Meta and NVIDIA Expand Partnership for AI Data Center Infrastructure

NVIDIA and Meta have established a long-term partnership to advance Meta’s AI infrastructure, involving large-scale deployment of NVIDIA’s latest hardware and networking technologies. This collaboration is part of Meta’s broader strategy to expand its data center capabilities and support AI workloads.

As part of the agreement announced in February 2026, Meta will use NVIDIA’s Blackwell and Rubin GPUs across its hyperscale data centers to handle AI training and inference. The partnership also includes the adoption of NVIDIA Grace CPUs, with Meta identified as the first major operator to implement these processors as standalone servers at scale.

Meta’s infrastructure roadmap features plans to introduce NVIDIA’s next-generation Vera CPUs in 2027, further increasing the company’s computing capacity. The collaboration also covers the integration of Spectrum-X Ethernet networking to enhance data center connectivity and performance.

In addition to hardware deployments, Meta will incorporate NVIDIA Confidential Computing into WhatsApp. According to available information, this step is intended to enable AI-powered features while maintaining user data confidentiality within the messaging platform.

What the numbers show

• Meta’s projected AI infrastructure capital expenditures for 2026 are between $115 billion and $135 billion
• Meta previously estimated access to 1.3 million GPUs by the end of 2025
• Plans include up to 30 data centers by 2028, with 26 in the United States

Meta’s investment strategy includes up to $600 billion allocated for U.S. data center and infrastructure projects by 2028. The company’s expansion involves building hyperscale data centers, such as the Prometheus site in Ohio with a capacity of 1 gigawatt and the Hyperion site in Louisiana with 5 gigawatts of capacity.

By the end of 2024, Meta estimated it would have purchased 350,000 H100 GPUs from NVIDIA, with the total number of GPUs accessible to the company expected to reach 1.3 million by the end of 2025. These resources are intended to support the increasing computational demands of Meta’s AI initiatives.

Meta’s data center expansion plan outlines the construction of up to 30 facilities by 2028. Of these, 26 are planned for locations within the United States, reflecting the company’s focus on domestic infrastructure growth.

The combination of new hardware, networking, and confidential computing measures is designed to support Meta’s AI services across its platforms. The partnership with NVIDIA provides Meta with access to multiple generations of CPUs and GPUs, supporting ongoing development and scaling of AI applications.

* This article is based on publicly available information at the time of writing.