The next chapter of AI is about more than powerful models. It’s about bringing AI closer to the data, applications, and systems that organizations depend on every day. For governments, regulated industries, and organizations operating critical infrastructure, this means making AI accessible in environments where data is sensitive, connectivity may be limited, and greater operational control is essential.
Microsoft’s approach to sovereign AI starts with the requirements of each workload. Our vision is to give organizations the choice and flexibility to determine where AI runs, which models they use, and how their data and infrastructure are managed. Depending on the workload, this can include control over where data and models reside, who can access and operate the environment, how AI workloads are governed, and whether critical operations can continue when cloud connectivity is restricted or unavailable. The goal is to help organizations meet their security, regulatory, and operational requirements while maintaining access to modern cloud and AI capabilities.
With Foundry Local, entering general availability on Azure Local at Ignite, we’re bringing advanced AI capabilities to customer-controlled environments. This enables organizations to build and run AI applications closer to their data, with greater control over how AI workloads are deployed and operated, extending the possibilities of AI beyond the public cloud.
Bringing the Microsoft Foundry experience to customer-controlled environments
AI delivers the greatest value when organizations can connect it securely to the data, applications, and processes that power their business.
Microsoft Foundry is our unified platform for building, optimizing, and governing AI applications and agents. Foundry Local extends these capabilities to customer-controlled environments, allowing organizations to deploy and run supported AI models on Azure Local infrastructure, with local inference, model management, and GPU acceleration. Built on Azure Arc-enabled Kubernetes, Foundry Local supports CPU- and GPU-backed deployments, model lifecycle management, and OpenAI-compatible APIs. Developers can select from an integrated model catalog or bring their own supported models. Models and inference workloads can run locally, including in supported disconnected environments.
We’re also expanding what organizations can build with Foundry Local, moving beyond standalone inference toward intelligent applications and agents that can securely interact with enterprise data.
These capabilities include:
- Model catalog (general available): Offers a curated model catalog, making it easy to discover and deploy AI models optimized for local and edge environments. Gives organizations greater flexibility to select models based on capability, performance, latency, resource requirements, and the sovereignty needs of individual workloads.
- Agentic Retrieval (general available): Enables applications to ground responses in enterprise knowledge through local data ingestion, indexing, and multistep retrieval, with citations that allow users to verify the underlying sources.
- Agents and enterprise connectors (available in preview): Emerging capabilities for orchestrating agents and connecting them to enterprise systems, including on-premises Exchange Server, SharePoint Server, and SQL databases through Model Context Protocol (MCP).
- Configurable AI guardrails (available in preview): Support for applying safety and compliance controls to local AI workloads, with flexibility to select guardrail models based on security requirements and available compute resources.
Together, these capabilities help organizations build AI applications where the data lives instead of requiring the data to move to the AI.
From local inference to sovereign AI applications
Running AI models locally is an important first step, but sovereign AI extends beyond inference. Organizations need to connect models to enterprise knowledge, build intelligent applications and agents, and maintain control over how those applications are developed, deployed, and operated.
Foundry Local helps make this possible by bringing AI capabilities closer to enterprise data and applications. Developers can use familiar Microsoft tools and technologies to build AI applications across cloud and customer-controlled environments, with greater flexibility over where models run and how they interact with organizational data. Azure Local provides the infrastructure foundation for running these workloads across datacenters, distributed locations, and edge environments, including sovereign private cloud scenarios. For organizations requiring greater operational independence, Azure Local disconnected operations enables supported infrastructure and services to operate through a local control plane without continuous connectivity to the public cloud.
Sovereignty also extends to the software development lifecycle. With GitHub Enterprise Local, becoming generally available at Ignite, organizations can bring familiar GitHub development capabilities to customer-controlled infrastructure on Azure Local. Developers can manage source code, collaborate, and automate build and deployment processes using GitHub Actions and self-hosted runners while maintaining greater control over their repositories, workflows, and artifacts. Foundry Local can also provide locally hosted models for AI-assisted development scenarios in supported disconnected environments.
Together, Foundry Local, Azure Local, and GitHub Enterprise Local provide a foundation for building, deploying, and operating AI applications within customer-controlled environments, extending sovereignty from the infrastructure to the models, applications, and development processes that run on it.
What this means for customers
The significance of these capabilities is ultimately about what they enable organizations to do.
- Keep sensitive data in the environment they control. Organizations can bring AI to data that may need to remain within a specific facility, geography, security boundary, or operational environment.
- Run AI where decisions need to happen. Manufacturers, retailers, transportation providers, healthcare organizations, governments, and others can bring intelligence closer to physical operations and distributed locations rather than depending on every interaction traveling to a remote cloud service.
- Support environments with limited or disconnected connectivity. Organizations operating in remote, isolated, or highly controlled environments can continue to pursue AI scenarios where constant public cloud connectivity may not be appropriate or possible.
- Maintain greater control over the AI stack. Customers can make deliberate choices across infrastructure, data, models, applications, operations, and governance based on the requirements of individual workloads.
- Turn enterprise data into useful AI experiences. Retrieval, connectors, agents, and model capabilities can help organizations build AI experiences grounded in their own information and business processes while keeping those capabilities closer to the systems they support.
Most importantly, customers do not have to define a single architecture for every workload. Some AI workloads will benefit from the scale and rapid innovation of the public cloud. Others will require greater local control. Many organizations will need both. Our strategy is to give customers the flexibility to choose the right architecture for each workload while maintaining a consistent foundation across their environments.
Sovereignty that enables innovation
As AI becomes more deeply embedded across applications, operations, and critical infrastructure, organizations will need architecture that give them both powerful AI capabilities and meaningful control over how those capabilities are deployed.
With Azure Local and Foundry Local, we’re continuing to bring Microsoft AI innovation to the places customers need it, from the cloud to customer-controlled datacenters and edge locations. That is the foundation of our approach to sovereign AI: giving every organization the control and choice to pursue its AI ambitions on its own terms.
Resources
- Join us at Microsoft Ignite in San Francisco to explore the latest innovations across Microsoft AI, Azure Local, and sovereign cloud.