In short: CoreWeave announced a multi-year agreement with Anthropic on April 10, 2026, giving the Claude maker access to Nvidia GPU capacity across US data centres for production-scale AI workloads. Financial terms were not disclosed. The deal arrives one day after CoreWeave announced a $21 billion expansion of its Meta partnership, and adds Anthropic to a customer roster that now covers nine of the ten leading AI model providers. CoreWeave generated $5.13 billion in revenue in 2025 and is guiding for more than $12 billion in 2026, backed by a contracted backlog exceeding $66 billion.
Ex-crypto miner becomes AI’s landlord
CoreWeave was founded in 2017 as Atlantic Crypto, an Ethereum mining operation that bought Nvidia graphics processing units in bulk to mine cryptocurrency and rent spare GPU capacity to other miners. When crypto margins compressed in 2019, the company renamed itself CoreWeave and pivoted to GPU-on-demand cloud services for general computing purposes. The timing proved transformative: the AI model training boom that began in earnest in 2023 turned CoreWeave’s stockpile of Nvidia hardware into one of the most strategically valuable infrastructure positions in technology. The company went public on Nasdaq under the ticker CRWV on March 28, 2025, at $40 per share, raising $1.5 billion and valuing it at approximately $23 billion. CoreWeave operates 32 data centres with more than 250,000 GPUs and 1.3 gigawatts of contracted power capacity. Its 2025 revenue of $5.13 billion represented a 168 per cent increase year-on-year, and management has guided for more than $12 billion in 2026 revenue against a contracted backlog that now exceeds $66 billion.
The company’s rapid growth has come with a significant concentration risk: Microsoft accounted for approximately 67 per cent of CoreWeave’s 2025 revenue, a dependence that investors and analysts flagged in the run-up to the IPO. Microsoft’s push to develop its own AI models adds a further strategic variable, raising the question of how much of Microsoft’s compute demand will eventually shift toward in-house infrastructure rather than third-party GPU cloud rental. The Anthropic deal, arriving the day after a $21 billion Meta expansion, represents CoreWeave’s most visible effort to build a diversified customer base that reduces its dependence on any single hyperscaler.
What Anthropic is paying for
Anthropic’s compute strategy has grown more complex alongside its revenue. The company’s annualised revenue run rate surpassed $30 billion in early April 2026, more than three times the $9 billion figure it recorded at the end of 2025. That rate of acceleration, driven by enterprise Claude adoption and the breakout growth of Claude Code, has required Anthropic to expand its infrastructure commitments across multiple chip architectures simultaneously. Its primary training workloads run on Amazon Web Services Trainium hardware via Project Rainier, a large-scale cluster spanning hundreds of thousands of AI chips across multiple US data centres. Three days before the CoreWeave announcement, Anthropic’s deal with Google and Broadcom for multi-gigawatt TPU capacity secured access to approximately 3.5 gigawatts of next-generation tensor processing unit compute expected to come online in 2027. The CoreWeave deal fills a third lane: Nvidia GPU capacity for production inference workloads, running at the scale and latency performance that enterprise Claude deployments require. Anthropic’s $100 million commitment to its Claude partner network earlier this year signalled the company’s intent to expand the ecosystem of developers and enterprises building on Claude, and that ecosystem expansion is now directly driving the compute procurement decisions behind deals like this one.
CoreWeave co-founder and CEO Michael Intrator framed the deal in terms that go beyond raw infrastructure capacity. “AI is no longer just about infrastructure, it’s about the platforms that turn models into real-world impact,” he said. “We’re excited to work with Anthropic at the centre of where models are put to work and performance in production shows up. It’s exactly the kind of real-world deployment of AI that CoreWeave was built for.” Anthropic will initially deploy compute under a phased infrastructure rollout, with the option to expand the arrangement over time. The specific Nvidia chip architectures involved have not been publicly disclosed, though CoreWeave’s estate spans current and next-generation Nvidia GPU generations. Nvidia’s Vera Rubin GPUs, unveiled at GTC 2026, represent the next major architecture in CoreWeave’s deployment roadmap, with volume shipments expected in the second half of 2026.
Nine of ten, two deals in 48 hours
The Anthropic agreement means that nine of the ten leading AI model providers now use CoreWeave’s platform, a market penetration figure the company cited in its press release. The customer roster built alongside Microsoft includes Meta, OpenAI, Mistral, Cohere, IBM, and Nvidia itself, as well as a sub-leasing arrangement through which Microsoft supplies some CoreWeave capacity to third-party clients. The Meta relationship deepened significantly on April 9, 2026, one day before the Anthropic announcement: Meta committed an additional $21 billion to CoreWeave for dedicated AI cloud capacity running from 2027 through December 2032, bringing the total value of the two companies’ infrastructure relationship to approximately $35 billion. CoreWeave also expanded its agreement with OpenAI by up to $6.5 billion earlier in 2026. The two announcements in 48 hours, covering Meta and Anthropic, illustrate how CoreWeave is converting its infrastructure position into long-duration contracted revenue rather than spot-market GPU rentals. CoreWeave raised $8.5 billion in a GPU-backed debt facility in March 2026, with the Meta relationship used as collateral. The Anthropic deal, while undisclosed in value, will contribute to a backlog that analysts are watching as the primary indicator of the company’s long-term revenue predictability.
The infrastructure tells a story about dependence
The same day CoreWeave announced the Anthropic agreement, reports emerged that Anthropic is exploring the design of its own custom AI chips — a move that would, if realised, eventually reduce its dependence on the Nvidia-powered infrastructure that CoreWeave provides. The irony is deliberate: Anthropic’s current infrastructure commitments across AWS, Google Cloud, and now CoreWeave reflect a company that is simultaneously expanding compute dependency in the short term and exploring the routes to architectural independence in the long term. That tension is not unique to Anthropic. Meta, OpenAI, and Google have all invested heavily in custom silicon programmes while continuing to rent third-party Nvidia capacity, because the timelines for custom chip maturity and the demand curve for AI compute do not align closely enough to allow a clean transition. CoreWeave’s position as the GPU landlord of choice for the AI industry is therefore both a statement about the current moment and a structural bet that Nvidia-native cloud capacity will remain competitively necessary for at least the duration of the contracts now being signed. As AI infrastructure spending accelerated through 2025, the GPU cloud market began to look less like a transitional gap-filler and more like a permanent layer of the AI stack, and CoreWeave, two deals in two days, is the clearest evidence of that shift.
