Meta Locks In $21B CoreWeave Deal, Betting Big on AI Compute Through 2032
When Llama 4 underperformed last year, Meta knew it had a compute problem. This week it showed how seriously it's addressing it.
image from grok
Meta has committed an additional $21 billion to CoreWeave through 2032, bringing total contracted revenue to $35 billion, securing early access to Nvidia's Vera Rubin chips—reportedly 2x faster than the current Blackwell platform. The deal underscores how AI compute has become the primary competitive moat, with Meta spending up to $135B on AI infrastructure this year after Llama 4 underperformed; the company has since built a dedicated superintelligence lab and released its Muse Spark model. The transaction also highlights the consolidation of AI compute supply around hyperscaler-adjacent providers like CoreWeave, with bond markets now underwriting these multi-billion-dollar infrastructure bets at scale.
- •Meta's $35B total commitment to CoreWeave reflects a strategic recovery play following Llama 4's underperformance and the establishment of its superintelligence lab.
- •Early access to Nvidia Vera Rubin chips is a key deal differentiator—chip allocation priority has become a material competitive advantage in foundation model development.
- •CoreWeave's 2026 capex guidance of $35B (up from $14.9B in 2025) signals accelerating GPU infrastructure buildout despite market concentration risks.
