Microsoft relaunches sovereign cloud: AI models and productivity even without connection

The sovereign cloud has become one of the most sensitive areas of Europe's digital transformation: an assumption that by now is (almost) no longer news, such is the attention that governments and large organisations are devoting to the study of architectures capable of reducing technological dependence and supporting a greater centrality of data in industrial strategies. All the main American platforms, from Amazon Web Services to Google Cloud to Oracle, have for some time been busy declining 'local' offers or partnerships with national operators to respond to the demands for control and compliance that meet the dictates of digital sovereignty. This is certainly not a new theme for Microsoft as well, and so the announcements made by the Redmond company in the last few hours should not come as much of a surprise, since they are aimed not only at guaranteeing the (sovereign) residency of data or specific levels of compliance, but also at ensuring operational continuity, governance and computing capacity even in extreme conditions, when the connection to the public cloud is unavailable or has to be intentionally interrupted.

In the blog that formalises the upcoming news for Microsoft Sovereign Cloud, Douglas Phillips, President and Chief Technology Officer of Microsoft Specialised Clouds, frames the rise of digital sovereignty as a strategic component with stark words: 'organisations are rethinking the way they deploy critical infrastructure and artificial intelligence capabilities, in a context of more stringent regulatory expectations and higher risk conditions'. It is therefore no longer an option, but a structural requirement for companies (especially those active in regulated sectors) and public administrations, and the North American giant's approach aims precisely at eliminating architectural fragmentation, relying on a 'package' of sovereign options between public and private environments. At the heart of this approach is a 'full stack' logic that integrates infrastructure, productivity and workloads in the cloud and (in Microsoft's intentions) allows customers to choose the level of control best suited to each workload, without increasing operational risk. Business confidence, Phillips points out, is based on the certainty that 'data remains protected, controls are enforceable and operations can continue under real-world conditions'. What marks this new round of announcements, however, is the extension of data processing capabilities even to completely disconnected environments, where (as the blog notes) 'connectivity can be intentionally limited and business continuity becomes a business imperative', especially for organisations that operate under stringent regulatory constraints and cannot afford network-related disruptions or vulnerabilities.

The new features of Sovereign Cloud are articulated along three lines and reflect the desire to make available to companies and public bodies a coherent and convergent model that standardises governance and operational practices between connected, hybrid and disconnected environments. The first relates to the availability of Azure Local in disconnected mode: the advantage promised to businesses and PAs served by Microsoft's cloud, Redmond confirm, is that of being able to enable even without connectivity mission-critical infrastructures in line with the governance dictated by the platform's policies (Azure), optimising operational continuity for sovereign and classified environments. Management, policy and execution of workloads remain largely within customer-controlled environments, allowing services to continue to run securely even when the environment needs to be isolated or the network is unavailable. The second 'new entry' is Microsoft 365 Local as a disconnected version: with this upgrade, key productivity workloads (Exchange Server, SharePoint Server and Skype for Business Server) can operate entirely within the customer's sovereign perimeter on Azure Local, with guaranteed support until at least 2035. The goal is to maintain active collaboration and communication even when disconnected from the cloud, ensuring that data, access and compliance remain under the direct control of the organisation itself. The third announcement is perhaps the most significant in perspective: Foundry Local is enhanced with greater infrastructure capabilities and (most importantly) support for large artificial intelligence models in fully disconnected sovereign environments. Thanks to the new generation of hardware provided by partners such as Nvidia, Microsoft is now able to enable local execution of multimodal models on customer systems, bringing advanced AI inference capabilities within strict sovereign boundaries and in even completely disconnected environments. Phillips speaks not coincidentally of a 'truly localised full-stack experience', fully scalable as inference needs grow, and thus of the ability of trained AI models to apply acquired knowledge to new data to make decisions or generate content in real time.