NVIDIA and Microsoft Expand AI Superfactory Partnership

NVIDIA and Microsoft are deepening their longstanding partnership with a major expansion of AI infrastructure capabilities announced at Microsoft Ignite this week. The collaboration centers on Microsoft's new Fairwater AI superfactory, which connects data centers in Wisconsin and Georgia using NVIDIA's latest Spectrum-X Ethernet networking technology.

The massive-scale infrastructure will integrate hundreds of thousands of NVIDIA Blackwell GPUs for AI training workloads, while Microsoft plans to deploy over 100,000 Blackwell Ultra GPUs globally for inference tasks. This represents one of the largest concentrated deployments of AI computing power to date, designed to handle the most demanding workloads from OpenAI, Microsoft's AI Superintelligence Team, and enterprise customers.

New Azure NCv6 Series virtual machines featuring NVIDIA RTX PRO 6000 Blackwell GPUs are now in public preview, expanding the Blackwell platform's reach across multiple workloads. The infrastructure supports everything from multimodal agentic AI and industrial digitalization to scientific simulation and visual computing, with flexibility extending from cloud to edge deployments through Azure Local.

Beyond raw computing power, the partnership focuses on building what NVIDIA calls a "fungible fleet" - flexible, continuously modernized infrastructure that can accelerate any workload efficiently. Continuous software optimizations have reportedly driven significant cost reductions, including over 90% price drops for popular GPT models on Azure within two years.

The collaboration extends deep into enterprise applications, with NVIDIA accelerating AI in Microsoft SQL Server 2025 through Nemotron open models and NIM microservices. Developers can now build enterprise-ready AI agents that integrate directly into Microsoft 365 applications including Outlook, Teams, and Word using the NVIDIA NeMo Agent Toolkit.

Security represents another major focus area, with the companies collaborating on adversarial learning models that claim 160x performance improvements over CPU s for real-time cybersecurity threat detection. The partnership also advances physical AI capabilities through NVIDIA Omniverse libraries on Azure, supporting robotics development and industrial digital transformation.

This expanded collaboration follows another partnership announcement between NVIDIA, Microsoft, and Anthropic earlier today, highlighting the rapid consolidation happening in the AI infrastructure space as major players align their technology stacks.