Nvidia Launches Vera CPU and Rubin Platform
Analysis based on 7 articles · First reported Mar 16, 2026 · Last updated Mar 17, 2026
Nvidia's launch of the Vera CPU and Vera Rubin platform significantly escalates competition in the data center CPU market, directly challenging Intel and AMD. This integrated CPU-GPU approach is expected to drive performance and efficiency gains for agentic AI workloads, potentially shifting market share and influencing future data center purchasing decisions.
Nvidia has launched its Vera CPU, an 88-core, Arm-based processor optimized for agentic AI and reinforcement learning workloads, at GTC 2026. This marks Nvidia's full-scale entry into the general-purpose CPU market, directly competing with Intel and AMD. The Vera CPU features custom-designed Olympus cores, high memory bandwidth, and FP8 precision support. It is a key component of the new Vera Rubin platform, a rack-scale AI supercomputer that pairs 72 Rubin GPUs with 36 Vera CPUs. This integrated system approach, leveraging NVLink 6 and other Nvidia technologies, aims to optimize the entire data path between CPU and GPU. Major cloud providers like Amazon===Amazon Web Services, Alphabet===Google Cloud Platform, Microsoft===Microsoft Azure, and Oracle Corporation===Oracle Cloud, along with AI companies such as OpenAI, Anthropic, and Mistral AI, have committed to offering Vera Rubin-based systems. Meta Platforms will also deploy Nvidia CPU-only systems. Early benchmarks from Redpanda show significant performance advantages for Vera over competing CPUs. Commercial systems are expected in the second half of 2026.
Set up alerts, explore entity relationships, search across thousands of events, and build custom intelligence feeds.
Open Dashboard