Partnerships
NVIDIA Expands Microsoft AI Collaboration with Vera Rubin and Nemotron

Key PointsNVIDIA and Microsoft are extending confidential computing to rack-scale AI…
View more
- NVIDIA and Microsoft are extending confidential computing to rack-scale AI infrastructure so regulated customers can verify environments before releasing sensitive data.
- EY adapted Nemotron models with Programmed Training and reported a 55% accuracy improvement alongside a 15x cost reduction versus proprietary frontier models.
- The partnership now reaches beyond Azure into on-premises systems, industrial edge devices and Windows Cloud PCs through Foundry Local and RTX PRO GPUs.
NVIDIA announced an expanded AI collaboration with Microsoft on November 19, 2026, at Microsoft Ignite, spanning Vera Rubin NVL72 systems running in Azure data centers, broader Nemotron open-model support in Microsoft Foundry, and deployment options for on-premises systems, industrial edge devices and Windows Cloud PCs.
The companies are expanding access to Azure virtual machines powered by Vera Rubin NVL72 through Azure’s Vera Rubin preview. The systems are designed for reasoning and agentic applications as well as large-scale model training and inference, connecting the preview to both model development and deployment.
Beyond compute capacity, Microsoft and NVIDIA are working to extend confidential computing to bare-metal and rack-scale AI infrastructure for proprietary and regulated information. NVIDIA Confidential Computing and attestation help customers verify the computing environment before releasing sensitive models, data or keys, with work under way to extend those protections across interconnected AI factory infrastructure. The companies are also collaborating on power efficiency through NVIDIA MaxLPS to help data centers obtain more performance from available power.
Microsoft is using NVIDIA technology in its own operations as well. Microsoft AI uses Dynamo for reinforcement-learning rollout inference during post-training, the computationally intensive process that generates experiences used to improve models. Dynamo coordinates inference across a large GPU fleet, optimizing performance and helping manage requests, model updates and recovery from failures in a deployment built on joint Microsoft AI and NVIDIA engineering. Microsoft also uses RTX PRO 6000 Blackwell GPUs for Bing search models.
Open-Model Customization in Foundry
For enterprise developers, Microsoft’s serverless Programmed Training offering provides control over post-training workflows while Foundry manages the underlying infrastructure. Foundry Managed Compute adds support for deploying custom model weights, connecting model adaptation with deployment through Microsoft’s managed services.
NVIDIA reported that EY used Programmed Training to adapt Nemotron open models for classifying corporate purchasing records. The reported results were a 55% accuracy improvement on EY’s evaluation dataset and a 15x cost reduction compared with proprietary frontier models, providing a measured example of customization for a specialized business task.
The expanded Nemotron 3.5 lineup on Foundry Managed Compute includes Super for long-running agents working across coding, research and business workflows, and Lightning for applications requiring fast responses and tool use. Embed and Rerank support information retrieval, Parse extracts structured information from documents, and Content Safety detects harmful content. NVIDIA listed Bolt for high-throughput, low-latency reasoning and Transcribe for speech recognition as coming soon.
Deployment Beyond Azure
Expanded Nemotron support in Foundry Local brings reasoning, coding and tool-using agents to Azure Local environments, while retrieval models help applications find evidence across text, document images, charts and tables. NVIDIA said Azure Local will support RTX PRO 6000 Blackwell GPUs from hardware partners for on-premises AI deployments.
Azure Local is also expanding to Jetson Thor and Jetson Orin for industrial uses including machine vision, robotics and autonomous operations, with Cosmos physical AI models being brought to Foundry Local. These additions extend the collaboration from centralized cloud infrastructure toward systems located closer to the data generated by industrial operations.
NVIDIA said the companies will demonstrate local large-language-model experiences on RTX Spark-powered Windows PCs at Ignite. For GPU-accelerated Windows 365 Cloud PCs, Direct Stream supports up to two 4K streams at 60 frames per second with low latency over challenging networks. A limited preview is open for sign-ups, with public preview planned for early 2027.
Scientific Tools and Enterprise Implications
NVIDIA’s BioNeMo Agent Toolkit and ALCHEMI Toolkit are also expanding into the Microsoft Discovery catalog, adding models for life sciences, drug discovery, materials science and chemical science to Microsoft’s agentic research platform. BioNeMo Agent Toolkit lets researchers use natural language to run computational-biology workflows, with agents coordinating the necessary tools and model calls.
The collaboration consequently covers both the supply and use of AI computing: Microsoft is running NVIDIA technology inside its own model development and search operations while offering related infrastructure and model services to enterprises. The announced rollout is staged across systems already running in Azure, a limited Direct Stream preview open for sign-ups and additional models still to come.












