Meta Expands AI Infrastructure with NVIDIA GPUs, Grace CPUs and Vera Plans
Meta enters a multi-year NVIDIA partnership to deploy millions of Blackwell and Rubin GPUs plus Arm-based Grace CPUs in its on-premises and cloud data centers. The agreement integrates Spectrum-X Ethernet switches into Meta’s Open Switching System, adds confidential computing in WhatsApp and plans Vera CPU adoption from 2027.
1. Multi-Year NVIDIA Partnership
Meta has entered a multi-year partnership with NVIDIA to expand its AI infrastructure across both on-premises and cloud-based data centers. The collaboration will deploy millions of NVIDIA Blackwell and Rubin GPUs alongside Arm-based Grace CPUs to strengthen Meta’s capabilities in AI training and inference workloads.
2. Network and Systems Integration
The agreement integrates NVIDIA’s Spectrum-X Ethernet switches into Meta’s Open Switching System platform, aiming to optimize networking performance and power efficiency. Meta will also implement NVIDIA GB300-based systems across its on-premises facilities and via cloud partnerships to achieve a unified architecture and streamlined operations.
3. Future CPU Roadmap and Confidential Computing
Both companies are collaborating on software optimizations for Grace CPUs and preparing for large-scale adoption of NVIDIA Vera CPUs starting in 2027 to enhance energy efficiency. Meta has introduced NVIDIA Confidential Computing in WhatsApp’s processing framework and plans to extend these privacy-focused features across additional products.