Meta (META) and GPU cloud operator CoreWeave (CRWV) have dramatically expanded a previously announced partnership, turning an initial commitment that topped out at $14.2 billion into a roughly $21 billion, multi-year pact that locks in dedicated AI compute for Meta through 2032.
Under the deal, CoreWeave will provide Meta with dedicated cloud capacity across multiple locations, including early deployments of NVIDIA’s next‑generation Vera Rubin platform. Vera Rubin is NVIDIA’s new supercomputing architecture designed to accelerate not just model training but large-scale, continuous AI inference — the always-on processing that powers live AI services. NVIDIA and industry observers say this next-gen stack could help shift AI from today’s “generative” models toward more “agentic” systems that proactively take steps to accomplish goals rather than just responding to prompts.
A major emphasis of the agreement is inference workload capacity, reflecting a broader industry move away from one-off training jobs toward sustained, large-scale AI operations. That trend makes long-term, guaranteed access to specialized GPUs and data-center capacity strategically important for companies building consumer-facing AI products.
“This is another example that leading companies are choosing CoreWeave’s AI cloud to run their most demanding workloads,” CoreWeave CEO Michael Intrator said. The market reacted: CoreWeave shares rose about 3% in pre-market trading after the announcement.
Why it matters: the deal underscores how hyperscalers and large tech firms are locking in long-duration compute commitments to secure the GPU horsepower needed for next-generation AI features and services. It’s also another major commercial win for NVIDIA’s ecosystem as Vera Rubin rolls out in production environments.
Related reading: Mark Zuckerberg’s Meta is reportedly planning a stablecoin comeback later this year.
Read more AI-generated news on: undefined/news
