CoreWeave Bond Yields Hit 15% as Investors Scrutinize $52 Billion in AI Cloud Debt
CoreWeave's bond yields hit 15% as investors question whether its $103.7B backlog can service $52B in debt and leases. What AI compute financing stress means for buyers.
What happened
CoreWeave's bond yields have touched 15% and its shares dropped sharply as investors questioned whether its $103.7 billion in contracted backlog can support $35.6 billion in debt and $16.3 billion in lease commitments. The loss-making AI cloud provider — one of the largest "neoclouds" renting GPU capacity — now faces about $52 billion in combined debt and lease obligations, according to reporting over the weekend, including a detailed investor defense published by Fortune.
The selloff reflects a broader scrutiny of the AI infrastructure trade: enormous contracted demand on one side of the ledger, enormous financing obligations on the other — with the whole structure depending on AI demand continuing to grow fast enough to service both.
Why this matters
CoreWeave is the purest bet on AI compute demand, and the market is now pricing in real doubt. Turbine costs are up more than 195% since 2019, according to Wood Mackenzie, and GE Vernova is offering delivery slots as far out as 2031. You can buy the chips; you cannot buy the power. The industry is responding with private power deals, fuel cells, repurposed plants — and Elon Musk's most extreme answer: move compute into space.
For the AI economy, the question is no longer whether demand for compute is real, but whether the financing stacked on top of it is.
What this means for your business
1. Stress-test your AI vendors' finances. If a critical workflow depends on a specialized AI cloud provider, check their balance sheet the way you would any key supplier. High yields mean expensive refinancing — and the risk lands on customers if terms change or capacity tightens.
2. Don't single-source your compute. Build your AI workloads to be portable across providers. The 15%-yield episode is a reminder that vendor concentration risk is financial risk, not just technical risk.
3. Price the power, not just the tokens. Electricity and grid access are becoming the binding constraint on AI infrastructure — and costs there will eventually reach buyers. Model your AI operating costs with energy pass-throughs in mind.
4. Watch contract enforceability. If you hold long-term compute contracts, confirm what protections you have if your provider restructures. Contracted backlog looks reassuring until it has to service debt it cannot refinance.
5. Separate AI demand from AI finance. The underlying demand for AI capability remains strong. What is under stress is the leverage built on top of it — an important distinction when planning multi-year AI investments.