NVIDIA and Microsoft Expand Foundry Partnership for Sovereign AI Deployments

Microsoft and NVIDIA have announced expanded integrations between Microsoft Foundry and NVIDIA's AI ecosystem, creating a unified platform for building and deploying agentic and physical AI systems across diverse environments. This collaboration addresses critical enterprise challenges around fragmented AI stacks, operational complexity, and strict data sovereignty requirements that have hindered AI adoption.

The partnership focuses on three key areas: specialized agentic systems powered by NVIDIA's Nemotron models, sovereign and on-premises AI deployments through Foundry Local, and production-grade physical AI workflows built on Azure and NVIDIA platforms.

Specialized AI Agents with Nemotron Models

Microsoft Foundry now provides access to NVIDIA Nemotron models through NVIDIA NIM microservices, giving developers production-ready open-weight reasoning models within a unified platform. The initial catalog includes Nemotron Nano 9B v2, Llama 3.1 Nemotron Nano VL 8B, Llama 3.3 Nemotron Super 49B v1.5, and NVIDIA Nemotron Super 3.

These models are designed for different use cases:

• Nemotron Nano: Optimized for low-latency, cost-efficient targeted agent tasks
• Nemotron Super: Built for deep research and high-accuracy reasoning
• Nemotron Ultra: Planned for large-scale multi-agent enterprise applications requiring maximum reasoning performance

Additional models coming include Nemotron Speech for enterprise-grade voice agents, Nemotron Vision for document intelligence and video understanding, and Nemotron AI Safety models for content moderation and security.

Later this year, Azure will offer Nemotron models through serverless pay-as-you-go APIs, eliminating infrastructure management overhead. The partnership with Fireworks AI also enables bring-your-own model deployments on Azure through Microsoft Foundry.

Sovereign AI with Foundry Local

For organizations with strict sovereignty requirements, Foundry Local extends Microsoft Foundry's capabilities into on-premises datacenters, edge locations, and sovereign private cloud infrastructure. This integration allows organizations to run advanced AI systems powered by NVIDIA GPUs, including the RTX PRO 6000 Blackwell Server Edition, with future support for NVIDIA Rubin platforms.

Azure Local infrastructure and Azure Arc provide unified management across distributed environments, while Azure Kubernetes Service and Foundry Local enable consistent deployment, operation, and scaling of AI models within sovereign boundaries. This approach allows governments and regulated industries to deploy powerful AI capabilities without compromising data sovereignty.

Physical AI and Robotics Workflows

The collaboration also advances physical AI development through an open Azure Physical AI toolchain that integrates NVIDIA's Physical AI Data Factory Blueprint with Azure services including IoT Operations, Microsoft Fabric Real-Time Intelligence, Microsoft Foundry, and GitHub Copilot.

This toolchain automates data curation, augmentation, and evaluation across perception, mobility, imitation learning, and reinforcement learning pipelines. The NVIDIA Metropolis VSS Blueprint accelerates video analytics AI agent development, while Cosmos world foundation models enable synthetic world generation and large-scale physical AI reasoning optimized for accelerated inference on Azure.

The upcoming NVIDIA Alpamayo open model will support advanced reasoning for autonomous driving systems, including data processing, closed-loop simulation, and evaluation workflows.

Unified AI Platform Vision

These announcements represent Microsoft's broader vision for Foundry as a unified AI platform spanning global public cloud, hybrid infrastructure, sovereign public clouds, and sovereign private environments. Through NVIDIA's accelerated computing integration, organizations can build specialized AI systems once and deploy them consistently across their entire infrastructure footprint.

Microsoft plans to share more details about this unified platform roadmap at Microsoft Build 2026, continuing to address the enterprise need for simplified, secure, and sovereign AI deployment at scale.

For developers interested in getting started, NVIDIA Nemotron models are available in the Microsoft Foundry AI Model Catalog, with documentation for Foundry Local deployments and Azure Physical AI toolchain access provided through Microsoft's resources.