SAN FRANCISCO — Nvidia on Tuesday announced a multiyear agreement to supply Meta Platforms with millions of its artificial intelligence (AI) chips, including both current and future models, as well as standalone central processing units (CPUs) that compete with products from Intel and Advanced Micro Devices.
Read More: NVIDIA tops 2025 tech giants by market value
The deal, whose value was not disclosed, covers Nvidia’s current Blackwell chips and the forthcoming Rubin AI chips. It also includes installations of Nvidia’s Grace and Vera CPUs, which were introduced in 2023 based on technology from Arm Holdings. Nvidia aims to position these CPUs for high-intensity workloads such as AI agent operations, data centers, and technical tasks including database management.
Ian Buck, general manager of Nvidia’s hyperscale and high-performance computing unit, highlighted the efficiency of the Grace CPUs, stating they can use half the power for common tasks like database operations. He added that early testing of the Vera CPU for Meta has shown promising results for next-generation performance.
The announcement comes as Meta develops its own AI chips and explores partnerships with Google to use its Tensor Processing Units (TPUs) for AI workloads. Analysts believe the deal underscores Nvidia’s continued influence in the AI chip market and its strong business relationship with Meta.
Read More: Nvidia outlines bold AI vision at CES 2026
Nvidia has not publicly disclosed the size of its sales to Meta, but the company is widely believed to be among four clients accounting for 61% of Nvidia’s revenue in its most recent fiscal quarter. The deal signals Nvidia’s growing traction with its CPU offerings alongside its established AI chip business.