Meta has signed a sweeping multiyear, multigenerational agreement with Nvidia to deploy millions of Blackwell and Rubin GPUs across its expanding data center network, according to an announcement made Monday. The deal also includes Nvidia's Grace CPUs and Spectrum-X Ethernet networking technology — and marks the first large-scale standalone deployment of Grace CPUs by any company.
Chip analyst Ben Bajarin of Creative Strategies described the financial scope plainly: "The deal is certainly in the tens of billions of dollars." Nvidia declined to specify a figure but confirmed Meta will deploy the chips across its own data centers and through Nvidia Cloud Partners.
The agreement accelerates Meta's already-massive infrastructure ambitions. The company has committed to spending $600 billion on U.S. data centers through 2028, with $135 billion earmarked for AI infrastructure this year alone. Two landmark projects are currently under construction: the Prometheus 1-gigawatt facility in New Albany, Ohio, and the 5-gigawatt Hyperion site in Richland Parish, Louisiana — one of the largest data centers ever built.
Beyond raw compute, the deal extends to privacy. Meta will adopt Nvidia's Confidential Computing framework to power WhatsApp's AI features while keeping user messages private, a rare case of AI capability and end-to-end encryption being engineered side by side.