NVIDIA Accelerates AI Hardware: 2.5M Grace Chips Shipped, Vera Rubin Targets 1,000 Racks Daily
NVIDIA has accelerated its CPU roadmap, shipping over 2.5 million Arm-based Grace CPUs to date, and is targeting a daily production capacity of 1,000 Vera Rubin racks. This production push is supported by approximately 12 manufacturing partners to meet the growing demand for AI infrastructure. This massive scaling solidifies NVIDIA's position as a full-stack system provider rather than just a GPU designer. The aggressive production target for the Vera Rubin platform, designed for Agentic AI, represents a massive revenue potential estimated at up to $630 billion per quarter. The Vera Rubin rack serves as the foundational building block for "AI factories," integrating the next-generation Vera CPU and Rubin GPUs. NVIDIA is collaborating with over 300 global partners to ensure a stable supply of the new Vera chips.
## BACKGROUND
Grace is NVIDIA's first proprietary data center CPU built on the Arm architecture, designed to handle high-performance computing and AI workloads. The upcoming Vera Rubin platform succeeds the Blackwell architecture, optimized specifically for Agentic AI—systems that can autonomously plan, use tools, and execute multi-step tasks to achieve goals.