CoreWeave powers Anthropic
CoreWeave signed a multi‑year deal to supply Anthropic with cloud computing capacity to run Claude at production scale, a move that highlights capacity contracts becoming central to AI strategy. The agreement sent CoreWeave shares higher and reinforces that compute availability — not just model quality — is shaping competitive advantage. Several major model providers are now on CoreWeave’s platform, making capacity partnerships a strategic asset across the industry. (reuters.com, cnbc.com)
A chatbot only looks instant on your screen because somebody already rented the machines behind it. On April 10, 2026, CoreWeave said Anthropic signed a multi-year deal for cloud capacity to develop and run the Claude family of models. (coreweave.com) That “capacity” is rows of specialized chips in data centers, not a software subscription. CoreWeave said the new compute for Anthropic will start coming online later in 2026, which means this deal is about securing future supply before demand hits. (coreweave.com) Investors treated the announcement like a supply win, not a branding exercise. Reuters reported CoreWeave shares jumped more than 13% on April 10 after the Anthropic deal was announced. (reuters.com) The timing was not random. One day earlier, CoreWeave said Meta had expanded its own commitment by another $21 billion, on top of a prior $14.2 billion arrangement for artificial intelligence cloud infrastructure. (cnbc.com) That puts CoreWeave in a new role: less like a typical cloud vendor selling generic server time, more like a landlord controlling scarce factory floor for artificial intelligence. CNBC said the Anthropic agreement adds another major model provider to a platform already serving companies such as Meta, OpenAI, Microsoft, and Cohere. (cnbc.com) Anthropic has been building exactly this kind of multi-supplier map. On April 6, 2026, Anthropic said it was expanding work with Google and Broadcom, while also saying Amazon Web Services remains its primary cloud provider and training partner for Project Rainier. (anthropic.com) Anthropic also spelled out why it wants more than one source of chips. The company said it trains and runs Claude on Amazon Web Services Trainium chips, Google tensor processing units, and Nvidia graphics processing units so it can match different jobs to different hardware and improve resilience. (anthropic.com) CoreWeave’s pitch is speed and specialization. The company said the Anthropic deal will support both development and deployment, which means the same supplier is helping with the expensive training phase and the always-on serving phase when millions of users ask Claude questions. (investors.coreweave.com) That is why these contracts keep getting bigger. If a model company runs short on chips, it cannot launch new features on time, cannot keep response times low at peak hours, and cannot easily shift customers to a rival model without rewriting products around a different system. (anthropic.com, cnbc.com) So the contest is no longer just who has the smartest model on a benchmark chart. It is also who locked up enough electricity, buildings, networking gear, and Nvidia processors months before everyone else tried to buy the same thing. (reuters.com, cnbc.com)
