CoreWeave finalizes a multi-year agreement with Anthropic as nine out of ten leading AI model providers integrate into its platform.

      In summary: CoreWeave declared a multi-year agreement with Anthropic on April 10, 2026, allowing the creator of Claude to utilize Nvidia GPU capacity across data centers in the U.S. for production-scale AI tasks. Financial details were not revealed. This announcement follows CoreWeave's recent $21 billion expansion of its partnership with Meta and adds Anthropic to a client list that now includes nine out of the ten top AI model providers. CoreWeave reported revenues of $5.13 billion in 2025 and anticipates exceeding $12 billion in 2026, supported by a contracted backlog exceeding $66 billion.

      From a crypto miner to AI’s service provider

      Established in 2017 as Atlantic Crypto, CoreWeave began as an Ethereum mining operation that procured Nvidia graphics processing units in bulk to mine cryptocurrency and lease excess GPU capacity to other miners. However, after crypto margins diminished in 2019, the company rebranded as CoreWeave and shifted its focus to GPU-on-demand cloud services for general computing. This pivot coincided with the AI model training surge starting in 2023, transforming CoreWeave's inventory of Nvidia hardware into a highly valuable asset in the technology sector. The company became publicly traded on Nasdaq under the ticker CRWV on March 28, 2025, at $40 per share, raising $1.5 billion and bringing its valuation to about $23 billion. CoreWeave currently runs 32 data centers with over 250,000 GPUs and 1.3 gigawatts of contracted power capacity. Its revenue of $5.13 billion in 2025 marked a 168 percent increase from the previous year, with projections exceeding $12 billion for 2026 against a contracted backlog of over $66 billion.

      The rapid growth of the company has introduced a significant concentration risk: Microsoft represented around 67 percent of CoreWeave’s revenue in 2025, a reliance that investors and analysts noted prior to the IPO. Microsoft’s ambition to create its own AI models introduces an additional strategic factor, raising concerns about how much of Microsoft's computational demand might eventually transition to its own infrastructure instead of relying on third-party GPU cloud services. The agreement with Anthropic, coming just a day after the $21 billion expansion with Meta, signifies CoreWeave's effort to diversify its customer base and decrease reliance on any single hyperscaler.

      What Anthropic is investing in

      Anthropic's computing strategy has become increasingly intricate alongside its revenue growth. By early April 2026, the company’s annualized revenue run rate had exceeded $30 billion, more than three times the $9 billion reported at the end of 2025. This rapid growth, fueled by enterprise adoption of Claude and the rapid expansion of Claude Code, has necessitated Anthropic to enhance its infrastructure commitments across various chip architectures simultaneously. Its primary training workloads utilize Amazon Web Services Trainium hardware through Project Rainier, a vast cluster comprising hundreds of thousands of AI chips in multiple U.S. data centers. Just three days before the CoreWeave announcement, Anthropic secured a deal with Google and Broadcom for multi-gigawatt TPU capacity, gaining access to approximately 3.5 gigawatts of next-generation tensor processing unit compute expected to become operational in 2027. The CoreWeave agreement adds a third component: Nvidia GPU capacity tailored for production inference workloads that meet the scale and latency performance needed for enterprise Claude deployments. Anthropic's commitment of $100 million to its Claude partner network earlier this year indicated the company’s intention to grow the ecosystem of developers and businesses working with Claude, which is now playing a direct role in its computing procurement strategies for deals like this one.

      CoreWeave's co-founder and CEO Michael Intrator described the arrangement in terms that extend beyond sheer infrastructure capacity. “AI is no longer solely about infrastructure; it’s about platforms that transform models into real-world outcomes,” he stated. “We are eager to collaborate with Anthropic at the core of where models are applied, and their production performance is realized. This aligns perfectly with the practical deployment of AI that CoreWeave was designed for.” Anthropic will initially implement compute through a phased infrastructure rollout, retaining the option for future expansion. Specific details concerning the Nvidia chip architectures involved have not been publicly disclosed, although CoreWeave's inventory includes both current and next-generation Nvidia GPUs. Nvidia’s Vera Rubin GPUs, introduced at GTC 2026, represent the next crucial architecture in CoreWeave’s deployment plan, with significant shipments anticipated in the latter half of 2026.

      Nine out of ten, two agreements in 48 hours

      With the Anthropic agreement, nine of the ten leading AI model providers now utilize CoreWeave’s platform, a market penetration figure highlighted in the company’s press release. The client list, which has grown alongside Microsoft, also includes Meta, OpenAI, Mistral, Cohere, IBM, and Nvidia,