Meta and NVIDIA seal a multi-year alliance to boost AI infrastructure

Meta will reinforce its data centers with millions of units from NVIDIA's Blackwell and Rubin GPU families, along with Grace and Vera CPUs from the same company. The alliance covers both Meta's own facilities and deployments with NVIDIA cloud partners, allowing the company to manage its AI workloads through a unified architecture.

One of the most notable aspects of the agreement is that it represents the first large-scale deployment based exclusively on NVIDIA Grace processors. These Arm-based chips are designed to deliver improvements in performance per watt, a relevant factor at Meta's scale of operations. Looking toward 2027, both companies plan to expand this collaboration with the future Vera processors.

For the interconnection of its data centers at AI scale, Meta has adopted the NVIDIA Spectrum-X Ethernet-based networking system, designed for large-scale AI environments. This technology aims to reduce latency and improve data transmission efficiency between data center servers.

On the privacy front, Meta has integrated NVIDIA Confidential Computing into WhatsApp. This solution allows AI-powered features to be activated within the messaging platform without exposing user data during processing. Both companies are also working to extend this capability to other applications in Meta's ecosystem.

Beyond hardware, the agreement includes technical collaboration between engineering teams at both companies to jointly optimize the AI models Meta uses across its platforms, which serve billions of users worldwide.

Key points

• Meta and NVIDIA sign a multi-year strategic alliance to modernize Meta's AI infrastructure.
• Meta's data centers will incorporate millions of GPUs from NVIDIA's Blackwell and Rubin families.
• The agreement includes the first large-scale deployment based exclusively on NVIDIA Grace CPUs.
• The future Vera processors, with greater energy efficiency, are planned for incorporation in 2027.
• Meta adopts NVIDIA Spectrum-X networking for the interconnection of its data centers at AI scale.
• NVIDIA Confidential Computing is integrated into WhatsApp to protect user data during AI processing.
• Engineering teams from both companies will work together on optimizing Meta's AI models.