Europe bets its AI sovereignty on a Milan startup most people have never heard of

The European Commission has selected Milan-based startup Domyn to lead the EUROPA consortium, tasked with building an open-source frontier AI model exceeding 400 billion parameters in all 24 EU languages within a year. The project aims to reduce Europe's dependency on US and Chinese AI providers, allocating up to 2.5% of EuroHPC supercomputing capacity. Domyn, a little-known scale-up valued at $1 billion, will collaborate with Fraunhofer-Gesellschaft and other European labs to deliver the model.

The European Commission has handed a relatively unknown Italian scale-up the job of building Europe's answer to GPT and Claude, a frontier AI model covering all 24 EU languages, to be delivered in a year. On June 19, the European Commission named EUROPA, a consortium led by Milan-based Domyn, as the winner of its Frontier AI Grand Challenge. The prize isn't cash exactly. It's access to up to 2.5% of total EuroHPC computing capacity for one year across the bloc's AI-optimised supercomputers, the largest single compute allocation ever offered to a European AI project. The task: build an open-source frontier model exceeding 400 billion parameters in all 24 official EU languages. The deadline: roughly twelve months. That's a serious mandate. For context, GPT-4 is widely estimated at around 1.8 trillion parameters, and Claude and Gemini operate at comparable scale. A 400-billion-parameter model won't be catching OpenAI tomorrow, but it would put Europe squarely on the frontier map for the first time, and doing it in 24 languages, from Finnish to Maltese, makes it more technically complex than almost anything attempted before at this scale. What the Commission is really buying isn't just a model. It's a hedge. As Il Sole 24 Ore reported in its coverage of the announcement, the EU's explicit motivation is reducing dependency on US and Chinese AI providers. That framing has grown more urgent through 2026 as the Trump administration has tightened export controls on AI chips and, as Sifted noted, the prospect of restricted access to American AI services has moved from theoretical to plausible. Europe's governments are reaching the uncomfortable conclusion that running critical public infrastructure on models controlled by foreign companies is a strategic liability. EUROPA is the Commission's most direct answer to that concern yet. Here's the thing about Domyn: until about two months ago, most people in European tech hadn't heard of it. The company was called iGenius until it rebranded, founded in Milan in 2016 by Uljan Sharka, an Albanian-born, Italian-raised entrepreneur who came to Italy at 16, taught himself the language, worked his way into Apple's certification program without a degree, spent time in Silicon Valley, and came back to Europe with a specific thesis: enterprise technology was orders of magnitude less accessible than consumer technology, and AI could fix that. iGenius built its reputation doing AI for regulated industries, financial services, government, heavy industry, the sectors where data sovereignty isn't a talking point but a legal requirement. It reached a billion-dollar valuation in 2024 and has around 150 staff from 25 countries. Then it rebranded to Domyn and went after something much larger. The Commission judged Domyn's proposal the most convincing in terms of strategic vision, execution capability and potential impact, and the consortium it leads includes Fraunhofer-Gesellschaft, the dominant European applied research organization, alongside other European AI labs and institutions. Fraunhofer's involvement matters: it's not decoration, it brings genuine model research infrastructure that a 150-person scale-up doesn't have on its own. Sharka told Reuters the model will launch within a year. You can read that as confidence or as the only answer you can give when the European Commission just handed you the continent's flagship AI project. Frankly, it's probably both. Training a 400-billion-parameter multilingual model from scratch in twelve months, on shared supercomputing time, with a consortium of institutions that haven't worked together before, is a serious undertaking for any organization. Doing it at Domyn's current headcount, even with Fraunhofer alongside, raises real questions about execution risk that nobody in Brussels seems especially eager to discuss publicly. The Frontier AI Grand Challenge itself was only launched in February 2026, four months before Domyn was announced as winner. That's a fast competition cycle for a project of this consequence, and it suggests the Commission was more focused on getting something moving than on exhaustive vetting. Whether that pace reflects confidence in the process or political pressure to show results ahead of an EU election cycle is worth watching. What's less debatable is the strategic logic. Europe has the compute, through EuroHPC. It has the research institutions. It has the regulatory framework, through the AI Act, that could actually make an open, auditable, multilingual model attractive to public-sector clients who can't use black-box American systems. What it has consistently lacked is a credible actor to pull those pieces together. Whether Domyn is that actor, or whether EUROPA ends up as another well-funded European AI initiative that produces a technical demonstration without real-world traction, will depend entirely on what ships in the next twelve months. Sharka has staked his company's entire identity on this moment. Domyn's rebrand, its pivot to sovereign AI, its positioning against OpenAI and Anthropic directly, all of it now points here. That kind of bet concentrates the mind. It also concentrates the risk. Europe will be watching whether a 150-person Milan startup can actually deliver the infrastructure the continent says it needs. Also read: Waymo registers a German subsidiary and starts recruiting in Berlin and Munich https://startupfortune.com/waymo-registers-a-german-subsidiary-and-starts-recruiting-in-berlin-and-munich/ • SpaceX's $25 billion bond frenzy is the clearest sign yet that markets have lost the plot https://startupfortune.com/spacexs-25-billion-bond-frenzy-is-the-clearest-sign-yet-that-markets-have-lost-the-plot/ • Unconventional AI raises $475 million at a $4.5 billion seed valuation to cut AI's power bill by a factor of 1,000 https://startupfortune.com/unconventional-ai-raises-475-million-at-a-45-billion-seed-valuation-to-cut-ais-power-bill-by-a-factor-of-1000/