NVIDIA Vera Rubin Enters Volume Production, Promising 10x Lower Inference Cost Than Blackwell

Summary

NVIDIA announced at GTC Taipei / COMPUTEX 2026 that its next-generation Vera Rubin platform is entering volume production, with more than 350 supply-chain partners across 30 countries — including Dell, HPE, Supermicro, and Lenovo — building systems for delivery later in 2026. The Vera Rubin NVL72 rack-scale system pairs 72 Rubin GPUs with 36 Vera CPUs over sixth-generation NVLink, and NVIDIA claims it delivers roughly 10x higher inference throughput at one-tenth the cost per token versus the cur