The Brief: At NVIDIA GTC, Microsoft presented updates across Microsoft Foundry, Azure AI infrastructure, and Physical AI integrations in collaboration with NVIDIA.
Foundry capabilities were extended to enable production-scale AI agents, highlighted by the general availability of Foundry Agent Service and enhanced observability features. Plus, the addition of NVIDIA Nemotron models and Fireworks AI integrations supports model customization and deployment across environments.
On the infrastructure front, Microsoft outlined Azure’s optimization for inference-driven workloads, including early rollout of NVIDIA Vera Rubin NVL72 systems and increased use of liquid-cooled GPUs.
The company also highlighted Physical AI advancements through deeper integration of Microsoft Fabric with NVIDIA Omniverse to support simulation-driven operations.
Analyst Perspective: Microsoft’s updates show a stronger focus on managing AI systems from start to finish. Instead of just helping build AI, Microsoft is also making it easier to monitor, deploy, and improve these systems in one place. This emphasizes long-term AI support, extending beyond the development phase.
Moreover, its closer partnership with NVIDIA highlights a focus on improving performance. By making sure hardware and software work well together, Microsoft is aiming to make it easier to turn new AI ideas into real-world applications.
Overall, this suggests a platform designed for continuous use, where AI systems can adjust and improve over time without needing constant manual updates.
Microsoft extended the capabilities of Microsoft Foundry to support enterprise-grade AI agent development and deployment. The Foundry Agent Service is now generally available, so teams can build agents capable of reasoning, planning, and executing tasks across connected systems.
Alongside this, the Foundry Control Plane brings observability features that give users visibility into agent behavior, which supports governance and operational oversight.
Additional updates include integration with NVIDIA Nemotron models that allows organizations to access and fine-tune open-weight models within the Foundry environment. Plus, the introduction of Voice Live API support in preview enables real-time, multimodal agent interactions.
Expanded integrations with security platforms like Prisma AIRS and Zenity further extend coverage across the agent lifecycle, from development to runtime protection.
Microsoft outlined its approach to scaling AI infrastructure to meet the demands of inference-heavy and reasoning-based workloads. Azure is being optimized to support these workloads through purpose-built systems designed for performance, efficiency, and scalability.
The company reported rapid deployment of liquid-cooled NVIDIA Grace Blackwell GPUs across its data centers, supporting high-density AI processing. It also announced early testing of NVIDIA Vera Rubin NVL72 systems, with plans to expand availability in the coming months.
Microsoft is extending these capabilities to regulated and sovereign environments through Azure Local to help organizations maintain control over data and operations. Integration with Azure Arc helps keep management and governance consistent while still supporting compliance and flexibility.
Microsoft and NVIDIA are collaborating to support Physical AI systems that connect digital models with real-world operations. Central to this effort is the NVIDIA Physical AI Data Factory Blueprint, which integrates with Microsoft Foundry to support large-scale deployment.
Microsoft introduced a Physical AI toolchain built on Azure services, enabling developers to create workflows that combine simulation, training, and operational execution. This includes a public GitHub repository designed to streamline development and deployment processes.
Further integration between Microsoft Fabric and NVIDIA Omniverse allows organizations to connect live operational data with simulation environments. This enables real-time analysis and decision-making across physical systems like manufacturing facilities, where AI can coordinate actions across machines and workflows.
Microsoft’s updates show a clear alignment between infrastructure, platforms, and real-world applications. The emphasis on AI systems that can operate reliably, along with production-ready agents, helps businesses shift from experimentation to practical use. This is particularly helpful for companies working in complex or regulated environments that require consistent oversight and control.
At the same time, integrating Physical AI introduces new opportunities for industries that rely on simulation and real-time coordination.
Still, there are challenges to consider, including how to integrate these systems, the need for specialized skills, and ensuring stable performance across different setups. Continued improvements in tools and automation will be key to addressing these issues.
In the long run, Microsoft seems focused on building tightly connected systems that support both development and real-world use, helping businesses adopt AI at a larger scale.
Position your event or publication with perspectives that resonate across the industry. Contact us at the Collab Collective to learn how our experts can support your media coverage or speaking lineup.