Meta, the social media company, and NVIDIA, the US chipmaker, have agreed a multiyear, multigenerational strategic partnership to build hyperscale data centres and deploy NVIDIA processors and networking technology across Meta's AI infrastructure.
The agreement covers on-premises, cloud and AI infrastructure and is designed to support Meta's long-term roadmap for training and inference at hyperscale, the companies said.
Meta will expand its deployment of Arm-based NVIDIA Grace central processing units (CPUs) in production, in what the companies describe as the first large-scale NVIDIA Grace-only deployment.
The partners are collaborating on software and CPU ecosystem library co-design to improve efficiency, and are also working together on NVIDIA Vera CPUs, with a potential large-scale deployment planned for 2027.
Meta will deploy NVIDIA GB300-based systems and adopt the NVIDIA Spectrum-X Ethernet networking platform, which the companies say will deliver low-latency, AI-scale networking while improving utilisation and power efficiency.
The partnership also covers security. Meta has adopted NVIDIA Confidential Computing for private processing on WhatsApp and plans to extend those capabilities to other Meta products.
NVIDIA and Meta engineering teams are engaged in deep co-design work to optimise and accelerate AI models across Meta's core workloads, including its personalisation and recommendation systems, which serve billions of users.
Jensen Huang, founder and chief executive of NVIDIA, said no organisation deployed AI at Meta's scale, integrating frontier research with industrial-scale infrastructure.
Related reading
- India's largest manufacturers adopt NVIDIA infrastructure for AI-driven factory upgrades
- NVIDIA says its newest AI servers cut inference costs by 35x, as cloud giants race to deploy
- OpenAI commits $7.5M to The Alignment Project
Mark Zuckerberg, founder and chief executive of Meta, said the company was excited to build leading-edge clusters using NVIDIA's Vera Rubin platform to deliver what he described as personal superintelligence to everyone in the world.
Financial terms of the partnership were not disclosed.
The recap
- Meta and NVIDIA announced a multiyear, multigenerational strategic partnership.
- Deployment includes millions of NVIDIA Blackwell and Rubin GPUs.
- Potential large-scale NVIDIA Vera CPU deployment targeted for 2027.
