Meta Commits Another $21 Billion to CoreWeave

CoreWeave (CRWV) and Meta Platforms (META) announced on April 9, 2026 an expanded long-term AI cloud capacity agreement of approximately $21 billion running through December 2032. The new commitment stacks on a prior $14.2 billion arrangement disclosed in September, lifting cumulative contracted volume between the two parties to $35.2 billion (Rogue Alpha roll-up from disclosed CoreWeave-Meta amounts). Initial deployments under the new agreement will include the NVIDIA Vera Rubin platform across multiple locations. CoreWeave separately disclosed plans to raise $3 billion in fresh debt the same day.

Meta is locking in AI computing capacity from CoreWeave for several years, instead of relying only on data centers it builds itself. That matters because it gives CoreWeave a large named customer commitment while showing how expensive the AI buildout has become.

CoreWeave (CRWV) is the direct counterparty carrying the contracted capacity and the matched debt expansion; its shares rose 3.5% on the disclosure. Meta Platforms (META) is the named buyer, with shares up 2.6%, and the agreement sits inside its stated $115 billion to $135 billion capex range for the year. NVIDIA (NVDA) is named as the silicon supplier through the Vera Rubin reference, with the contracted deployments scheduled to include some of the platform's initial installations.

NVIDIA (NVDA) gains a publicly named anchor for early Vera Rubin deployment volume, with downstream readout for HBM supplier Micron Technology (MU) and networking supplier Arista Networks (ANET) once delivery cadence becomes visible in CoreWeave filings. Server integrators Dell Technologies (DELL) and Super Micro Computer (SMCI) sit in the build path for the multi-location dedicated capacity. Power and thermal infrastructure providers Vertiv Holdings (VRT) and Eaton (ETN) are exposed through the distributed deployment footprint. Other hyperscalers — Microsoft (MSFT), Alphabet (GOOGL), Amazon (AMZN) — and competing capacity provider Oracle (ORCL) now face a public reference contract defining duration, dollar size, and platform-cycle alignment for hyperscaler third-party AI cloud commitments.

This commitment does not establish revenue timing for CoreWeave; capacity delivery dates and recognition schedules are not in the announcement. It does not guarantee deployment cadence for the NVIDIA Vera Rubin platform beyond the named initial deployments. It does not imply equivalent third-party cloud commitments from Microsoft, Alphabet, or Amazon, and it does not disclose unit pricing per GPU-hour or per cluster.