By Michał Puchała · 2026-07-16 · 3 min read
EU cloud news, week of 2026-07-16
Airbus selected Scaleway for sovereign cloud workloads, while EU experts named compute and energy as Europe's urgent AI priorities. Germany's Soofi S put sovereign infrastructure to work, and Exoscale launched dedicated AI inference in European data centres.
European cloud sovereignty became easier to measure this week. Airbus put a major industrial workload decision behind a European provider, while new research and product launches showed where European AI infrastructure is gaining substance and where it still lacks scale.
Airbus selects Scaleway for sovereign cloud workloads Airbus has selected Scaleway to provide sovereign cloud infrastructure for selected enterprise applications that require strong governance, resilience and legal protection. The decision is a useful procurement signal because it comes from a company whose operations span regulated markets, complex supply chains and long-lived industrial systems. For boards, it shows that a European cloud can now enter the shortlist for more than isolated or experimental workloads. Technical teams will still need to examine service coverage, availability design and migration requirements, since the announcement does not provide that architectural detail.
EU experts identify compute and energy as Europe's urgent AI priorities The EU AI Office has published findings from more than 100 experts on Europe's position in frontier AI. Their assessment credits Europe with strong research and talent, but says its computing infrastructure and growth-stage capital remain below the scale needed for leading model development. Compute capacity and the energy to run it are described as the most urgent priorities for the next two years, with the coming one to two years potentially decisive. For business leaders, the practical message is that an AI strategy now needs a dependency map covering model access, hosting, energy and the ability to change suppliers.
Soofi S puts Germany's sovereign AI infrastructure to work A German research consortium has presented Soofi S, an open-source German-English model trained on Deutsche Telekom's Industrial AI Cloud in Munich. The model has about 30 billion parameters but uses roughly 3 billion for each piece of text it processes, an approach intended to reduce the computing needed during use. Its team also plans to release weights, training and evaluation code, and detailed accounting of the source material used for training. The performance results come from the project's own report and need independent testing, but the broader achievement is concrete: European infrastructure has supported the full training of an auditable model designed for practical deployment.
Exoscale launches dedicated AI inference in European data centres Exoscale's new Dedicated Inference product gives organisations a managed environment with dedicated NVIDIA GPUs in European or Swiss data centres. Customers can run a model of their choice without operating the underlying hardware and can adjust capacity as demand changes. That reduces an important operational barrier for teams that want more control over where their AI workloads run without building an internal GPU platform. CTOs should still examine model licences, logging, support arrangements and portability, because European hosting and dedicated hardware solve only part of the dependency question.
The common direction is clear: European cloud and AI sovereignty is moving from policy language into procurement decisions, working infrastructure and products that technical teams can assess and deploy with confidence.
Thinking about migration? Book a free consultation to discuss your situation.