Over the last year, as its CEO Sam Altman preached a gospel of insatiable compute, OpenAI has created a web of deals that tie a meaningful chunk of Silicon Valley’s AI buildout to its own trajectory. Big names like Nvidia, Oracle and SoftBank have all inked infrastructure contracts with the ChatGPT maker, but there is one company perched further out on a limb than the rest: CoreWeave, an AI cloud company with a roughly $60 billion market cap. The Wall Street Journal reported Monday that OpenAI missed internal projections for revenue and user growth. It claimed OpenAI CFO Sarah Friar is worried the company may not be able to pay for future computing contracts. If that’s even directionally right, it will land hardest on CoreWeave—which counts OpenAI as one of its biggest customers and has borrowed more than $40 billion in mostly high-interest debt used to finance GPUs and data centers. CoreWeave’s view, at least publicly: it can ride out turbulence as long as demand for AI compute keeps outrunning supply. "OpenAI is a terrific partner, but not our only one,” a CoreWeave spokesperson said, namechecking other big-name customers including Meta, Anthropic, Microsoft and Google. “As more companies build and deploy AI, demand for compute continues to grow. We continue to see demand exceed supply across the AI ecosystem.” Problem is: that “AI ecosystem” is not a broad-based consumer market so much as a coterie of spenders writing very large checks. Trillions of dollars’ worth of infrastructure commitments are concentrated in a few places: big tech balance sheets (Oracle, Meta, Microsoft and Nvidia) and a handful of newer entrants that buy AI capacity and then rent it out (like CoreWeave, Nebius and Nscale). CoreWeave’s model—buy GPUs, spin up data centers, lease the capacity to labs—turns that concentration into both opportunity and fragility. By Phoebe Liu, Reporter Richard Nieva, Senior Writer. Learn more about your ad choices. Visit megaphone.fm/adchoices