Timed with the Microsoft Ignite conference running this week, NVIDIA is expanding its collaboration with Microsoft, including through the adoption of next-generation NVIDIA Spectrum-X Ethernet switches for the new Microsoft Fairwater AI superfactory, powered by the NVIDIA Blackwell platform.
The collaboration brings new integrations across Microsoft 365 Copilot, as well as the public preview of next-generation Azure NC Series VMs powered by NVIDIA RTX PRO 6000 Blackwell Server Edition GPUs, NVIDIA Nemotron integrations to accelerate AI for Microsoft SQL Server 2025, capabilities for onboarding AI agents in Microsoft 365 and optimizations for high-performance inference, cybersecurity and physical AI.
Microsoft’s AI Superfactory connects the landmark Fairwater data center in Wisconsin with a new, state-of-the-art facility in Atlanta, Georgia. This massive-scale infrastructure will integrate hundreds of thousands of NVIDIA Blackwell GPUs for large-scale training. In addition, Microsoft is deploying more than 100,000 Blackwell Ultra GPUs in NVIDIA GB300 NVL72 systems being deployed globally for inference.
“Our collaboration with NVIDIA is built on driving innovation across the entire system and full stack, from silicon to services,” said Nidhi Chappell, corporate vice president of product management at Microsoft. “By coupling Microsoft Azure’s unmatched data center scale with NVIDIA’s accelerated computing, we are maximizing AI data center performance and efficiency, which is of paramount importance for our customers leading the new AI era.”
The most demanding workloads for OpenAI, the Microsoft AI Superintelligence Team, Microsoft 365 Copilot and Microsoft Foundry services will be powered by this infrastructure. Customers like Black Forest Labs are also using NVIDIA GB200 NVL72 systems to train next-generation multimodal FLUX models that power visual intelligence.
To connect this massive infrastructure, Microsoft is deploying next-generation NVIDIA Spectrum-X Ethernet switches in its Fairwater AI data center — the largest and most sophisticated AI factories ever built — delivering the performance, scale and efficiency required for OpenAI to run large-scale AI models and applications.
New Azure NCv6 Series VMs with NVIDIA RTX PRO 6000 Blackwell GPUs are now in public preview on Azure, expanding the Blackwell platform to provide right-sized acceleration for multiple workloads including multimodal agentic AI, industrial digitalization with NVIDIA Omniverse libraries, scientific simulation and visual computing. This flexibility extends from the cloud to the edge with Azure Local, enabling powerful sovereign AI solutions while bringing low-latency, real-time AI to wherever data needs to reside.
This allows enterprises to seamlessly develop, deploy and manage AI-powered digital twins and generative AI applications with NVIDIA RTX PRO 6000 Blackwell GPUs from the Azure cloud directly to their factory floors, on-premises data centers or secure edge locations.
Software Optimizations Deliver a Fungible AI Fleet
The NVIDIA platform on Azure, spanning NVIDIA Blackwell and Hopper GPUs, accelerates the latest models from the Microsoft AI Superintelligence Team, including text (MAI-1-preview), real-time voice (MAI-Voice-1) and high-fidelity image generation (MAI-Image-1) — bringing new multimodal experiences across Bing Image Creator and Microsoft Copilot.
Central to NVIDIA’s collaboration with Microsoft is building a fungible fleet — a flexible, continuously modernized infrastructure that can accelerate any workload with maximum efficiency. This is achieved through continuous, full-stack software optimizations that deliver compounding performance gains and maximize throughput across the entire AI lifecycle and across multiple NVIDIA architectures on Azure. The gains also extend to workloads beyond generative AI, including data processing, vector search, databases, digital twins, scientific computing and 3D design.
This co-engineering saves significant costs for customers, making AI projects that were once theoretical now economically viable. For example, the continuous full-stack optimization work has directly contributed to an over 90% drop in the price of popular GPT models for end users on Azure in two years.
Ongoing optimization work now extends to Microsoft Foundry, where the NVIDIA TensorRT-LLM library helps boost throughput, reduce latency and lower costs for a wide range of popular open models.
NVIDIA and Microsoft have also partnered to optimize their fleet for AI workload performance through the NVIDIA DGX Cloud Benchmarking suite. Engineering teams from both companies worked closely together to identify bottlenecks and implement infrastructure tuning, driving performance gains. By achieving 95% of the performance possible using the NVIDIA reference architecture, Microsoft was named an Exemplar Cloud for H100 training.
From Intelligent Data to AI Agents
NVIDIA and Microsoft are integrating AI into the core of the enterprise, unlocking decades of proprietary data stored in one of the world’s most trusted databases.
NVIDIA is accelerating AI in the new Microsoft SQL Server 2025 by integrating it with NVIDIA Nemotron open models and NVIDIA NIM microservices. This solution delivers GPU-optimized, secure and scalable retrieval-augmented generation directly where enterprise data lives, in the cloud or on premises.
Plus, the collaboration extends to the new frontier of agentic AI in the workplace. The NVIDIA NeMo Agent Toolkit now connects with Microsoft Agent 365, enabling developers to build, deploy and onboard compliant, enterprise-ready AI agents directly into the Microsoft 365 app ecosystem, including Outlook, Teams, Word and SharePoint.
To power these new enterprise agents, Microsoft Foundry now offers NVIDIA Nemotron models for digital AI and NVIDIA Cosmos models for physical AI as secure NIM microservices. Developers can use them to build enterprise-grade agentic AI for a vast range of applications that benefit from multimodal intelligence, multilingual reasoning, math, coding and physical AI capabilities.
The collaboration is also tackling cyber threats for enterprises. Microsoft and NVIDIA are collaborating on research for new adversarial learning models, built on the NVIDIA Dynamo-Triton framework and the NVIDIA TensorRT suite of tools, that can help enterprises defend against real-time cybersecurity threats with a 160x performance speedup compared with CPU methods.
Physical AI and Industrial Digitalization
NVIDIA and Microsoft are building the future of physical AI. With NVIDIA Omniverse libraries available on Microsoft Azure, NVIDIA is unlocking end-to-end reindustrialization in the cloud through its developer ecosystem. Developers are transforming industrial workflows, from computer-aided engineering with Synopsys to factory operations with Sight Machine and SymphonyAI.
Robotics developers can tap into the NVIDIA Isaac Sim open-source robotics simulation framework to unlock critical workflows, from synthetic data generation to software-in-the-loop testing for all types of robot embodiments. Hexagon is building its AEON humanoid robot primarily using NVIDIA’s full robotics stack on Azure. Similarly, the robotics platform, Wandelbots NOVA, running on Azure integrates Isaac Sim and Isaac Lab to simplify and speed up simulation to real-world deployment.
In addition, NVIDIA and Microsoft are using a standardized approach for digital engineering to enable seamless OpenUSD interoperability across 3D workflows, making simulation and digital content creation accessible in the cloud.
This expanded collaboration comes on the heels of a partnership announced with Anthropic and Microsoft earlier today. NVIDIA and Anthropic will collaborate on design and engineering to optimize Anthropic models for performance, efficiency and total cost of ownership, as well as optimize future NVIDIA architectures for Anthropic workloads.
Learn more about NVIDIA and Microsoft’s collaboration and sessions at Microsoft Ignite.