Meta and NVIDIA Announce Multi-Generational AI Infrastructure Alliance: Millions of GPUs, First Large-Scale Grace CPU Deployment

Meta and NVIDIA Forge Landmark AI Infrastructure Alliance

Meta and NVIDIA officially announced a multiyear, multigenerational strategic AI infrastructure partnership on February 17, 2026. The deal—worth tens of billions of dollars according to Bloomberg—is part of Meta's stated commitment to invest $600 billion in US data centers and infrastructure by 2028.

What's in the Deal

The partnership encompasses several major components:

• Millions of NVIDIA Blackwell and Rubin GPUs for both AI training and inference workloads across hyperscale data centers
• First large-scale Grace-only CPU deployment: Meta becomes the first company to deploy NVIDIA Grace central processing units as standalone chips at large scale, delivering significant performance-per-watt improvements
• Potential large-scale deployment of NVIDIA Vera CPUs, targeting 2027
• Integration of NVIDIA Spectrum-X Ethernet switches into Meta's Facebook Open Switching System platform

Why It Matters

The announcement confirms two things: NVIDIA's dominance in AI infrastructure is deepening, not weakening—even as AMD and custom silicon compete for share. And Meta is accelerating its AI infrastructure buildout at a scale that rivals the biggest hyperscalers, spanning Llama model training, Metaverse infrastructure, and AI-native products.

The first large-scale Grace CPU deployment also signals NVIDIA's push into general-purpose compute beyond GPUs—a significant expansion of the company's addressable market.

Source: Meta Builds AI Infrastructure With NVIDIA — NVIDIA Newsroom