AI Sovereignty
When Telefónica's Chief Digital Officer Sebas Muriel Herrero presented the EURO-3C project at the Mobile World Congress in Barcelona in late February — a federated European cloud and AI infrastructure backed by more than 70 organizations — he chose his words carefully: "It's very difficult that Europe can create a hyperscaler from zero, from scratch." So they're trying a different approach: networking existing national infrastructures rather than going head-to-head with Amazon, Microsoft, and Google. It's a sentence that precisely captures the European AI debate in the spring of 2026: they know what they don't have — and are looking for ways to build something of their own regardless.
The numbers behind the catch-up effort sound impressive at first. The EU Commission's AI Continent Action Plan totals 200 billion euros. Two billion are earmarked for so-called AI Factories, with the InvestAI initiative aiming to mobilize a further 20 billion for "gigafactories" — high-performance data centers capable of training frontier models. In January 2026, the Council adopted the legal framework for this. Thirteen AI Factories have already been selected across 17 member states. Nine new AI-optimized supercomputers are set to go online in 2025 and 2026.
But behind the billions lie structural problems. The European Chips Act, celebrated as a breakthrough in 2023, has stumbled: private investors failed to meet their commitments, Intel foremost among them, whose European fab plans stalled. The Commission is planning a revised Chips Act for the first quarter of 2026 — this time, it is said, with a "long-term strategy rooted in geopolitical realities." In advanced chip manufacturing below five nanometers, the gap to TSMC and Samsung remains enormous.
In this vacuum, Mistral AI has become the most important symbol of European AI ambitions — and a test case for whether Europe can compete in the model-development race. The Paris-based company, founded in 2023, closed a 2.8-billion-euro funding round in September 2025, led by Dutch chip equipment maker ASML. The valuation: 11.7 billion euros. With its acquisition of infrastructure startup Koyeb in February 2026 and the launch of Mistral Forge — a platform for on-premise training on sovereign infrastructure — Mistral is deliberately positioning itself as a European counter-narrative to OpenAI and Anthropic.
The Mistral 3 model series, unveiled in December 2025, demonstrates in nine specialized variants that European models can be competitive for everyday use. Le Chat, Mistral's chatbot, performs comparably to ChatGPT for routine tasks. The partnership with SAP offers full European data residency. Around one billion euros are set to flow into high-performance chips and infrastructure in 2026.
And yet a truth remains, one that technology blogger Stefan Pfeiffer recently put bluntly: "If Europe doesn't invest massively, Mistral will probably fail." After Aleph Alpha's retreat from the model business, Mistral is Europe's only remaining answer to the American frontier labs. That's thin redundancy for a continent aspiring to sovereignty.
The AI Act, phasing in since February 2025, was supposed to be Europe's regulatory edge — the counterpart to its industrial weakness. But here too, things are grinding. Over a hundred technology companies, including not just US giants like Alphabet and Meta but also European firms like Mistral and ASML, called for a delay under the banner "Stop the Clock." More than 40 CEOs from Siemens, Airbus, Mercedes-Benz, and others pushed for a two-year suspension to develop "clear and workable standards." Thomas Regnier, the Commission's spokesperson, pushed back: "There is no pause, no transition period, and no stopping the clock."
The Commission responded nonetheless — with the Digital Omnibus on AI of November 2025, addressing feedback from both industry and civil society. Industry largely welcomed it. On the other side, 31 organizations led by European Digital Rights signed a letter opposing any dilution. It is the tension that has run through Europe's AI policy for years: regulation as a trust advantage or as an innovation brake?
Digital Minister Karsten Wildberger set the tone at the Summit on European Digital Sovereignty in November 2025: "Product first, regulation second." And Macron, co-hosting alongside Chancellor Merz, put it more sharply still: "If we leave the field to the US and China, we'll have good regulation, but in the end we'll be regulating nothing."
Europe has a surprising strength in AI talent: roughly 30 percent more AI professionals per capita than the US and nearly three times as many as China. But the best ones leave. European countries are losing "significant AI talent, both national and international, to the United States," research organization Interface found in 2024. US salaries for AI engineers run 30 to 70 percent above European levels.
Yet the wind is shifting — from an unexpected direction. Since the Trump administration proposed massive cuts to the research budget in 2025 and political uncertainty grew across US academia, applications from American researchers to European institutions have surged. The number of US applicants for early-career grants from the European Research Council has nearly tripled, from 60 in 2024 to 169 in 2026. At German universities, US applications doubled in February 2025 — in some cases tripled. It is a geopolitical irony: Europe's greatest chance to attract AI talent owes nothing to its own appeal and everything to America's self-inflicted wounds.
Mario Draghi, whose report on European competitiveness shaped the debate in 2024, warned in February 2026: "Europe risks becoming subordinated, divided, and deindustrialized." What holds true for the economy as a whole is magnified in AI: every layer of the value chain — chips, cloud, models, applications — is currently dominated by non-European players.
The question is whether the EU will draw the right conclusions. The answer won't come in a single step. It will be measured by whether EURO-3C becomes more than an acronym at a trade fair in Barcelona. Whether Mistral's billions are enough to stay at the technological frontier. Whether the Chips Act delivers in its second version what it promised in its first. And whether Europe not only welcomes the talent now knocking at its door, but offers them conditions that make them stay.
Digital sovereignty, as one observer put it, works like a spectrum, not a switch. Europe doesn't have to build everything itself. But it must be able to hold its own options at the points that matter. In the spring of 2026, that is still a promise, not a reality.
AI doesn't change whether people are needed. It changes what they're needed for. Companies that understand this pull the decisive lever. The rest invest blindly. Five imperatives for decision-makers who want to make the difference.
Artificial intelligence optimises networks, improves customer service through personalised chatbots, and enables new data-driven business models. However, challenges such as data privacy must be addressed. AI offers great potential for the industry.
In our latest article, we highlight the central role of compliance and quality management in the development of AI systems. We show how innovative approaches can be reconciled with regulatory requirements — an important step for companies that want to drive sustainable innovations in the fast-moving world of technology.
