In short: CoreWeave announced a multi-year agreement with Anthropic on April 10, 2026, granting access to Nvidia GPU capacity across US data centres for production-scale AI workloads. Financial terms remain undisclosed. This announcement follows CoreWeave's recent $21 billion expansion of its partnership with Meta, adding Anthropic to its growing clientele that already includes nine of the ten leading AI model providers. CoreWeave reported $5.13 billion in revenue for 2025 and is projecting over $12 billion in 2026, supported by a contracted backlog exceeding $66 billion.

Ex-crypto Miner Transitions to AI Infrastructure

Founded in 2017 as Atlantic Crypto, CoreWeave initially operated as an Ethereum mining platform, purchasing Nvidia graphics processing units in large quantities to mine cryptocurrency and leasing out excess GPU capacity to fellow miners. However, as cryptocurrency margins declined in 2019, the company rebranded to CoreWeave, shifting its focus to GPU-on-demand cloud services for broader computing needs. This strategic pivot coincided with the explosion of AI model training beginning in 2023, transforming CoreWeave’s Nvidia hardware into a critical asset in the tech infrastructure landscape. CoreWeave made its Nasdaq debut under the ticker CRWV on March 28, 2025, priced at $40 per share, raising $1.5 billion and achieving an approximate valuation of $23 billion. The company operates 32 data centres, housing over 250,000 GPUs and possessing 1.3 gigawatts of contracted power capacity. Its 2025 revenue of $5.13 billion marked a remarkable 168 percent increase year-on-year, with management forecasting revenues exceeding $12 billion in 2026, supported by a backlog surpassing $66 billion.

Despite its impressive growth, CoreWeave faces significant concentration risk, with Microsoft accounting for roughly 67 percent of its 2025 revenue. This dependence has raised concerns among investors and analysts, especially as Microsoft increasingly develops its own AI models, prompting speculation about how much of its compute demand might shift from third-party GPU cloud rentals to in-house solutions. The recent Anthropic agreement, which follows the substantial $21 billion Meta expansion, signifies CoreWeave's proactive approach to diversify its customer base and mitigate reliance on a single hyperscaler.

Anthropic's Growing Compute Needs

Anthropic's compute strategy has evolved alongside its revenue growth. By early April 2026, its annualized revenue run rate exceeded $30 billion, a significant increase from the $9 billion reported at the end of 2025. This acceleration, fueled by enterprise adoption of Claude and rapid growth in Claude Code, necessitates a robust expansion of Anthropic's infrastructure commitments across various chip architectures. Its primary workloads currently leverage Amazon Web Services Trainium hardware through Project Rainier, a large-scale cluster encompassing hundreds of thousands of AI chips across multiple US data centres. Just three days prior to the CoreWeave announcement, Anthropic secured a deal with Google and Broadcom for multi-gigawatt TPU capacity, ensuring access to around 3.5 gigawatts of next-generation tensor processing unit compute expected to be operational by 2027. The CoreWeave partnership complements this by providing Nvidia GPU capacity tailored for production inference workloads, crucial for the performance and scale demanded by enterprise Claude deployments. Earlier this year, Anthropic's $100 million investment in its Claude partner network underscored its commitment to fostering an ecosystem of developers and enterprises utilizing Claude, directly influencing its compute procurement strategies.

CoreWeave's co-founder and CEO, Michael Intrator, articulated the significance of the deal beyond mere infrastructure capacity: “AI is no longer just about infrastructure; it’s about the platforms that turn models into real-world impact.” He expressed enthusiasm for collaborating with Anthropic, emphasizing the deployment of AI in real-world applications, a core mission for CoreWeave. Anthropic will initiate its compute deployment in phases, with an option for future expansion. While specific Nvidia chip architectures involved have not been publicly disclosed, CoreWeave’s facilities encompass both current and next-generation Nvidia GPUs, including the newly unveiled Vera Rubin GPUs expected to ship in the latter half of 2026.

Market Penetration and Strategic Moves

The agreement with Anthropic signifies that nine out of ten leading AI model providers now utilize CoreWeave’s platform, a noteworthy achievement highlighted in the company's press release. CoreWeave’s customer roster, established alongside Microsoft, includes major players like Meta, OpenAI, Mistral, Cohere, IBM, and Nvidia, in addition to a sub-leasing arrangement where Microsoft provides some CoreWeave capacity to third-party clients. The relationship with Meta deepened on April 9, 2026, just one day before the Anthropic announcement, with Meta committing an additional $21 billion to CoreWeave for dedicated AI cloud capacity extending from 2027 until December 2032, bringing the cumulative value of their partnership to approximately $35 billion. Earlier in 2026, CoreWeave also expanded its agreement with OpenAI by up to $6.5 billion. These two major announcements within 48 hours, involving Meta and Anthropic, illustrate CoreWeave’s strategy of transforming its infrastructure advantage into long-term contracted revenue streams rather than relying solely on spot-market GPU rentals. Additionally, CoreWeave secured an $8.5 billion GPU-backed debt facility in March 2026, leveraging its relationship with Meta as collateral. While the financial details of the Anthropic deal remain undisclosed, it is expected to contribute significantly to a backlog that analysts view as a key indicator of the company’s long-term revenue stability.

The Tension of Dependency and Independence

On the same day that CoreWeave revealed the Anthropic agreement, reports surfaced indicating that Anthropic is considering the design of its own custom AI chips. This move could potentially reduce its reliance on the Nvidia-powered infrastructure provided by CoreWeave over time. This scenario is particularly ironic, as Anthropic’s current commitments across AWS, Google Cloud, and now CoreWeave illustrate a company that is both increasing its short-term compute dependency while exploring routes to architectural independence for the future. This dilemma is not unique to Anthropic; other major players like Meta, OpenAI, and Google have similarly invested in custom silicon projects while continuing to utilize third-party Nvidia capacity, given the misalignment between custom chip maturation timelines and the surging demand for AI compute. CoreWeave’s positioning as the preferred GPU landlord for the AI sector represents both a reflection of the current landscape and a strategic bet that Nvidia-native cloud capacity will remain essential for the duration of the newly forged contracts. As AI infrastructure spending surged throughout 2025, the GPU cloud market began to solidify its role as a permanent component of the AI ecosystem, with CoreWeave's two deals in quick succession serving as compelling evidence of this shift.

Source: TNW | Anthropic News