Microsoft is bundling its sovereignty offering under the name Microsoft Sovereign Cloud and, with Azure Belgium Central, has had its own cloud region in Belgium since late 2025. The promise: extra control, transparency, and enforceable governance on top of the familiar hyperscale cloud, although digital sovereignty is not automatically guaranteed by this alone.
Digital sovereignty dominates every cloud conversation today. The US Cloud Act, geopolitical tensions, and the rise of AI are making organizations question their dependence on American tech companies. Microsoft feels that pressure and is responding with a sovereign offering that now has its own name, an extensive toolbox, and even a Belgian home base.
Jimmy Bandell, Product Manager Public Cloud at Cegeka, guides organizations through its implementation and explains exactly what you get—and what you don’t.
Not just another Azure
First, the naming issue: Microsoft Sovereign Cloud is the evolution and expansion of the former Microsoft Cloud for Sovereignty. More importantly, what it is not: a separate public cloud. “It is a set of capabilities, deployment models, tooling, and reference architectures on top of Microsoft cloud services such as Azure and Microsoft 365,” says Bandell. “It’s not just another Azure, but an extra layer of governance, control, and demonstrability.”
That extra layer should enable three things: translating regulatory requirements into technical controls, making those controls repeatable and scalable, and better supporting audits with platform data and logging. For Belgian organizations already on Azure, this primarily solves a governance and evidence problem. Today, compliance agreements are often scattered across documents, processes, and loose configurations. With Sovereign Landing Zones and Azure Policy, you enforce a standardized baseline.
It’s not just another Azure, but an extra layer of governance, control, and demonstrability.
Jimmy Bandell, Product Manager Public Cloud at Cegeka
“This shifts the focus from ‘we think this is set up correctly’ to ‘we can demonstrate which controls are active’,” Bandell summarizes. However, he warns immediately: the sovereign layer does not automatically make compliance fully settled. It primarily helps to consistently enforce and demonstrate technical controls.
Five building blocks for control
Which sovereignty mechanisms does Microsoft Sovereign Cloud actually bring together? Bandell distinguishes five domains. First, data residency: more control over where data is stored and processed, within the chosen region or geography, although this depends on the Azure service and configuration used. Second, customer-controlled encryption via Customer Managed Keys, for example with Azure Key Vault or Managed HSM. External Key Management, where key material remains outside Microsoft infrastructure, is currently available as a limited preview.
Third, policy-driven governance: Azure Policy and Sovereign Landing Zones automatically enforce technical controls around location, encryption, network, and configuration.
The fourth domain, operational sovereignty, revolves around the question of who can access your environment. Mechanisms such as Customer Lockbox and Data Guardian restrict Microsoft’s support access with approval flows and logging. Data Guardian, part of the offering since 2025, must also guarantee that European employees oversee access to European customer environments. Finally, there is transparency and auditability: better audit trails and tamper-evident logging for provider operations and support interactions.
From public cloud to fully disconnected
Microsoft Sovereign Cloud comes in different flavors: public with controls, local (disconnected), and national partner. The Sovereign Public Cloud adds sovereignty controls within existing European data center regions: the same services as the global cloud, but with extra locks on them. The Sovereign Private Cloud bundles Azure Local, Microsoft 365 Local, and Foundry Local for those who want to run workloads within their own operational boundary, if necessary fully disconnected from the public cloud.

Since early 2026, Azure Local and Microsoft 365 Local can also operate without cloud connectivity, and Foundry Local brings large AI models to fully disconnected environments. Additionally, there are National Partner Clouds, where a trusted local partner operates a sovereign cloud environment.
That choice is a trade-off. Those moving towards Azure Local disconnected sacrifice functionality compared to the full hyperscaler, trading services for extra sovereignty. The market has yet to clarify the right use cases for these most shielded variants: will they serve as daily work environments or rather as a safe haven when the public cloud fails?
Azure Belgium Central: data on Belgian soil
Since November 18, 2025, Belgium has had its own Azure region. Azure Belgium Central consists of three availability zones in the outskirts of Brussels. For Belgian organizations, this primarily means local data residency: workloads and services available in Belgium Central can be physically placed in Belgium, and the data remains there at rest.
A Belgian Azure region is not the same as full digital sovereignty. It primarily helps with the question: where is my data located?
Jimmy Bandell, Product Manager Public Cloud at Cegeka
Additionally, the region can lower latency for Belgian users and systems, although this depends on the workload and network architecture. And there is the regulatory aspect: a Belgian region helps when dealing with government, customers, and regulators when local anchoring is important.
Bandell does add an important caveat. “A Belgian Azure region is not the same as full digital sovereignty. It primarily helps with the question: where is my data located? Sovereignty is also about who has access, how keys are managed, how support access is handled, and how controls are demonstrably enforced.”
AI and Copilot under the same regime
The question remains how AI fits into this picture, as these workloads are currently the growth engine of every company. For Bandell, the answer is simple: AI and Copilot fit within a sovereign framework as soon as they fall under the same governance as the rest of the platform. Microsoft 365 Copilot adopts existing permissions and does not correct historical oversharing; well-configured access rights and data classification therefore remain necessary. Microsoft also states that prompts, responses, and Microsoft Graph data from Copilot are not used to train the underlying foundation models.
With Microsoft Purview, you manage AI interactions through classification, DLP, auditing, retention, eDiscovery, and insider risk controls. Those building their own solutions can use Azure OpenAI or Foundry with their own data within a controlled Azure environment, where data residency depends on the chosen region, deployment type, and service configuration.
Finally, AI workloads fit into the same sovereign landing zones, with identical policy, network, identity, and compliance frameworks as other Azure workloads. “The key question is not whether or not to use AI, but how to use AI in a controlled manner,” says Bandell. “With correct permissions, data classification, logging, policy enforcement, and clear choices regarding region and model deployment.”
Trust must be demonstrated
When taking stock, it’s clear that Microsoft Sovereign Cloud is primarily a governance layer that you activate on top of what is already there: no move to a different cloud, but extra locks, keys, and logs. The Belgian region strengthens that story but does not solve it on its own, and the homework regarding data classification and governance remains the responsibility of the organization itself.
Ultimately, sovereign cloud is not about a single region, a single tool, or a single policy.
Jimmy Bandell, Product Manager Public Cloud at Cegeka
“Ultimately, sovereign cloud is not about a single region, a single tool, or a single policy,” Bandell concludes. “It’s about provable control: knowing where your data is, who has access, how you enforce compliance, and how you deploy AI safely and responsibly. In an AI-first world, trust is no longer something you just promise, but something you must be able to demonstrate technically.”
This is an editorial contribution in collaboration with Cegeka. For more information about their solutions, you can go here.
