Nvidia and Microsoft Expand Infrastructure Collaboration from Azure to Windows PCs
Nvidia and Microsoft have broadened their technical alliance at Microsoft Ignite, introducing Vera Rubin systems on Azure, expanded open model tools, and edge computing support.
By The Daily Circuit AI Desk (automated, AI-generated)
3 min read
AI-generated image (flux-1-schnell), not a photograph of the events described.
Key facts
- Microsoft is making Azure virtual machines powered by Nvidia Vera Rubin NVL72 systems available to customers.[1]
- EY used Programmed Training in Microsoft Foundry to adapt Nvidia Nemotron open models, improving accuracy by 55% and lowering costs by 15x.[1]
- In MLPerf Training v6.1, Azure scaled to 10,240 Nvidia GPUs with GB200 NVL72 systems on the DeepSeek-V3 model.[1]
- Microsoft is incorporating Nvidia RTX PRO 6000 Blackwell GPUs to accelerate Bing search models.[1]
- Direct Stream capabilities for Windows 365 Cloud PCs will enter public preview in early 2027.[1]
Nvidia and Microsoft are broadening their collaborative efforts to integrate full-stack technologies across Azure infrastructure, industrial environments, and Windows PCs [1]. The expanded partnership was announced at Microsoft Ignite, introducing new integrations spanning cloud data centers, developer tooling, and edge devices [1].
Expanding Cloud and Infrastructure Capabilities
Under the expanded collaboration, Microsoft is making Azure virtual machines powered by Nvidia Vera Rubin NVL72 systems available to customers [1]. These data center systems are engineered to handle demanding reasoning workloads, agentic applications, and large-scale model training and inference [1]. Furthermore, the companies are collaborating on power efficiency using Nvidia MaxLPS to help data centers optimize available power [1].
Performance benchmarks highlighted the underlying infrastructure integration. In MLPerf Training v6.1, Azure scaled to 10,240 Nvidia GPUs using GB200 NVL72 systems on the DeepSeek-V3 model, marking the largest scale in the testing round [1]. Additionally, in MLPerf Inference v6.1, Azure systems running Nvidia GB300 NVL72 led cloud submissions on DeepSeek-R1 interactive throughput by over 10%, according to Signal65 [1].
To safeguard sensitive proprietary and regulated datasets, Microsoft and Nvidia are extending confidential computing protections to bare-metal and rack-scale AI infrastructure [1]. Nvidia Confidential Computing and attestation mechanisms enable customers to verify computing environments before releasing sensitive models, data, or cryptographic keys [1].
Enterprise Tools and Custom Open Models
Developers gain broader access to Nvidia open models and customization capabilities through Microsoft Foundry [1]. Microsoft’s serverless Programmed Training service provides developers with governance over post-training workflows while managing underlying infrastructure [1]. Foundry Managed Compute introduces support for deploying custom model weights directly through managed services [1].
Enterprise deployment examples include EY, which utilized Programmed Training within Microsoft Foundry to adapt Nvidia Nemotron open models for classifying corporate purchasing records [1]. This implementation improved evaluation dataset accuracy by 55% and reduced costs by 15 times compared to proprietary frontier models [1].
The extended Nvidia Nemotron 3.5 lineup on Foundry Managed Compute features specialized variants [1]. Nemotron 3.5 Super caters to long-running coding, research, and business agents, while the Lightning model suits applications requiring rapid responses and tool invocation [1]. Complementary models include Nemotron Embed and Rerank for information retrieval, Nemotron Parse for document structure extraction, and Nemotron Content Safety for harmful content detection, with Nemotron 3.5 Bolt and Transcribe slated for release [1].
Edge, Local, and Scientific Computing
Deployment options are extending beyond cloud environments into local and edge settings. Foundry Local brings Nemotron 3.5 capabilities and physical AI models like Cosmos to Azure Local environments [1]. Hardware options for on-premises deployments are also expanding to include Nvidia RTX PRO 6000 Blackwell GPUs from hardware partners, alongside support for Nvidia Jetson Thor and Jetson Orin edge platforms in industrial machine vision and robotics [1].
On the client side, Nvidia and Microsoft are showcasing local large language model experiences on Windows PCs powered by Nvidia RTX Spark [1]. For Cloud PC users, Direct Stream capabilities are being brought to GPU-accelerated Windows 365 Cloud PCs, supporting graphics applications with up to two 4K streams at 60 frames per second over challenging networks, with a limited preview open for sign-ups and public preview scheduled for early 2027 [1].
In scientific research, the Nvidia BioNemo Agent Toolkit and Nvidia ALCHEMI Toolkit are entering the Microsoft Discovery catalog [1]. An initial integration utilizes the RFD4-Proteina model, developed with the University of Washington’s Institute for Protein Design, to generate candidate protein structures and amino acid sequences [1].
Internal Acceleration at Microsoft
Alongside customer-facing integrations, Microsoft is deploying Nvidia technology within its own internal services [1]. Microsoft AI utilizes Nvidia Dynamo to coordinate inference across large GPU fleets during reinforcement learning rollout inference, optimizing performance, handling requests, and managing model updates and recovery [1]. Furthermore, Microsoft uses Nvidia RTX PRO 6000 Blackwell GPUs to accelerate Bing search models [1].
Context
The collaboration builds upon long-standing technical integration between Nvidia and Microsoft, combining Nvidia's accelerated computing hardware and software stacks with Microsoft's cloud, developer tools, and operating system ecosystems.
Why it matters
The expanded alliance shapes how enterprise developers, scientific researchers, and cloud providers build, secure, and deploy artificial intelligence applications across diverse computing environments ranging from hyperscale cloud infrastructure to local edge devices.
What could happen next
AI-generated analysis based on the sources below. Not a forecast or recommendation.
The integration of Nvidia's hardware and open models directly into Microsoft Foundry and Azure Local could accelerate enterprise adoption of customized AI agents by simplifying infrastructure management and data security compliance.
Risks and open questions
Performance at scale, seamless cross-platform confidential computing attestation across complex multi-accelerator topologies, and the practical reliability of edge deployments remain operational variables for enterprise adopters.
Sources
- 1.NVIDIA (primary source): NVIDIA and Microsoft Expand Accelerated Computing From Azure to Windows PCs, 20 November 2026
How this article was produced. This story was researched and written by an automated system using Google’s Gemini models, working only from the sources listed above. Automated checks verified that every citation, quote and figure matches those sources before publication. It was not written by a human reporter. Editorial policy · Report an error
- NVIDIA
- Microsoft
- Azure
- AI
- Cloud Computing
Related stories
Jensen Huang Resigns From Nvidia Following Extramarital Affairs And National Security Scrutiny
Nvidia co-founder Jensen Huang has resigned as CEO following revelations of extramarital affairs and federal scrutiny over chip sales to China.
3 min read
Signal Blocks Screenshots on Mobile Devices to Prevent AI Agents From Reading User Data
Encrypted messaging platform Signal is restricting screenshots on Android and iOS devices to stop AI assistants like Gemini and Siri from accessing user messages.
2 min read
Anthropic Raises Billions in Record-Breaking Initial Public Offering
AI lab Anthropic has gone public in a massive initial public offering, raising nearly $105 billion while its seven founders each cashed out about $280 million.
2 min read
Anthropic Prices IPO at $83 Per Share, Valuing Claude Maker at $1.5 Trillion
Anthropic’s initial public offering is priced above its projected range, raising $81.3 billion and valuing the AI company at roughly $1.5 trillion on a fully diluted basis.
3 min read