
Not just a speed problem
Bill Eigen isn’t worried about whether AI is growing. He’s worried about how fast it’s still growing. On CNBC, the J.P. Morgan Asset Management CIO said investors are fixated on the headline numbers, but the real tell is the “second derivative” — basically, whether growth is accelerating or starting to wobble.
And that’s where his eyebrows go up. Eigen says signs of deceleration are already showing up in capital spending and private-market valuations tied to AI labs. In other words: the AI story may still look shiny, but the rate at which the machine is becoming shinier might be slowing.
The balance-sheet boomerang
Here’s the part that sounds a little like a finance thriller. Eigen compared the AI cycle to a real estate cycle, pointing to long-dated debt and lease commitments backing data centers while the chips inside them may lose value far faster.
That mismatch matters because:
- bonds can stretch for 30 years
- AI hardware can age out in 3 to 6 years
- a lot of the obligations sit off the balance sheet, where they’re easier to ignore until they aren’t
Goldman Sachs estimates roughly $1.5 trillion in hyperscaler lease commitments, and a Wall Street Journal report put Alphabet, Amazon, Meta, and Microsoft at about $3 trillion in off-balance-sheet commitments. That’s not pocket change. That’s “are we still pretending this is casual?” money.
The market is calm. Maybe too calm.
Eigen also noted that credit spreads are still hanging out near historic tights, even as some AI-related credit default swaps — including Nvidia’s — have started to widen. That’s the kind of divergence that makes cautious investors squint at the scoreboard.
He’s not shorting the trade, because timing a cycle reversal is a fool’s errand. But he’s also not chasing it. And honestly, that’s probably the mood on Wall Street right now: nobody wants to miss the next AI leg up, but nobody wants to be the last one holding the expensive server rack.
Big picture: the AI narrative is still alive, but this warning says investors may need to watch financing, commitments, and growth momentum as closely as the chips themselves.
