Meta Platforms has signed a multiyear agreement with Nvidia to expand its data center infrastructure with millions of Grace and Vera CPUs, along with Blackwell and Rubin GPUs.
Although Meta has used Nvidia hardware for its artificial intelligence products for years, Nvidia said this agreement represents the first large-scale deployment built solely around its Grace CPUs.
The company added that the Grace platform is expected to deliver significant performance per watt gains inside Meta’s data centers. The agreement also includes plans to introduce Nvidia’s next-generation Vera CPUs into Meta facilities in 2027.
Meta is developing its own in-house chips to run AI models. However, the Financial Times reported that the company has faced technical issues and delays in rolling out its custom silicon strategy.
Nvidia, meanwhile, is navigating concerns related to depreciation and chip-backed loans used to fund AI infrastructure expansion. It is also facing increasing competitive pressure from Advanced Micro Devices (AMD) and Google.
CNBC reported that Nvidia shares fell four percent after a November report said Meta was considering using Tensor chips from Google for AI workloads. Later in the year, AMD announced chip supply arrangements with OpenAI and Oracle.
Neither Nvidia nor Meta disclosed the financial terms of the deal. However, total AI spending this year by Meta, Microsoft, Google, and Amazon is estimated to exceed the cost of the entire Apollo space program, which was over $250 billion in today’s USD value.
