The Competition to Create Europe’s DeepSeek Has Begun

Given the current circumstances, Europe’s dependence on American AI technology increasingly resembles a vulnerability. In a scenario that experts deem unlikely, the US could opt to restrict access to AI services and essential digital infrastructure. More realistically, the Trump administration might leverage Europe’s reliance as part of ongoing trade negotiations. “That dependency is a liability in any negotiation—and we are going to be negotiating increasingly with the US,” notes Taddeo.

Requests for comments from the European Commission, the White House, and the UK Department for Science, Innovation and Technology went unanswered.

To mitigate these risks, European countries have sought to bring AI production back onshore through funding initiatives, targeted deregulation, and collaborations with academic institutions. Some of these efforts have been aimed at developing competitive large language models for native European languages, such as Apertus and GPT-NL.

However, as long as ChatGPT or Claude consistently outshine European chatbots, America’s dominance in AI will only escalate. “These domains are typically winner-takes-all. When a platform excels, it attracts everyone,” explains Nejdl. “Not being able to create cutting-edge technology in this area signifies a perpetual lag. You end up merely providing the larger players with your insights, enabling them to improve further while you fall behind.”

Mind the Gap

There remains uncertainty regarding how far the UK or EU plans to pursue “digital sovereignty,” according to lobbyists. Does sovereignty necessitate absolute self-sufficiency within the extensive AI supply chain, or is it sufficient to enhance capabilities in specific areas? Does it require excluding US-based suppliers or simply ensuring the availability of local alternatives? “The definition is quite ambiguous,” remarks Boniface de Champris, senior policy manager at the Computer & Communications Industry Association, a trade organization for technology firms. “At this point, it seems to be more of a narrative.”

Moreover, there is no consensus on which policy measures to implement for fostering European self-sufficiency. Some local suppliers advocate for a strategy mandating, or at least incentivizing, European businesses to procure from domestic AI companies—akin to China’s reported practices in its local processor sector. Unlike grants and subsidies, this strategy could stimulate demand, argues Ying Cao, CTO at Magics Technologies, a Belgium-based firm focused on developing AI-specific processors for space applications. “That’s more critical than just access to capital,” states Cao. “The key factor is that you can successfully market your products.” Yet, proponents of open markets and deregulation warn that attempting to exclude US AI companies could disadvantage local enterprises, leaving them to select from the best available AI solutions. “From our perspective, sovereignty means having options,” says de Champris.

Despite the differences in policy details, there is widespread agreement that closing the performance gap with American leaders is very much achievable, even for labs with limited budgets and resources, as demonstrated by DeepSeek. “If I believed we could not catch up, I would not even try,” asserts Nejdl. The SOOFI project, which is focused on open-source model development and in which Nejdl participates, aims to release a competitive general-purpose language model boasting approximately 100 billion parameters within the next year.

“Advancements in this field will primarily no longer rely on the largest GPU clusters,” asserts Nejdl. “We will be the European DeepSeek.”