Microsoft CEO Satya Nadella shared a data center photo today, showing that the first production units of NVIDIA Vera Rubin have arrived. Nvidia subsequently confirmed that Vera Rubin is now entering full production.
Vera Rubin is the next-generation rack-scale AI computing platform, following Blackwell. A single NVL72 system integrates 72 Rubin GPUs and 36 Vera CPUs. Nvidia says that compared to the GB200 NVL72, it cuts per-million-token inference costs to about one-tenth and reduces the number of GPUs needed to train MoE models to one-quarter.
https://twitter.com/satyanadella/status/2090929770779521310?s=46