CoreWeave’s newest deal with Meta, up to $14.2 billion for cloud compute infrastructure through December 2031 (with a possible extension into 2032), says a lot more than simply that there’s hunger for AI capacity. CoreWeave has been shaped by earlier stress points: where it has shored up weak spots, amplified its strengths, and tried to prove its business model to investors. The Meta deal is another corrective step in its evolution.
When it went public in March 2025, CoreWeave was exposed in plain view: two customers provided 77% of its revenue, with Microsoft alone making up around 62–72%. That degree of customer concentration concerned investors. If one big customer reduced demand, delayed payment, or negotiated harder, CoreWeave’s growth would suffer severely. “Customers are heavily concentrated,” a Forbes article noted in reviewing the IPO.
At a time when compute demand is surging, especially for the kinds of GPU work Meta and OpenAI are doing, that kind of risk is too big to ignore. CoreWeave’s subsequent deals with OpenAI, expanding to about $22.4 billion in 2025, already pushed it toward a more diversified customer base. The Meta contract strengthens that diversification. It gives CoreWeave a way to reduce its dependency on Microsoft. That matters for negotiating leverage and stability.


