The race to build DeepSeek in Europe is on

Against this background, Europe’s reliance on American-made artificial intelligence is increasingly starting to look like a liability. In a worst-case scenario, although experts consider this possibility remote, the United States could choose to block access to artificial intelligence services and critical digital infrastructure. What makes more sense is that the Trump administration could use Europe’s reliance as leverage as the two sides continue to reach a trade deal. “This dependency is an obstacle in any negotiations, and we will increasingly negotiate with the United States,” Taddeo says.

The European Commission, the White House and the UK Department of Science, Innovation and Technology did not respond to requests for comment.

To hedge against these risks, European countries have tried to bring AI production in-house, through funding programmes, targeted deregulation, and partnerships with academic institutions. Some efforts have focused on building competitive large language models for native European languages, such as Apertus and GPT-NL.

As long as ChatGPT or Claude continues to outperform European-made chatbots, America’s lead in AI will grow. “A lot of times, these are winner-take-all areas,” Najdel says. “When you have a very good platform, everyone goes for it.” “Not being able to produce the latest technology in the field means you won’t be able to catch up. You will always be feeding the big players with your input, so they will improve and you will be further behind.”

Pay attention to the gap

It is unclear precisely how far the UK or the EU intend to go further with the push for “digital sovereignty”, as lobbyists claim. Does sovereignty require complete self-sufficiency across the sprawling AI supply chain, or merely improved capability in a narrow set of disciplines? Does it require the exclusion of US-based providers, or does it only require the availability of local alternatives? “It’s quite ambiguous,” says Boniface DeChambris, senior policy director at the Computer and Communications Industry Association, a membership organization for technology companies. “It seems more like a narrative at this point.”

There is no broad agreement on which policy tools we should use to create the conditions necessary to enable Europe to achieve self-sufficiency. Some European suppliers are calling for a strategy whereby European companies would be required, or at least incentivized, to buy from local AI companies – similar to the approach announced by China towards its domestic processor market. Unlike grants and subsidies, such an approach would help boost demand, says Ying Cao, CTO at Magics Technologies, a Belgium-based company developing AI processors for use in space. “This is more important than just access to capital,” says Kao. “The most important thing is that you can sell your products.” But those who advocate open markets and deregulation argue that trying to exclude US-based AI companies risks putting domestic companies at a disadvantage against their global counterparts, as they are left free to choose which AI products suit them best. “From our point of view, sovereignty means having a choice,” de Chambres says.

But for all the disagreement over the finer details of the policy, there is widespread belief that closing the performance gap for American leaders remains largely possible even for labs with limited budget and resources, as DeepSeek has shown. “If I really thought we wouldn’t catch up, I wouldn’t do it [try]”, says Najdel. SOOFI, an open source model development project in which Najdel is involved, intends to roll out a competitive general-purpose language model with approximately 100 billion parameters over the next year.

“Progress in this area will no longer depend in large part on the largest GPU clusters,” Najdel says. “We will be the European DeepSeek.”