CoreWeave signs $11.9bn AI infrastructure deal with OpenAI; OpenAI takes $350m stake

CoreWeave, an Nvidia-backed cloud provider specialising in GPU rental, agreed to supply OpenAI with up to $11.9bn of dedicated AI compute capacity under a multi-year contract. As part of the arrangement, OpenAI took a $350m equity stake in CoreWeave through a stock issuance, becoming an investor in the company that supplies it with infrastructure.

CoreWeave chief executive Michael Intrator said the deal underscored the company’s “proven ability to deliver reliable and performant infrastructure services,” while OpenAI’s Sam Altman characterised it as complementary to the company’s existing compute relationships with Microsoft and Oracle, and to the Stargate joint venture with SoftBank. The contract added a fourth major compute supplier to OpenAI’s roster in the space of months, reflecting both the scale of capacity the company said it needed to train and serve increasingly large models and a deliberate strategy of not depending on any single provider.

The timing was notable: the agreement was signed weeks before CoreWeave’s own Nasdaq initial public offering, and a contract of this size with a customer as prominent as OpenAI featured prominently in the listing’s pitch to investors. The deal became an early, concrete example of the circular financing arrangements that characterised the AI infrastructure build-out through 2025 and 2026 — compute providers and model developers taking stakes in one another alongside supply contracts, a structure critics said obscured how much real demand underpinned the spending and made revenue figures harder to interpret independently.