CoreWeave bonds slide as Meta plans competing in-house AI cloud infrastructure
CoreWeave's euro-denominated junk bonds fell nearly 2 cents over two days to 96.8 cents on the dollar, while its $1.25 billion dollar bond (also issued in June) dropped 1.3 cents to 97.3. The €2 billion ($2.3 billion) euro note was the first euro-denominated junk bond issued by a U.S. AI infrastructure company, opening European pension funds and insurers to AI infrastructure debt for the first time. The bond weakness reflects growing investor scrutiny of AI infrastructure financing even as new capital continues to flow into the sector.
CoreWeave has raised more than $20 billion through stock and debt sales in 2026, a remarkable pace. The company holds $66+ billion in contracted revenue backlog but carries approximately $30 billion in total debt, roughly triple what it was a year earlier. While Meta committed $35 billion total to CoreWeave infrastructure (including an additional $21 billion announced in April for 2027-2032), the company now faces competition from Meta building its own AI cloud services and other hyperscalers pursuing direct capacity buildout instead of outsourcing to specialist providers.
The credit market is shifting from exuberance to scrutiny: higher spreads reflect demand for increased compensation for risk, not outright rejection of the AI buildout. Core risks include GPU collateral depreciation (Nvidia refreshes chips annually), customer concentration (Meta, OpenAI, Anthropic control terms), and the unknown useful life of hardware backing $30B+ in debt. CoreWeave's argument rests on contracted revenue visibility, but falling GPU prices and rising refinancing costs pose structural headwinds.
For capital market practitioners and LPs, this signals an inflection in AI infrastructure financing: equity-stage risk (startup execution, TAM capture) has transferred to credit risk (asset depreciation, customer churn, fixed-cost overhangs). CoreWeave's widening spreads and competing Meta capacity suggest the junk-bond wave for AI infrastructure is pricing in execution risk that equity investors have not yet fully discounted.