NVIDIA and Microsoft Expand Full-Stack AI Collaboration Across Azure, Windows PCs, and Industrial Edge

·2026.11.20 05:00

Key point

The partnership adds Vera Rubin NVL72 systems to Azure, expands Nemotron model customization via Microsoft Foundry, and brings local AI capabilities to Windows PCs and industrial edge devices.

Details

  • NVIDIA Vera Rubin NVL72 systems are now running in Azure data centers, supporting confidential computing for bare-metal and rack-scale AI infrastructure to protect sensitive models and data.
  • Azure’s NVIDIA GB300 NVL72 systems led cloud submissions in MLPerf Inference v6.1 for DeepSeek-R1 interactive throughput by over 10%, while GB200 NVL72 systems scaled to 10,240 GPUs in MLPerf Training v6.1.
  • Microsoft Foundry expands support for NVIDIA Nemotron 3.5 models, enabling enterprises to customize open models; EY used Programmed Training to improve accuracy on its evaluation dataset for classifying corporate purchasing records by 55% and lower costs by 15x compared to proprietary frontier models.
  • Local AI deployment expands to Azure Local with Nemotron 3.5 support, NVIDIA RTX PRO 6000 Blackwell GPUs, and Jetson Thor/Orin platforms for industrial applications, while Windows PCs powered by NVIDIA RTX Spark will showcase local LLM experiences.

Scientific and Internal Acceleration

  • NVIDIA BioNemo Agent Toolkit and ALCHEMI Toolkit integrate into Microsoft Discovery, using the RFD4-Proteina model to generate candidate protein structures for life sciences research.
  • Microsoft AI utilizes NVIDIA Dynamo inference software to support reinforcement learning rollout inference during post-training, coordinating inference across large GPU fleets for performance optimization and failure recovery.

This summary was generated automatically by AI. Check the original for the author's claims and context. Copyright belongs to the original author.

Our guide explains how the AI works. Report summary errors, attribution issues, or removal requests via Contact.