Nvidia has announced a massive multiyear agreement to supply Meta Platforms with millions of its high-performance artificial intelligence chips. The deal covers a broad spectrum of Nvidia’s hardware, including the current Blackwell architecture and the forthcoming Rubin AI chips. Crucially, the agreement also includes Nvidia’s Grace and Vera central processing units (CPUs), signalling a direct challenge to the data centre dominance of Intel and AMD.
Read: Starlink escalates public campaign for South African operating licence
While Nvidia is the undisputed leader in AI GPUs, this announcement highlights its aggressive expansion into the CPU market. Based on Arm technology, Nvidia’s Grace and Vera processors are designed to handle “workaday” technical tasks—such as running massive databases—while serving as efficient companions to AI workloads.
By selling standalone installations of these CPUs to Meta, Nvidia is proving that its silicon is capable of more than just generative AI. Ian Buck, Nvidia’s GM of hyperscale computing, noted that the Grace CPU has already demonstrated the ability to complete common tasks using half the power of traditional processors. Meta has reportedly already begun testing the next-generation Vera chips, with Buck describing the initial performance results for high-intensity back-end operations as “very promising.”
See also
The timing of this deal is a significant strategic move for Nvidia. Meta is currently a “hyperscaler” powerhouse, believed to be one of the four key customers responsible for a staggering 61% of Nvidia’s recent quarterly revenue. However, Meta is also hedging its bets by developing its own in-house AI silicon and discussing the use of Google’s Tensor Processing Units (TPUs).
As data centres face increasing scrutiny over energy consumption, Nvidia is positioning its CPU line as the “green” choice for the AI era. The Grace and Vera chips are engineered to maximize “performance-per-watt,” which is critical for companies like Meta that operate at a global scale. If Nvidia can convince the industry that its CPUs are the most efficient choice for running AI agents and background databases, it could fundamentally shift the balance of power in the broader semiconductor industry.
