NVIDIA and Microsoft extend AI partnership across Azure, Foundry and Windows devices
Announced at Microsoft Ignite, the collaboration covers Vera Rubin systems in Azure, open-model customization in Foundry, edge deployment options and confidential computing work.
In plain words
NVIDIA and Microsoft say they are widening their joint AI work so companies can build and run AI applications on Azure, on their own premises and on Windows PCs. The announcement covers new Azure virtual machines using NVIDIA's Vera Rubin systems, NVIDIA open models in Microsoft Foundry, and tools for running AI closer to where business data lives.
What was announced
The companies said Vera Rubin NVL72 systems are now running in Azure data centers for reasoning, agentic applications and large-scale training and inference. They are also working to extend confidential computing to bare-metal and rack-scale infrastructure, so customers can verify the computing environment before releasing sensitive models, data or keys. Microsoft cited benchmark results as context: in MLPerf Training v6.1, Azure scaled to 10,240 NVIDIA GB200 GPUs on the DeepSeek-V3 model, and its GB300 NVL72 systems led cloud submissions on DeepSeek-R1 interactive throughput by over 10%, per Signal65.
On the software side, Microsoft Foundry is adding NVIDIA Nemotron 3.5 open models and customization options, including a serverless Programmed Training offering and Managed Compute for deploying custom weights. EY used Programmed Training to adapt Nemotron models for classifying purchasing records, reporting a 55% accuracy improvement on its evaluation dataset and a 15x cost reduction versus proprietary frontier models. Beyond the cloud, Foundry Local gains Nemotron 3.5 support in Azure Local environments, which will also support RTX PRO 6000 Blackwell GPUs and Jetson Thor and Orin platforms for industrial uses. Microsoft is additionally using NVIDIA Dynamo inference software in its own post-training workflows.
What it does not show
- All claims come from a single NVIDIA blog post announcing the partnership; no independent testing or confirmation is reflected, and benchmark figures are vendor-reported or attributed to Signal65 within that post.
- Some items are previews or future plans, such as the Windows 365 Direct Stream limited preview with public availability planned for early 2027, and Nemotron 3.5 Bolt and Transcribe models described as coming soon; current availability is not established for those.
For security and IT teams
- Scope
- The announcement spans cloud infrastructure, on-premises and edge deployment through Azure Local, and local AI on RTX Spark Windows PCs, rather than a single product launch.
Sources
Source-support review completed 20 Nov 2026. This story was drafted with AI assistance and checked against retained source passages. Automated checking is not independent verification.
Updates and corrections
No corrections so far.
First published 20 Nov 2026
New to the subject? AI from the Ground Up starts from the beginning.