According to NVIDIA, on July 21, the Vera Rubin NVL72 platform entered production ramp-up, with supply chain coverage now spanning more than 350 factories across 30 countries and over 300 partners participating in deployment.
The company disclosed that Vera CPU was delivered to OpenAI, Anthropic, and SpaceX in June. In production environment testing via DeepInfra, Vera CPU demonstrated up to 1.6x more parallel AI agents and 2.2x task orchestration speed. CoreWeave reported 10x improvement in tokens generated per megawatt-second when running DeepSeek-R1 on Vera Rubin NVL72 compared to the previous Grace Blackwell NVL72. Google Cloud has launched the A5X cloud instance based on Vera Rubin NVL72.