Meta and NVIDIA have announced a sweeping multi-year strategic partnership to build hyperscale AI data centers optimized for both training and inference. The deal covers millions of NVIDIA Blackwell and next-generation Rubin GPUs, making it one of the largest AI chip agreements to date. Notably, Meta becomes the first company to deploy NVIDIA Grace CPUs as standalone processors in its data centers, rather than pairing them with GPUs.

The partnership also integrates NVIDIA Spectrum-X Ethernet switches into Meta's Facebook Open Switching System platform. Both companies are collaborating on deploying NVIDIA's upcoming Vera CPUs, with potential large-scale deployment in 2027. The deal emphasizes performance-per-watt efficiency and advances the broader Arm software ecosystem.

While the exact financial terms were not disclosed, chip analysts estimate the deal is worth tens of billions of dollars. Meta CEO Mark Zuckerberg has previously announced plans to spend over $60 billion on AI infrastructure in 2026 alone.