'How many canaries have we killed in the coal mine?' — Wall Street debates if the latest AI bubble warnings are another head fake or the real thing

Wall Street keeps hunting for the moment the AI trade finally cracks, and it keeps coming up empty-handed. Since ChatGPT arrived in late 2022, every new wobble has been treated like a possible last warning: Nvidia’s march past $2 trillion, then $3 trillion, then $4 trillion and $5 trillion in market value; the debt-fueled spending sprees at AI hyperscalers; the DeepSeek shock in early 2025; Sam Altman saying a bubble exists; Michael Burry’s bearish bets; the Iran war; South Korea’s market slide; and those “circular” funding deals between AI companies. Each time, the alarms got louder. Each time, stocks mostly recovered and went on to new highs.

Now the latest stress point is CoreWeave, the neocloud firm whose shares have tumbled and whose credit default swaps have jumped. That has revived the familiar question: is this just another false alarm, or the first real crack in the wall? Dan Greenhaus, a strategist at Solus Alternative Asset Management, says people have been declaring victory over the AI boom at every hiccup for three years. In his view, CoreWeave is not a canary in the coal mine so much as another bird that Wall Street has already misread several times.

Still, the backdrop is getting less forgiving. Borrowing costs have climbed as the Iran war and oil prices pushed bond yields to their highest levels in more than 20 years, and the Federal Reserve is raising rates again to fight stubborn inflation. That matters because AI expansion has leaned harder and harder on debt financing. Greenhaus says CoreWeave’s unsecured debt is the “riskier of the riskier AI buildout play,” and he does not think AI companies can keep funding themselves forever if rates keep moving up. The question, in his telling, is not whether pressure will show up. It is whether this is the breaking point or just another 25 or 50 basis points away.

CoreWeave is pushing back. CEO Mike Intrator told employees the company’s backlog is more than enough to cover its debt obligations, and Nvidia’s Jensen Huang said the company does not need a financial backstop. Huang also said CoreWeave’s strength is improving and that its cost of capital is actually falling. Plenty on Wall Street still see the recent selloff as a buying chance. But not everyone is buying the story.

Gartner warned on Wednesday that AI data center owners would need $8.5 trillion in cumulative revenue through 2030 to cover spending, financing costs, and a 12% return. Gartner’s estimate for total enterprise and consumer AI revenue heading to data centers tops out at $3.6 trillion, which leaves more than $4 trillion in asset value exposed to write-down risk through 2032. Bank of America took the longer view on Friday: this looks like a classic capital cycle, the kind that usually ends badly even when the underlying technology turns out to be real. Competition is already rising, especially from Chinese open-weight models, and business spending on AI has started to dip as higher token volumes fail to make up for lower token prices. That is the part that should make investors uneasy, not the latest round of dramatic hand-wringing.