Europe’s AI Independence Push Intensifies as Researchers Chase a “European DeepSeek”

Key Takeaways

  • European R&D groups are betting on open collaboration to narrow the AI capability gap with the United States
  • Geopolitical tensions and regulatory clashes are amplifying concerns about dependence on US-controlled AI infrastructure
  • A new 100‑billion‑parameter model from the SOOFI consortium is positioned as a major test of Europe’s ability to build competitive foundational AI

Across research labs and government offices in Europe, the conversation has taken a sharper turn in recent months. What used to be framed as a long-term aspiration—building autonomous AI capabilities—now feels more like an urgent strategic requirement. It’s not just about innovation anymore. It’s about leverage, negotiation, and in some scenarios, basic resilience.

Part of this shift comes from the visible strain in transatlantic politics. What was once treated as a stable, almost automatic alliance now features recurring friction. Disputes over trade subsidies, tariffs, and tech regulation have been piling up, and they’re no longer viewed as isolated incidents. European policymakers have noticed that these tensions increasingly intersect with technology, and especially artificial intelligence.

Here’s the thing: American companies still control most of the infrastructure that powers modern AI. Nvidia dominates processors. OpenAI, Google, Meta, and Anthropic set the pace on model development. Capital, talent, and market gravity continue to flow toward the US ecosystem. Belgium’s top cybersecurity official summarized the current reality bluntly earlier this year, saying Europe had “lost the internet.” Many took that comment as defeatism; others as a wake-up call.

Not everyone is buying the narrative that Europe must simply accept a dependent position. Rosaria Taddeo of the University of Oxford put it plainly, warning that Europe has been “too gullible” about assuming innovation is solely an American domain. Her argument taps into something deeper: if critical AI services sit outside Europe’s control, then every negotiation—trade, security, regulatory—carries an implicit vulnerability. Even if the US would never cut off access outright, the possibility of pressure is enough to shift the balance.

That said, this isn’t just geopolitics. The technical side of the story is shifting too. DeepSeek’s rise rattled assumptions about what determines AI competitiveness. The Chinese lab demonstrated that clever model engineering can offset limited compute resources. As a result, European researchers see a more open path. If massive GPU farms aren’t the only ticket to top-tier performance, then Europe’s constraints matter less than they once did.

This belief is fueling projects like SOOFI, Apertus, and GPT‑NL. Wolfgang Nejdl, a leading figure within SOOFI, argues that openness is Europe’s strategic advantage. American labs keep their architectures and datasets tightly guarded. European groups are designing their models in full view of research partners, startups, and even competitors. The theory is simple: shared discovery accelerates faster than siloed development. Whether this approach can overcome the winner‑takes‑all dynamics of the AI market remains an open question.

And those dynamics are real. Once a platform pulls ahead—whether a chatbot, developer assistant, or enterprise automation model—it quickly becomes the default choice. Users generate more data. Investors direct more funding. Developers build more tools around it. The cycle strengthens itself. This is one reason Europe’s AI gap feels both bridgeable and precarious. If the performance delta grows too large, domestic alternatives risk becoming permanent underdogs.

A quick tangent here: the debate about “digital sovereignty” doesn’t help clarify things. The term is everywhere, yet lobbyists and policymakers still disagree on what it actually requires. Some argue for end‑to‑end self-sufficiency. Others just want credible domestic options. A few envision a regulatory push to prioritize European suppliers, not unlike China’s approach to its chip ecosystem. Meanwhile, free‑market groups counter that forcing European companies to avoid US AI systems would put them at a global disadvantage.

Adding complexity, hardware innovators note that government grants matter far less than steady demand. As Ying Cao from Magics Technologies explained, selling competitive products beats receiving one‑off subsidies. That perspective underscores a quieter truth: Europe’s AI ambitions depend not only on research breakthroughs but also on cultivating a home market willing to adopt and trust domestic solutions.

For now, Europe is pushing ahead with a mix of optimism and urgency. Hundreds of millions of dollars have already been allocated to boost training infrastructure, support academic labs, and fund emerging model developers. The SOOFI consortium’s planned 100‑billion‑parameter model will be an important benchmark. If it demonstrates competitive performance—or even a credible trajectory—Europe’s open‑source strategy may look far more viable.

Will clever engineering be enough to counter the massive scale of American AI giants? That’s the question lingering over many of these initiatives. Some researchers admit privately that they don’t know. But they also argue that waiting longer would guarantee failure. As Nejdl put it, if they believed catching up were impossible, they wouldn’t be trying at all. Europe may have missed the first internet wave. Artificial intelligence, however, still feels unsettled. And in that uncertainty, Europe sees an opening—narrow, but still open—to shape its own technological future.