CoreWeave has announced a massive $21 billion expansion of its cloud capacity agreement with Meta, securing the social media giant’s AI infrastructure needs through December 2032. The eight-year deal positions CoreWeave as a primary compute provider for Meta’s increasingly complex generative AI and large language model (LLM) development.
A key component of the agreement is the early-access deployment of NVIDIA’s Vera Rubin platform. This next-generation architecture is designed to optimize performance and system resilience for high-concurrency AI workloads, offering the scalability required for Meta’s global operations. The infrastructure will be distributed across multiple regional data centers to ensure low-latency access and redundant capacity.
According to Michael Intrator, Co-Founder and CEO of CoreWeave, the deal reinforces the industry-wide shift toward specialized AI clouds for demanding enterprise workloads. Since its Nasdaq listing in March 2025, CoreWeave has aggressively expanded its footprint to meet the skyrocketing demand for dedicated GPU compute, effectively establishing a long-term “compute moat” for Meta as it scales its AI ecosystem.