Headaches grow as AI-related investment drives 50% of US GDP growth in H1 2025

AI is now carrying the entire economy, and the scale of it is no longer something anyone can brush aside.

The turbulence that hit AI-linked stocks last week showed how exposed the country is, as growth has leaned so heavily on machine-learning spending and the wealth tied to it that a sudden collapse would hit the broader system hard.

Business investment linked to AI may have made up half of all inflation-adjusted GDP growth in the first six months of the year.

Rising AI stocks have also pushed up household wealth in recent months, and that extra wealth has fed straight into consumer spending.

Peter Berezin, chief global strategist at BCA Research, said the picture would look very different without the AI surge. “It’s certainly plausible that the economy would already be in a recession” without the boom, he said.

Job creation improved in September, but hiring has slowed through the year. The unemployment rate is rising. Deutsche Bank says private business investment that doesn’t include AI has stayed mostly flat since 2019. And outside data centers, commercial construction has weakened. Shopping centers, office towers; none of them are seeing real action.

Big Tech pours money into capex as other investment stalls

Stephen Juneau, an economist at Bank of America, said the quiet part out loud: “It’s the only source of investment right now.” And the spending is massive. Bank of America estimates that Microsoft, Amazon, Alphabet, and Meta will pour $344 billion into capital expenditures this year, about 1.1% of GDP, up from $228 billion last year.

Barclays estimates that software, data center projects, and computer hardware together boosted GDP growth by roughly one percentage point annualized in the first half of 2025. Most of that came from AI.

AI chips, mainly from Nvidia, make up the largest slice of the bill, but they’re mostly imported. Once imports are subtracted, Barclays still finds that AI spending added 0.8 percentage points to output in the first half of the year. GDP growth stood at 1.6%. Without AI, it would’ve been 0.8%.

Some companies may be rushing purchases ahead of tariff changes, but analysts still expect spending to rise next year. Nvidia told the market Wednesday that it expects $65 billion in fourth-quarter sales, above predictions. And Bank of America sees Microsoft, Amazon, Alphabet, and Meta lifting capital spending again in 2026, this time to $404 billion.

Long-term hopes that AI will boost worker productivity remain unproven. But in the meantime, rising AI stock prices have fed what economists call the wealth effect. JPMorgan Chase says the jump in AI stocks added 0.9%, or $180 billion, to consumer spending over the last year. Consumer spending has grown 5.6% over the year through August before inflation adjustments.

Data center hiring grows while other sectors weaken

AI’s impact on jobs is uneven. Finished data centers don’t need large teams, and tech employment is down since 2022.

But data center construction has become a rare strong spot in a market hit by high rates, a weak real estate sector, and tighter immigration rules.

Ben Kaplan, managing director at Turner Construction, said data center builds now require 100 to 5,000 workers each. Data center work is now 35% of Turner’s U.S. backlog, up from 13% five years ago.

But the pace is straining supplies. Lead times for generators, switchgear, and similar equipment now stretch by months. “Every element of the supply chain is being stressed right now,” Kaplan said.

The risks match the scale of the boom. Valuations are elevated . The S&P 500 fell about 2% last week on concerns about an AI bubble before rising 1% Friday. A decline in stocks would flip the wealth effect.

Barclays senior U.S. economist Jonathan Millar estimates that a 20% to 30% market drop could cut GDP growth by 1 to 1.5 percentage points over a year. If AI investment growth slows, that could remove another 0.5 point. If it stops, that becomes a full point.

Debt is the other pressure point.

Oracle’s debt has grown past $100 billion after the company sold $18 billion in bonds, with part of the proceeds likely tied to AI build-outs. CoreWeave and similar firms that rent GPUs and storage are also borrowing aggressively.

Berezin says the amount of AI-linked debt isn’t enough to trigger a crisis by itself, but he warns that financial markets are connected, and trouble in one part of the system can spill into another very quickly.

Join Bybit now and claim a $50 bonus in minutes