Microsoft at NVIDIA GTC: New solutions for Microsoft Foundry, Azure AI infrastructure and Physical AI
Microsoft is enhancing artificial intelligence by merging accelerated computing with large-scale cloud engineering. For many years, we’ve partnered with NVIDIA to fuse hardware, software, and infrastructure, enabling significant advancements in AI technology.
Exciting Updates from NVIDIA GTC
- We’ve broadened the capabilities of Microsoft Foundry, allowing users to create, deploy, and manage AI agents ready for production on NVIDIA accelerators and open NVIDIA Nemotron models.
- A new Azure AI infrastructure, fine-tuned for inference-heavy, reasoning-centric tasks, marks the launch of the first hyperscale cloud using next-generation NVIDIA Vera Rubin NVL72 systems.
- Enhanced integration among Microsoft Foundry, Microsoft Fabric, and NVIDIA Omniverse now supports Physical AI systems, bridging simulation and real-world operations.
Transforming Frontier Models into Ready-to-Use Agents
At the core of our strategy lies Microsoft Foundry, serving as the operating system for AI development, deployment, and management at an enterprise scale. Foundry harnesses Azure to combine models, tools, data, and observability within one cohesive system designed for production agents. We’re now expanding these capabilities for Foundry Agent Service and NVIDIA Nemotron models.
The next-generation Foundry Agent Service and Observability in Foundry Control Plane are now generally available. This means organizations can develop and manage AI agents efficiently at scale. The Foundry Agent Service enables teams to create agents that can reason, plan, and take action across tools, data, and workflows. Once established, Foundry Control Plane grants developers comprehensive oversight into agent performance, boosting productivity and fostering trust at the enterprise level. Companies like Corvus Energy are already using Foundry to transition from manual inspections to agent-driven operational intelligence throughout their global fleet.
We’ve also made the journey from prototype to production smoother with the public preview of Voice Live API integration with Foundry Agent Service. This allows developers to create voice-first, multimodal, real-time experiences. Additionally, a revamped Microsoft Foundry portal with expanded integrations for Palo Alto Networks’ Prisma AIRS and Zenity enhances builder experiences and runtime security throughout the agent lifecycle.
NVIDIA Nemotron models are now accessible through Microsoft Foundry, joining the most extensive array of models available on any cloud, including the latest reasoning, frontier, and open models. This complements our recent partnership announcement about Fireworks AI on Microsoft Foundry, allowing customers to refine open-weight models like NVIDIA Nemotron into low-latency resources that can be deployed at the edge.
Scaling AI Infrastructure for High-Demand Workloads
AI inference workloads are transforming the cost, performance, and design principles of systems. To effectively implement agentic AI at scale, businesses require specially designed infrastructure that focuses on inference-heavy, reasoning-centric tasks, consistently operable across global and regulated landscapes.
Microsoft’s approach to AI infrastructure has been meticulously crafted to integrate cutting-edge NVIDIA systems into Azure data centres, designed for efficiency in power, cooling, networking, and quick upgrades. This grants our customers swift and agile operation, as they stay ahead with each new generation.
In less than a year, we’ve rolled out hundreds of thousands of liquid-cooled Grace Blackwell GPUs across our global data centre network. We’re also thrilled to be the first hyperscale cloud to activate NVIDIA’s latest Vera Rubin NVL72 in our labs. Over the coming months, we’ll be deploying Vera Rubin NVL72 across our state-of-the-art, liquid-cooled Azure data centres.
Our infrastructure breakthroughs with NVIDIA are also available in sovereign and regulated environments, allowing customers to control how and where their AI operates. Recently, we introduced Foundry Local support for modern infrastructure and expanded AI models, and today we’re launching initial support for the NVIDIA Vera Rubin platform on Azure Local. This expands accelerated AI capabilities into customer-managed environments while maintaining Azure-consistent governance and security through our unified software layer with Azure Arc and Foundry Local.
Integrating AI into the Physical World
As AI expands beyond digital realms, Microsoft and NVIDIA are joining forces to advance the next generation of Physical AI. At GTC, this initiative focuses on the NVIDIA Physical AI Data Factory Blueprint, with Microsoft Foundry serving as the platform for hosting and managing Physical AI systems at cloud scale.
By merging this blueprint with Azure services into a Physical AI Toolchain, Microsoft empowers developers to craft, train, and run physical AI and robotics workflows. These workflows connect physical assets, simulations, and cloud training into efficient, enterprise-grade pipelines. We are launching a public Azure Physical AI Toolchain GitHub repository, which is integrated with the NVIDIA Physical AI Data Factory and key Azure services.
To enhance the impact of AI in real-world physical scenarios, Microsoft and NVIDIA are deepening their collaboration between Microsoft Fabric and NVIDIA Omniverse libraries. This integration links live operational data to accurately simulate digital twins, allowing organizations to monitor their physical systems in real time and use AI to determine the next best actions. Businesses in manufacturing and operations are adopting this strategy to move beyond mere monitoring to coordinated, AI-driven responses across machinery, facilities, and workflows.
From Innovation to Real-World Results
Microsoft is committed to delivering reliable, large-scale AI by uniting its robust global AI infrastructure, platforms, and real-world systems with cutting-edge NVIDIA innovations. For our customers, this means the ability to continuously operate intelligent systems while handling inference-heavy, reasoning-focused, and physical AI workloads with the necessary performance, security, and governance suited for commercial and regulated contexts.
From enabling always-on agents to scaling advanced AI infrastructure and deploying smart systems in factories, energy facilities, and controlled environments, Microsoft and NVIDIA are helping customers accelerate their transition from insights to actions.
Yina Arenas leads product strategy and execution for Microsoft Foundry, overseeing the entire AI product portfolio, infrastructure, developer experiences, and model integration across OpenAI, Anthropic, Mistral, DeepSeek, and others. She is dedicated to providing an enterprise-ready, production-grade AI platform that global clients trust for secure, reliable, and scalable AI.
Share this content:
