Artificial intelligence is becoming critical infrastructure – as important as energy or connectivity. And with that shift comes one question: who really controls the data, models, and compute that businesses now run on?
A 2025 Gartner survey found that 61% of IT leaders in Western Europe plan to increase their use of local or regional cloud providers, mainly because of geopolitical factors. This is not a passing trend, but a new phase of transformation in which technological sovereignty becomes a condition for adopting AI.
This is not about an abstract principle, but about three very practical concerns:
- where sensitive data ends up – from customer records to industrial data and intellectual property;
- how much control an organization keeps over the models it comes to depend on;
- and how exposed it is to decisions taken outside its jurisdiction, from changes in service access to demands to hand over data.
It is geopolitics and data sensitivity, not a passing trend, that are pushing companies to ask about sovereignty.
What is sovereign AI?
Sovereign AI is the ability of an organization or a state to build and use artificial intelligence while keeping full legal and technical control over its data, models, and infrastructure. In practice, this means data does not leave a chosen jurisdiction; systems are run by local personnel, and access by entities outside that legal domain is excluded or tightly supervised.
The term was popularised by NVIDIA CEO Jensen Huang. He argues that every country should own “the production of its own intelligence”, and that AI should be treated as infrastructure – much like electricity or roads.
Sovereignty can be split into four layers:
- data localization,
- operational independence,
- cryptographic control,
- and compliance with local law.
The more of these an organization controls, the higher its level of sovereignty.
Why sovereign AI is gaining importance
The first reason is geopolitics. Rising tensions and dependence on non-European providers make organizations want certainty about where their data physically sits and who – under political pressure – could gain access to it. For sensitive sectors such as public administration, defense, finance, and healthcare, the assurance that data and models stay within European jurisdiction is often a precondition for starting any AI project at all.
The second is data sensitivity. AI feeds on exactly the information organizations guard most closely – customer data, medical records, trade secrets, industrial data. In a cloud model, that information flows into training, prompts, and inference, so the question of who can see it stops being theoretical.
The third is control over the models. Building core processes on a model offered as an external API means depending on the provider’s decisions – on pricing, versions, availability, and usage terms. A sovereign approach lets you choose a model, run it in your own environment, and keep operating even when an external service changes or becomes unavailable.
AI sovereignty isn’t ideology – it’s risk management. You need to know where the data sits, who can access it, and which law governs it, and ask that at the start of a project, not after it goes live.
Marcin Mosiołek, AI Offering Lead at Sii Poland
Microsoft Azure – the most fully developed answer to sovereignty needs
Among the major providers, Microsoft has built the most complete offering for data and AI sovereignty. In June 2025, Satya Nadella unveiled the Microsoft Sovereign Cloud in two variants: public and private. Both keep data in European data centers, managed by European personnel, and External Key Management encryption removes the provider’s own technical ability to access customer data.
But Microsoft goes further than data localization. Azure Local lets organizations run the cloud with a fully local control plane and, from early 2026, even completely disconnected from the internet. In February 2026, the company announced that large AI models can also run securely in that isolated mode (Foundry Local). That matters for defense, the public sector, and critical infrastructure.
The offering is rounded out by Microsoft 365 Local (generally available in the second half of 2025), which runs services such as Exchange and SharePoint locally, and by support for advanced AI hardware, including NVIDIA GPUs. Microsoft has also launched a Digital Sovereignty specialization in its partner program, validating partners’ expertise in delivering sovereign solutions on Azure and Microsoft 365.
Azure’s strength is that sovereignty doesn’t stop at data location. From public cloud to fully disconnected environments, we can match the level of control to a client’s real risk.
Bogusław Kosęda, Cloud Solutions Architect at Sii Poland
AWS European Sovereign Cloud – independent infrastructure in Europe
Amazon Web Services answered the sovereignty need by launching the AWS European Sovereign Cloud – a separate cloud built, operated, and controlled in Europe, with an independent governance structure. General availability was announced in January 2026. The first region was established in Brandenburg, Germany, and AWS committed to a €7.8 billion investment through 2040.
From the outset, the environment offers more than 90 services, including AI services on a “sovereign-by-design” foundation: Amazon Bedrock, Amazon Q, and Amazon SageMaker. AWS plans to extend coverage beyond Germany — with sovereign Local Zones in Belgium, the Netherlands, and Portugal. The most demanding customers can use Dedicated Local Zones and AWS Outposts, deployed at a chosen location, including their own data center.
Google Cloud – a broad portfolio and institutional partnerships
Google Cloud is developing one of the broadest portfolios of sovereign solutions. It spans three models of increasing isolation: Data Boundary, Dedicated, and Air-Gapped. In 2025, the provider strengthened them with an upgraded Data Boundary, a User Data Shield mechanism, and Dedicated deployments run by local operators in Europe.
The scale of commitment shows in its investments – $10 billion in Belgium and $5.5 billion in Germany – and in the first European sovereign cloud hub, opened in Munich in November 2025. The maturity of the offering is also reflected in a multi-million-dollar contract with NATO’s agency (NCIA) to deliver secure, sovereign cloud and AI capabilities.
Artificial Intelligence
We deliver AI solutions tailored to your business, driving efficiency and boosting productivity within your teams.
AI offeringHow to choose the right sovereignty model
The three largest providers now have genuine, production-grade sovereign solutions, but they differ in philosophy and level of control – from public cloud with localization guarantees, through dedicated environments, to fully disconnected installations. So the question is not “which provider is best”, but “which level of sovereignty fits my organization’s data, its sensitivity and its risk”.
It is this stage – translating business needs and data sensitivity into a concrete architecture – that determines a deployment’s success. Sii Poland’s experts support clients in assessing data sensitivity, selecting a sovereignty model, and designing and implementing AI solutions within the Microsoft Azure ecosystem and other platforms, so that innovation goes hand in hand with control over data and models.
Want to deploy AI with full control over your data and the models behind it? You can count on the support of Sii Poland’s experts.
Sources
- Gartner, “Geopolitics Will Drive 61% of CIOs… to Increase Reliance on Local Cloud Providers” (November 12, 2025) – gartner.com
- NVIDIA, “What Is Sovereign AI?” (Jensen Huang) – blogs.nvidia.com
- Microsoft Azure Blog, “Microsoft strengthens sovereign cloud capabilities with new services” (June 2025) – azure.microsoft.com
- Microsoft Blog, “Microsoft Sovereign Cloud… support for large AI models… completely disconnected” (February 24, 2026) – blogs.microsoft.com
- AWS, “Opening the AWS European Sovereign Cloud” (January 2026) – aws.amazon.com
- About Amazon EU, “AWS plans to invest €7.8 billion into the AWS European Sovereign Cloud” – aboutamazon.eu
- Google Cloud, “Launches First Sovereign Cloud Hub in Munich” (November 12, 2025) – googlecloudpresscorner.com
- PR Newswire, “NATO and Google Cloud Sign Deal for AI-Enabled Sovereign Cloud” – prnewswire.com
Leave a comment