CoreWeave Exposes Cracks in the AI Debt Boom

The AI infrastructure boom is meeting a less enthusiastic audience: credit investors who want clearer evidence that expensive computing capacity can generate enough cash to repay its debt.

What happened

CoreWeave's bonds fell to distressed levels as investors reassessed the risks behind debt-funded AI infrastructure. The company has approximately $50 billion of debt, while the cost of insuring that debt against default climbed to around 1,200 basis points. Yields on some bonds reached roughly 15% to 16%.

CoreWeave has large customer commitments from companies including Microsoft, OpenAI and Meta. Some of its financings are secured by GPUs, infrastructure assets and long-term customer contracts rather than relying only on the resale value of chips.

Those protections can strengthen an individual loan, but they do not remove concentration risk when repayment depends heavily on a small group of large technology customers.

Why it matters

Higher financing costs could change the economics of the entire neocloud sector. Lenders may demand lower leverage, stronger covenants, wider interest-rate spreads and projects that are closer to completion before providing capital.

That matters because AI infrastructure providers must spend heavily before data centres begin producing revenue. A small change in borrowing costs can materially affect whether a project is financially viable.

The bigger picture

Private credit could be asked to finance as much as $1 trillion of the wider AI-compute buildout. CoreWeave is therefore more than an isolated credit story: it is becoming a market test for how much debt investors will tolerate while AI demand, utilisation and customer concentration remain uncertain.

The market is not rejecting AI infrastructure. It is starting to price its risks more seriously.