At a Glance
- European AI labs are pursuing open-source models to counter US dominance
- Governments have pledged hundreds of millions to reduce reliance on American AI
- DeepSeek’s success proved that massive GPU clusters aren’t the only path to competitiveness
- Why it matters: Europe’s push for “digital sovereignty” could reshape global AI supply chains and trade negotiations
European governments and research labs are mounting an urgent campaign to break their dependence on American artificial intelligence, committing hundreds of millions of dollars and betting on open-source collaboration to close a yawning performance gap with US leaders.
US firms currently outpace European rivals at every stage of AI development-from chip design and datacenter capacity to model training and applications. The US also captures the lion’s share of global AI investment, a trend that boosted domestic stocks and economic growth last year. Some officials contend that Europe has already “lost the internet” and must accept reliance on US infrastructure.
The Open-Source Gambit
Rather than chase ever-larger GPU clusters, European labs are experimenting with transparent development. By publishing models for public use and modification, researchers hope breakthroughs will compound as collaborators refine them.
> “You are multiplying the power of these models,”
> said Wolfgang Nejdl, computer-science professor at Germany’s Leibniz Universität Hannover and director of the L3S Research Center.
Nejdl’s consortium is building a large language model optimized for Europe. The approach mirrors China-based DeepSeek, whose breakout success last year undercut the belief that only the biggest processor fleets win.
> “We have been too gullible to the narrative that innovation is done in the US-that we lost the AI train and should not even think about it,”
> said Rosaria Taddeo, professor of digital ethics and defence technologies at the University of Oxford.
> “That’s a dangerous narrative.”
Geopolitical Urgency
Transatlantic tensions have added urgency to Europe’s push. Recent clashes include:
- A $140 million fine levied on X by the European Commission in December
- A UK regulator’s investigation into AI-generated sexualized images on X, prompting US threats of retaliation
- Ongoing disputes over Greenland sovereignty, tariffs, and immigration
> “The geopolitical situation has changed the way we should interpret sovereignty… This technology is an infrastructure-and an infrastructure we do not produce,”
> Taddeo said.
> “We have to start moving in that direction. It’s not possible to ignore it anymore.”
Officials fear the US could restrict access to AI services or use Europe’s dependence as leverage in trade talks.
> “That dependency is a liability in any negotiation-and we are going to be negotiating increasingly with the US,”
> Taddeo added.
Onshoring Efforts
European nations are trying to onshore AI production through:
- Targeted funding programs
- Selective deregulation
- University partnerships
- Language-specific models such as Apertus and GPT-NL
Yet until home-grown chatbots rival ChatGPT or Claude, America’s lead may only widen.
> “These domains are very often winner-takes-all. When you have a very good platform, everybody goes there,”
> Nejdl said.
> “Not being able to produce state-of-the-art technology in this field means you will not catch up.”
Defining Sovereignty
Lobbyists note that “digital sovereignty” remains loosely defined. Questions linger:
| Aspect | Possible Interpretations |
|---|---|
| Scope | Total self-sufficiency across AI supply chain OR focus on select disciplines |
| Providers | Exclude US firms OR ensure domestic alternatives exist |
| Policy | Mandatory buying preferences OR open markets |
> “It’s quite vague,”
> said Boniface de Champris, senior policy manager at the Computer & Communications Industry Association.
> “It seems to be more of a narrative at this stage.”
Some European suppliers advocate purchase mandates or incentives for domestic AI, akin to China’s support for home-grown processors. Ying Cao, CTO at Belgium-based Magics Technologies, argues guaranteed demand is more critical than capital.
> “The most important thing is that you can sell your products,”
> Cao said.
Others counter that shutting out US firms risks handicapping European businesses that benefit from choosing the best global tools.
> “From our perspective, sovereignty means having choice,”
> de Champris said.
The DeepSeek Precedent
Despite policy disagreements, researchers see room to catch up. DeepSeek proved that imaginative model design can rival brute-force scale. Nejdl’s SOOFI project plans to release a 100-billion-parameter open-source language model within a year.
> “Progress in this field will not to the larger part depend anymore on the biggest GPU clusters,”
> Nejdl said.
> “We will be the European DeepSeek.”
Key Takeaways
- Europe is pouring public funds into open-source AI to reduce US dependency
- Success hinges on whether collaborative development can outpace the closed ecosystems of Nvidia, Google, and OpenAI
- Without competitive performance, Europe risks remaining a data supplier that strengthens American platforms
