NVIDIA Vera Rubin enters full production, delivering 10x agent throughput compared with NVIDIA Grace Blackwell platform

NVIDIA announced that Vera Rubin — its POD-scale AI platform comprising five purpose-built racks operating as one supercomputer — is ramping into full production. Taiwan's top server makers and global supply chain leaders are manufacturing Vera Rubin-based systems at scale, with 150 ecosystem partners in Taiwan alone across 350+ factories and 30 countries.

Vera Rubin unifies NVIDIA Vera Rubin NVL72 systems, NVIDIA Vera CPU, NVIDIA Groq 3 LPX, NVIDIA Vera BlueField-4 STX storage and NVIDIA Spectrum-6 SPX Ethernet racks into a fully integrated system designed for agentic workloads (AI systems that autonomously reason, retrieve information, use tools, and generate responses).

The platform delivers 10x agent throughput at scale compared with the previous-generation NVIDIA Grace Blackwell platform. Top system builders including Dell Technologies, HPE, Lenovo, Supermicro, and others are in full-scale production of Vera Rubin.