Two major AI infrastructure deals in two days. That’s the pace CoreWeave is moving at right now, and the market noticed. CNBC confirmed CoreWeave’s stock rose 11% on news of a multi-year infrastructure agreement with Anthropic, coming one day after a separately reported $21 billion expanded agreement with Meta. The Anthropic deal is valued at a reported $6.8 billion, according to trade coverage, though that specific figure has not been confirmed in primary source text. What is confirmed: the agreement exists, it’s multi-year, and it’s significant enough to move CoreWeave’s stock double digits.
The terms are worth unpacking carefully. CoreWeave’s investor relations page confirms a multi-year agreement with Anthropic, describing a phased infrastructure rollout. The $6.8 billion valuation of that agreement appears widely in trade coverage but comes from secondary reporting rather than primary company disclosure, readers tracking this deal should treat that figure as reported, not confirmed. The GPU specification cited in some coverage, NVIDIA Vera Rubin architecture, similarly requires qualification: it appears in the Wire’s research summary but isn’t substantiated in the source material available for this review. Under the infrastructure services agreement, Anthropic will run production-scale workloads on CoreWeave’s platform.
The framing matters here. CoreWeave isn’t a legacy public cloud provider. It’s a GPU-specialized infrastructure company that went public in 2025, at an initial valuation of approximately $23 billion fully diluted, per reporting from The Information, and its market valuation has risen sharply since. That trajectory tells a story about where frontier AI labs are choosing to place their compute: not primarily with AWS, Azure, or GCP, but with infrastructure providers built from the ground up around GPU density and AI workload optimization. The CoreWeave/Anthropic agreement is new data in that pattern.
This isn’t an isolated deal. It’s the latest evidence in a thesis this hub established earlier when covering Anthropic’s gigawatt-scale compute commitments: frontier AI labs are making multi-year, multi-billion-dollar infrastructure bets that lock in vendor relationships at a scale that changes the competitive landscape for cloud providers. CoreWeave is winning those bets right now. The Meta agreement, reported at $21 billion, positions the company as infrastructure backbone for two of the most resource-intensive AI development programs in the world simultaneously.
For enterprise technology leaders, the practical implication is a narrowing field of specialized AI cloud providers capable of operating at this scale. That concentration creates both efficiency opportunities and vendor dependency risks. The labs making these deals are essentially co-investing in infrastructure providers, Anthropic’s long-term workload commitments are, in effect, a capital allocation decision as much as a technical one.
What to watch: CoreWeave’s infrastructure buildout cadence over the next 90 days will indicate whether the company can absorb back-to-back hyperscale commitments without execution risk. Watch also for whether other frontier labs, OpenAI, Google DeepMind, Meta’s own internal teams, accelerate their own specialized cloud agreements in response. The $21 billion Meta deal was announced first; the prior Anthropic infrastructure brief covering Google and Broadcom TPU capacity remains relevant context for how Anthropic is diversifying its compute base even as it adds CoreWeave.
The pattern is accelerating. Two hyperscale infrastructure deals in 48 hours, targeting the same provider, from two of the most compute-intensive AI programs in the world. Whatever figure ultimately gets confirmed for the Anthropic agreement, the direction of capital concentration is unmistakable.