markets are now close to doing exactly that with artificial intelligence.
Three Factors Driving AI Domination
Oxford Economics economist Michael Pearce breaks the effect into three pieces. First, there’s the direct spending: tech companies buying compute, building data centers, and writing enormous checks for chips and software. Second, that construction boom is creating jobs and tax revenue in the towns where the data centers go up. Third, and maybe less obvious, is the wealth effect. AI stocks have driven a chunk of this year’s market gains, and richer households tend to spend more.
“Without this investment boom, I think it’s pretty clear the economy would be running cooler,” Pearce told the WSJ.
Barclays economist Jonathan Millar agrees: “It’s very much an AI-driven economy right now.”
The numbers behind that statement are substantial enough not to ignore. US corporate AI investment has hit an annualized $1.5 trillion, up 50% from just two years ago. Data center construction spending reached $68.3 billion in June, according to Commerce Department figures. Meanwhile, every other category of private construction, from housing to hospitals, actually shrank by over $100 billion over the same period.
The Federal Reserve Bank of St. Louis found that AI-adjacent investment contributed nearly a full percentage point to GDP growth in the first three quarters of 2025, which worked out to 39% of total growth. That share is well above the 28% contribution IT investment made at the peak of the dot-com boom.
Capital Economics puts the US figure at roughly a third of growth over the past year too, and mentions another noteworthy observation: the trend is not limited to the United States. China’s economy is now leaning on AI even more heavily, with more than half of its recent quarterly growth tied to AI activity in some way.
Expert Opinions on AI Hyperconcentration
Ask around Wall Street and you get a split reaction. JPMorgan CEO Jamie Dimon has said he isn’t losing sleep over the data center boom fading. Asked directly by CNBC whether a slowdown in AI spending would threaten the wider economy, he answered simply: “No, not really.” His reasoning is that even if hyperscalers pull back on new spending, the data centers already built will keep generating real value. He’s flagged bumps ahead, particularly around whether individual projects pan out as planned, but he doesn’t see the concentration itself as the danger.
Others are less relaxed. Peter Berezin at BCA Research has argued that without the AI surge, “it’s certainly plausible” the US would already be in a recession. In particular, he was pointing to a labor market that’s been cooling all year and non-AI business investment that’s stayed roughly flat since 2019. Outside of data centers, commercial construction, such as office towers, shopping centers, etc. has been going nowhere.
Why Fintech Should Watch AI Growth Carefully
Payment processors, BNPL lenders, and consumer credit models have spent the past two years underwriting against a backdrop where consumer spending kept holding up better than the labor market alone would suggest. Part of the reason it held up is that household wealth swelled alongside the stock market, and AI stocks did a lot of that lifting.
If AI capex growth slows sharply, or if monetization keeps lagging the scale of the spending (in fact, that’s a gap several of the hyperscalers’ own earnings reports have already exposed), the wealth effect that’s been quietly supporting consumer spending could go into reverse. That’s a real exposure for any underwriting model built on the assumption that spending trends from the past year or two will simply continue. It’s less about whether AI itself is a bubble and more about how much of today’s “normal” consumer behavior actually depends on it holding together.
