CoreWeave, the Nasdaq-listed AI cloud infrastructure provider, announced Thursday that it has struck an expanded $21 billion agreement with Meta Platforms to supply dedicated GPU computing capacity through December 2032 — its largest single commercial contract since going public last year. The new deal builds on a prior $14.2 billion arrangement signed in September 2025, bringing the total value of commitments between the two companies to approximately $35.2 billion, making Meta one of CoreWeave's anchor customers alongside Microsoft.
Under the terms of the agreement, CoreWeave will deploy dedicated AI cloud capacity across multiple data center locations globally, including some of the first large-scale commercial deployments of Nvidia's Vera Rubin platform — the next-generation AI supercomputing architecture expected to power a shift from today's generative AI models toward more proactive "agentic" AI systems capable of executing complex, multi-step tasks autonomously. The distributed, multi-node deployment architecture is designed to optimize performance and scalability for Meta's increasingly demanding AI inference workloads, which serve billions of users daily across Facebook, Instagram, WhatsApp, and its growing suite of AI products.
"This is another example that leading companies are choosing CoreWeave's AI cloud to run their most demanding workloads," said CoreWeave CEO Michael Intrator in a statement. The deal underscores the degree to which even the largest technology companies — those with the financial firepower to build their own infrastructure — continue to rely on specialized AI cloud providers to bridge immediate capacity gaps and secure access to cutting-edge hardware. Meta's capital expenditure plans for 2026 range between $115 billion and $135 billion, nearly double what the company spent in 2025, and the CoreWeave contract represents a core pillar of that build-out strategy.
CoreWeave's data centers are populated with hundreds of thousands of Nvidia GPUs optimized specifically for AI training and inference workloads, distinguishing it from general-purpose hyperscalers such as Amazon Web Services, Google Cloud, and Microsoft Azure. The company also serves Google, Microsoft, OpenAI, and others, and has said the new Meta contract will ensure no single customer represents more than 35% of total revenue — a significant diversification milestone for a company that had historically derived 62% of its 2024 revenue from Microsoft alone. CoreWeave also said Thursday it plans to raise $3 billion in fresh debt to support continued data center expansion.
Shares of CoreWeave rose approximately 3.5% following the announcement Thursday, while Meta stock climbed 2.6%. For investors, the deal is a read-through signal for Nvidia, whose Vera Rubin chips sit at the center of the CoreWeave infrastructure stack, as well as for broader AI infrastructure plays including data center REITs, power utilities serving hyperscale campuses, and optical networking companies that supply the high-speed interconnects tying GPU clusters together. The announcement further validates a capital expenditure cycle that Wall Street analysts widely describe as durable through at least 2027.