While everyone was watching oil prices and war headlines this week, JPMorgan Chase did something that barely made the news — but probably should have. The nation’s largest bank quietly began marking down the value of loans tied to private-credit portfolios, with many of those loans going to software companies. When the biggest bank in America starts repricing collateral in the hottest lending market on the planet, it’s time to pay attention.
Here’s the backdrop. Private credit has exploded into a multi-trillion-dollar industry over the past decade, filling the gap left after regulators tightened traditional bank lending post-2008. Borrowers get flexibility, lenders get yield. But the whole machine depends on one thing: borrowers generating enough cash flow to service their debt. And a growing share of those borrowers are now pouring money into the AI infrastructure buildout — data centers, cloud capacity, chips — where spending is certain but profits are anything but.
Take Oracle as a case study. The stock jumped 14% after reporting strong earnings and assuring investors its AI data center economics still work. CEO Clayton Magouyrk said they’re getting better at building and running these facilities more cheaply. Sounds great — until you run the numbers. Oracle signed a $300 billion cloud deal to deliver 4.5 gigawatts of computing power to OpenAI through 2032. Building that capacity costs roughly $225 billion ($50 billion per gigawatt). The margin for error on $75 billion in gross profit across six years is razor-thin, especially when AI computing is essentially a commodity — every provider runs the same Nvidia chips on the same platform, so pricing comes down to who blinks first.
The stress signals aren’t just theoretical. Private credit giant Blue Owl Capital recently faced a surge of redemption requests from investors in one of its funds, forcing it to restrict withdrawals and sell roughly $1.4 billion in loans to raise liquidity. While AI loans weren’t the specific trigger, a massive share of Blue Owl’s lending book is tied to software companies whose business models may themselves be disrupted by AI — a cruel irony where the very technology they’re financing could undercut the borrowers paying the bills.
JPMorgan’s markdown mirrors steps it took during COVID, which should tell you something about how seriously its risk team is taking this. It’s not a panic signal — it’s a positioning signal. The smart money is recalibrating before the music stops, not after.
For investors, the takeaway isn’t that AI is a bust. It’s that the companies financing the AI buildout are taking on enormous risk that the market hasn’t fully priced in. The picks-and-shovels trade everyone loved in 2024 and 2025 is getting crowded, expensive, and dependent on revenue projections that remain, as one analyst put it, “as inscrutable as a Delphic oracle.” If you’re chasing the AI infrastructure trade, make sure you understand who’s holding the bag when the spending doesn’t produce the returns Wall Street is modeling.