JPMorgan Just Quietly Flagged the Biggest Risk in AI

While everyone was watching oil prices and war headlines last week, JPMorgan Chase did something that barely made a ripple — but probably should have. The nation’s largest bank quietly started marking down the value of certain loans tied to private credit portfolios. Most of those loans? Made to software companies.

On the surface, a bank adjusting collateral values doesn’t scream “breaking news.” But when you understand what’s underneath, the picture gets a lot more interesting — and a lot more uncomfortable for anyone all-in on the AI trade.

Here’s the connection most people are missing: private credit has ballooned into a multi-trillion-dollar industry over the past decade. After the 2008 crisis, regulators tightened bank lending, and private lenders rushed in to fill the gap. For borrowers, the appeal is flexibility. For lenders, it’s yield. But the whole system rests on one assumption — that borrowers keep generating enough cash flow to service their debt. A growing share of those borrowers are now technology companies pouring enormous sums into AI infrastructure. Data centers, cloud buildouts, GPU clusters — the spending is staggering.

Take Oracle as a case study. Shares jumped 14% last week after the company reassured investors it could finance its aggressive AI expansion without raising additional debt in 2026. Wall Street cheered. But dig into the math and things get dicey. Oracle signed a $300 billion cloud deal to provide 4.5 gigawatts of computing power to OpenAI between 2027 and 2032. Each gigawatt costs roughly $50 billion to build — $35 billion for Nvidia chips alone, plus another $15 billion for supporting infrastructure. That’s $225 billion in capital expenditure just to fulfill one contract.

The revenue from that deal? About $300 billion over five years, or $60 billion annually. Subtract the build costs, maintenance, and financing — and the margins start looking razor-thin. This isn’t a guaranteed gold mine. It’s a massive bet that AI demand will not only persist but accelerate enough to justify the spend. And Oracle isn’t alone. Nearly every major cloud and tech company is making a version of this same wager.

That’s what makes JPMorgan’s move so telling. When the biggest bank in the country starts quietly reducing the value of loans to the very companies powering this boom, it’s not panic — it’s prudence. They’ve seen this movie before. The parallels to the early days of the telecom bubble aren’t exact, but the pattern rhymes: massive infrastructure spending funded by debt, justified by demand projections that may or may not materialize.

None of this means AI is a bust. The technology is real, the demand is real, and certain companies will generate enormous returns. But there’s a growing gap between what’s being spent and what’s being earned — and that gap is being financed by an increasingly complex web of private debt. When JPMorgan starts tapping the brakes, even gently, smart investors pay attention. The AI trade isn’t dead. But the easy money phase might be.