“Your data, your models, your decisions.” That’s how Microsoft sums up sovereign AI, and it was the thread running through my session at NVIDIA GTC Berlin 2026. If you couldn’t make it to Berlin, or you want to watch it again, the recording of Sovereign AI, From Azure to the Edge (SB61158) is now available on NVIDIA On-Demand.
Watch the session: Sovereign AI, From Azure to the Edge on NVIDIA On-Demand
Why sovereign AI matters
Almost every organization I talk to wants to use AI on its most valuable data. And almost all of them ask the same follow-up questions: Where is my data processed? Who controls the models? And what happens if we need to switch models, or the connection drops?
For me, digital sovereignty is fundamentally about risk management. It’s not about picking one location for everything. It’s about starting with the workload, understanding its requirements and applying the right level of control. AI makes this a bit more interesting, because an AI workload is more than stored data. Prompts, retrieved context, embeddings, models, responses and telemetry each come with their own residency and access questions.
What I cover in the session
In the session, I show how Microsoft Sovereign Cloud helps you use AI with control, across a continuum that goes from Azure into your own datacenter and out to the edge.
Microsoft Sovereign Cloud and sovereign AI. Microsoft Sovereign Cloud spans Sovereign Public Cloud, Sovereign Private Cloud and a sovereign partner ecosystem. Sovereign AI is one of its core scenarios, next to policy and governance, cloud operator access restrictions, data residency and confidentiality, and cloud resilience and business continuity.
Sovereign AI in Azure. Many AI workloads can run in Sovereign Public Cloud with the right controls: region choice and the EU Data Boundary for residency, Azure confidential computing to help protect data while it’s being processed, and Microsoft Purview to classify, protect and monitor the data your AI touches.
Bringing AI to your data with Azure Local. When data must stay under your own authority, you bring AI to where the data lives. Foundry Local on Azure Local brings the Foundry experience into your own datacenter, connected or fully disconnected, with a catalog of Microsoft, partner, open-weight and open-source models. Microsoft has announced its general availability for Microsoft Ignite 2026. Combined with NVIDIA accelerated computing on Azure Local, this supports demanding inference and agentic workloads on your own hardware.
All the way to the edge. The same operating model extends from your datacenter to edge locations, including environments that need to run fully disconnected from the cloud. That’s where resilience becomes part of the sovereignty conversation: which AI functions must keep running if the connection drops?
Control, choice, flexibility and resilience. These are the four principles of Microsoft’s sovereign AI framework, developed together with NVIDIA. One design choice I highlight in the session: decouple the model from the platform. AI moves too fast to assume one model will remain the right answer for every workload.
Key takeaways
- Sovereign AI is about control over data, models and operations, not just location.
- Start with the workload and apply controls per layer: source data, embeddings, inference, model and outputs.
- Sovereign Public Cloud covers many AI workloads. Foundry Local on Azure Local brings AI to your data when it must stay under your authority, even disconnected.
- Keep model choice open. Flexibility is a resilience control.
Learn more
- Watch the session: Sovereign AI, From Azure to the Edge (SB61158)
- Sovereign AI: Accelerate innovation with control and choice
- Advancing digital sovereignty with greater control and choice
- Sovereign Private Cloud: Bringing more choice and control to AI and cloud infrastructure
- Sovereignty implementation guidance on Microsoft Learn
- My earlier post: Azure Local Disconnected Operations: Running Sovereign Cloud, Productivity, and AI in Air-Gapped Environments
I hope you enjoy the session. If you have any questions, or want to share how you approach sovereign AI in your organization, leave a comment below or reach out to me on LinkedIn.
