NVIDIA details expanded Microsoft AI collaboration
NVIDIA says its expanded collaboration with Microsoft spans Azure AI infrastructure, model development and industrial systems. The companies are also working on confidential computing and bringing NVIDIA models and GPUs into Microsoft products and services.

Key takeaways · 4
- 01
If assessing Azure infrastructure for demanding AI workloads, note that Microsoft is expanding access to Vera Rubin NVL72-powered virtual machines.
- 02
Foundry’s serverless Programmed Training lets developers control post-training workflows while Microsoft Foundry manages the infrastructure.
- 03
EY used Programmed Training to adapt NVIDIA Nemotron open models for corporate purchasing records; NVIDIA says the work improved evaluation accuracy by 55% and reduced costs 15-fold compared with proprietary frontier models.
- 04
For local deployments, expanded Nemotron 3.5 support in Foundry Local brings reasoning, coding and tool-using agents to Azure Local environments.
Azure infrastructure and security
NVIDIA says Vera Rubin NVL72 systems are running in Azure data centers, while Microsoft is expanding access to Azure virtual machines powered by them.[1] The infrastructure is designed for demanding reasoning, agentic applications, and large-scale model training and inference.[1] NVIDIA and Microsoft are also working to extend confidential computing to bare-metal and rack-scale AI infrastructure.[1] NVIDIA says this is intended to protect sensitive models and data while they are processed, and that its confidential-computing and attestation tools let customers verify the environment before releasing sensitive assets.[1]
Foundry training and Nemotron models
Microsoft Foundry’s serverless Programmed Training gives developers control over post-training workflows while Foundry manages the infrastructure; Foundry Managed Compute adds support for deploying custom model weights.[1] NVIDIA says EY used Programmed Training to adapt Nemotron open models for classifying corporate purchasing records.[1] NVIDIA reports a 55% accuracy improvement on EY’s evaluation dataset and costs 15 times lower than with proprietary frontier models.[1] The model lineup includes Nemotron 3.5 Super for long-running agents, Lightning for fast responses and tool use, and Parse and Content Safety for document extraction and harmful-content detection.[1]
Local AI and industrial systems
Expanded Nemotron 3.5 support in Foundry Local brings reasoning, coding and tool-using agents to Azure Local environments.[1] Azure Local will support NVIDIA RTX PRO 6000 Blackwell GPUs from hardware partners, and is expanding to Jetson Thor and Jetson Orin for industrial uses including machine vision, robotics and autonomous operations.[1] NVIDIA says its Cosmos physical AI models are also being brought to Foundry Local.[1] NVIDIA and Microsoft will showcase local large-language-model experiences at Ignite; the described capabilities support up to two 4K streams at 60 frames per second, even over challenging networks.[1]
Research, inference and previews
Models for life sciences, drug discovery, and materials and chemical science are being brought into Microsoft’s agentic research platform.[1] NVIDIA says its BioNeMo Agent Toolkit lets scientists use natural language to run computational biology workflows in Microsoft Discovery.[1] Microsoft AI uses NVIDIA Dynamo for reinforcement-learning rollout inference during post-training; NVIDIA says Dynamo coordinates inference across a large GPU fleet and helps manage requests, model updates and failure recovery.[1] NVIDIA RTX PRO 6000 Blackwell GPUs are also for Bing search models.[1] A limited preview of Direct Stream is open for sign-ups, with a public preview planned for early 2027.[1]
For professionals, the announcement connects infrastructure choices with model customization, deployment location and workload-specific tools. Teams can use those details to identify which parts of the collaboration merit evaluation for their own needs, while treating NVIDIA’s reported performance results as company-provided claims.
Why it matters
AI training for your team
kju.ai is built for teams: daily sessions, team analytics, and rollout support for your organization.
How this developed
20 November 2026
NVIDIA details expanded Microsoft AI collaboration