Voices
Jensen Huang: Vera Rubin Platform Now in Full Production — 10x Inference Cost Reduction, 2x Supply Chain Scale vs Blackwell
At Computex, Huang confirmed the Vera Rubin platform has entered full production with a supply chain twice as large as Grace Blackwell. The five purpose-built racks operate as one AI supercomputer delivering up to 10x reduction in inference token cost and 4x reduction in GPUs needed to train MoE models vs. Blackwell. AWS, Google Cloud, Microsoft, and OCI will deploy Vera Rubin instances in H2 2026. For builders: if inference costs drop 10x, the economics of running persistent AI agents changes fundamentally.
↳ Follow the thread