CoreWeave signs multi-year compute agreement with Anthropic
The deal added CoreWeave to Anthropic's existing mix of AWS Trainium, Google TPU and Nvidia GPU capacity, but neither company disclosed a dollar figure or exact capacity.
- Compute & infrastructure
- Minor
Anthropic and CoreWeave announced a multi-year agreement under which Anthropic would run production workloads on CoreWeave’s cloud platform, with capacity expected to come online later in 2026 and the arrangement structured to expand over time. Neither company disclosed a dollar value, exact capacity commitment or specific deployment sites — a contrast with the same-week $21 billion figure CoreWeave and Meta attached to their own expanded agreement.
CoreWeave chief executive Michael Intrator framed the deal around “real-world deployment of AI” rather than around any headline number, and the announcement noted that nine of the top ten AI model providers by CoreWeave’s own count now used its infrastructure in some capacity.
For Anthropic, the agreement added a fourth compute source to a portfolio that already spanned AWS Trainium chips, Google TPUs and Nvidia GPUs, continuing a pattern in which the company spread its training and inference workloads across multiple providers rather than committing exclusively to one. The deal came a day after CoreWeave’s much larger publicised agreement with Meta and in the same week as Anthropic’s own expanded TPU deal with Google and Broadcom, illustrating how compressed the announcement cadence for frontier-lab infrastructure deals had become by early 2026.