CoreWeave Adds 9 of Top 10 AI Model Providers, Including Anthropic

The strategic value of specialized AI infrastructure is on full display as CoreWeave secures a major new partnership. The company announced a multi-year agreement with Anthropic on April 10, 2026, granting the creator of Claude access to its extensive network of Nvidia GPUs across U.S. data centers. This deal, coming just one day after a massive $21 billion expansion with Meta, solidifies a remarkable customer base. CoreWeave now counts nine of the ten leading AI model providers as clients, a testament to its central role in powering production-scale artificial intelligence. The company’s financial trajectory underscores this demand, with revenue guidance for 2026 exceeding $12 billion, supported by a contracted backlog that has ballooned to over $66 billion.

This ascent traces back to an unlikely origin. Founded in 2017 as a cryptocurrency mining venture, the company pivoted to offering GPU-on-demand cloud services when crypto profits dwindled. That decision positioned it perfectly for the AI boom. Its accumulated stockpile of Nvidia hardware became a critical asset, fueling a public listing in March 2025 and explosive growth. CoreWeave now operates 32 data centers housing more than 250,000 GPUs. Its 2025 revenue of $5.13 billion marked a 168 percent annual increase, but this growth has highlighted a significant concentration risk. Microsoft alone accounted for roughly 67 percent of last year’s revenue, a dependency that raises questions as the tech giant develops its own AI models. The back-to-back deals with Meta and Anthropic represent a deliberate push to diversify its customer base and reduce reliance on any single hyperscaler.

For Anthropic, this partnership addresses a specific and urgent need within a complex compute strategy. The company’s annualized revenue has skyrocketed, surpassing $30 billion in early April 2026. Supporting this breakneck growth, driven by enterprise adoption of Claude, requires infrastructure across multiple platforms. Anthropic already runs primary training on Amazon’s Trainium chips and has secured future tensor processing unit capacity through Google and Broadcom. The CoreWeave agreement establishes a crucial third lane: high-performance Nvidia GPUs optimized for production inference workloads. This capacity is essential for delivering the scale and low latency that enterprise clients demand, directly supporting Anthropic’s ecosystem expansion.

CoreWeave’s leadership emphasizes that their role extends beyond mere hardware provision. “AI is no longer just about infrastructure, it’s about the platforms that turn models into real-world impact,” stated CEO Michael Intrator. He described working with Anthropic at the point “where models are put to work” as the core of the company’s mission. The partnership will begin with a phased infrastructure rollout, with flexibility to expand. While the specific Nvidia chips involved were not disclosed, CoreWeave’s inventory spans current and next-generation architectures, including the forthcoming Vera Rubin GPUs slated for volume shipment in late 2026.

The scale of CoreWeave’s market penetration is now unmistakable. Its client roster, built alongside anchor tenant Microsoft, includes giants like Meta, OpenAI, Mistral, Cohere, and IBM. The Meta expansion announced on April 9 commits an additional $21 billion for capacity from 2027 through 2032, bringing the total value of that relationship to approximately $35 billion. These consecutive announcements signal a fundamental business shift. CoreWeave is moving beyond short-term rentals, locking in long-duration contracted revenue that provides predictable, visible growth. This model was further validated in March 2026 when the company raised an $8.5 billion debt facility backed by its GPU assets and the Meta contract.

A revealing irony emerged on the same day as the Anthropic news. Reports indicated Anthropic is exploring designs for its own custom AI chips, a path that could eventually lessen its need for third-party Nvidia infrastructure. This tension defines the current moment. Leading AI firms are making massive, long-term commitments to rented GPU capacity while simultaneously investing in proprietary silicon for future independence. The timelines for custom chip development simply cannot keep pace with immediate compute demand. CoreWeave’s success is a structural bet that Nvidia-native cloud capacity will remain a competitively necessary layer of the AI stack for the duration of these decade-long contracts. The market is evolving from a transitional solution into a permanent fixture, and CoreWeave’s landmark week is the clearest evidence of that enduring shift.