NVIDIA and Microsoft bring Vera Rubin NVL72 to Azure and Nemotron 3.5 to Foundry

NVIDIA Blog · rss · 2026-11-20

At Microsoft Ignite, NVIDIA and Microsoft expanded their full-stack collaboration across Azure, enterprise AI, and Windows PCs. Vera Rubin NVL72 systems are now running in Azure data centers, with confidential computing extending to bare-metal rack-scale infrastructure; Azure scaled to 10,240 GB200 GPUs on DeepSeek-V3 in MLPerf Training v6.1 and led cloud DeepSeek-R1 interactive throughput by over 10% on GB300 NVL72. Microsoft Foundry adds NVIDIA open models: EY used Programmed Training to adapt Nemotron models for purchase-record classification, improving accuracy 55% at 15x lower cost than proprietary frontier models. Nemotron 3.5 Super and Lightning are live, with Bolt and Transcribe coming. Edge options include Nemotron 3.5 in Foundry Local, RTX PRO 6000 Blackwell GPUs and Jetson Thor/Orin support in Azure Local, and local LLM demos on RTX Spark PCs. Microsoft AI also uses NVIDIA Dynamo for RL rollout inference during post-training, and BioNeMo/ALCHEMI toolkits join Microsoft Discovery with the RFD4-Proteina protein-generation model.

Original post →

More from Embodied

Embodied channel →