The US economy is relying on AI capex by Microsoft, Amazon, Alphabet, Meta and Nvidia, as traditional sectors stagnate, risking recession if the boom ends
AI has grown from a useful tool, to something driving economic growth.
Investment in data centres and computing infrastructure now represent a huge portion of GDP in developed economies.
In the US, this dependence has reached a point where economists are questioning what happens if the boom ends.
Recent volatility in AI stocks has exposed how tied the US economy has become to AI-related investment and wealth creation.
The numbers show that business investment in AI may have accounted for half of the growth in gross domestic product, adjusted for inflation, during the first six months of this year.
“It’s certainly plausible that the economy would already be in a recession” without the AI boom, says Peter Berezin, Chief Global Strategist at BCA Research, an investment research firm, according to The Wall Street Journal.
Strip away AI spending and the picture looks worrying.
Private business investment excluding AI-related categories has flatlined since 2019, according to Deutsche Bank.
Commercial construction outside of data centres – shopping centres, office buildings and the like – is in decline. Job creation has slowed and unemployment is creeping upward.
“It’s the only source of investment right now,” says Stephen Juneau, an economist at Bank of America.
It’s becoming an uncomfortable reality for policymakers watching traditional economic engines sputter.
How Microsoft, Amazon, Alphabet and Meta are leading an AI spending surge
Four companies are carrying the weight.
Bank of America estimates that Microsoft, Amazon, Alphabet and Meta will make US$344bn in capital expenditures this year. That’s 1.1% of GDP – up from US$228bn last year.
The impact on growth is huge.
Barclays reckons that investment in software, computer equipment and data centres boosted GDP growth by around one percentage point in the first half of 2025.
Chips such as those sold by Nvidia make up the bulk of AI spending, though most are imported.
Even accounting for that, AI spending still increased output by 0.8% in the first half of the year.
GDP grew by 1.6% during the period, which means that without AI-related spending, growth would have limped along at 0.8%.
Nvidia says that it anticipates US$65bn in sales in the fourth quarter, beating analyst predictions.
Bank of America expects the big four tech companies to pour another US$404bn into capital expenditures next year, though the pace of growth is expected to moderate.
The construction sector is feeling the effects directly.
Data centres now account for around 35% of Turner Construction’s backlog in the US, up from around 13% five years ago.
Building one requires anywhere from 100 to 5,000 people, says Ben Kaplan, Managing Director of the advanced technology group at Turner Construction.
The surge in demand is straining supply chains. Lead times for electrical generators, switchgear and other equipment have stretched by months in some cases.
“Every element of the supply chain is being stressed right now,” says Ben.
Why stock valuations create a recession risk
The concentration of growth in AI investment creates vulnerabilities.
Stock price to earnings ratios are hovering near record highs – and if profit predictions prove overly optimistic, share prices could tumble.
The S&P 500 fell about 2% last week despite a Friday rally, with bubble concerns weighing on investor sentiment.
The wealth effect matters here.
JPMorgan Chase calculates that rising prices of AI stocks alone boosted consumer spending by US$180bn over the past year, as households spend a portion of paper gains.
If those gains evaporate, consumption could fall sharply.
Jonathan Millar, Senior US Economist at Barclays, estimates that a 20% to 30% stock market decline could shave one to 1.5% off GDP growth over roughly a year.
If AI investment simply stopped growing, that could knock another 0.5 point off. If it collapsed entirely, a full percentage point disappears.
The scale of AI-related borrowing compounds the risk.
Oracle’s debt recently sailed past US$100bn after the database software and cloud computing company sold US$18bn in bonds for AI infrastructure.
Companies such as CoreWeave, which lease data centres and rent servers to tech companies, have borrowed heavily to fund rapid expansion.
If revenues don’t materialise to service that debt, the pain could spread through credit markets.
Peter warns that the combination of factors could tip a weakening economy into contraction: “If you take a fragile labour market and you kick it with a capex bust, you’re probably going to get a recession out of it,” he says.