The warning signs began in the bond market, where investors who had financed the AI data-center rush started asking a less glamorous question than stock traders: will they be repaid? Credit analyst Robert Schiffman said “some of the AI boom’s weaknesses can show up first” in credit markets, especially among smaller, debt-dependent neocloud operators.¹

CoreWeave became the clearest test case. The company borrows to build data centers and lease computing power, but its funding needs are immense: at the end of June, it carried roughly $35 billion in debt, another $16 billion in lease liabilities and reported negative free cash flow of $5.7 billion in its latest quarter.²

By Wednesday, the anxiety had become a rout. Its bonds sold off for a second consecutive day, driving yields above 15%, while the stock dropped about 15.5%. PitchBook’s John Atkins called CoreWeave “a bellwether for the riskier cusp of hyperscaler sentiment,” as investors recoiled from borrowing that may need refinancing at punishing rates.³ CoreWeave CEO Michael Intrator pushed back internally, arguing that markets can “misunderstand and act irrationally” when a new technology company is being built.³

The divide is now stark. Evan Schlossman of CoreWeave investor Neostellar argues that $103.7 billion in remaining performance obligations — much of it tied to non-cancellable, take-or-pay contracts — gives lenders a reliable route to repayment. “We are confident in our backlog, which more than supports all of our debt obligations,” the company said.⁴

Skeptics are unconvinced. Accountant and investigative journalist Francine McKenna warned against assuming Microsoft would rescue a supplier if AI-computing demand weakens: “Microsoft is ruthless and will eat CoreWeave for lunch” if that serves its interests.⁴ The market’s verdict is not yet a collapse, but it is a demand for proof: AI demand alone will not settle the question if capital remains expensive.