As tensions rise between the US and its European allies, a quiet but urgent effort is underway across the continent to reduce reliance on American dominance in Artificial Intelligence. While US firms – Nvidia, Google, OpenAI, and others – currently control the vast majority of AI development and market share, European labs are seeking new paths to competitiveness. This isn’t just about technological pride; it’s about national security and economic leverage in a rapidly changing geopolitical landscape.
The American Lead: A Difficult Reality
For years, the US has held a clear advantage in the AI space. From chip manufacturing to datacenter capacity and model design, American companies consistently outperform their European rivals. Some analysts believe this gap is unbridgeable, mirroring the long-standing dependency on US cloud services. Belgium’s national cybersecurity chief recently stated Europe has “lost the internet,” and must accept reliance on American infrastructure.
This dependence isn’t simply an inconvenience; it’s a strategic vulnerability. The US could theoretically withhold access to critical AI services, or use Europe’s reliance as leverage in trade negotiations.
China’s DeepSeek: A New Blueprint
However, the success of China’s DeepSeek AI lab has shattered the notion that sheer computing power alone determines AI leadership. DeepSeek demonstrated that imaginative model design and efficient research can overcome hardware disadvantages. This has galvanized European researchers to pursue alternative strategies.
“We have been too gullible to the narrative that innovation is done in the US,” argues Rosaria Taddeo, a digital ethics and defense technology professor at Oxford. “That’s a dangerous narrative.”
Open-Source Collaboration: Europe’s Potential Edge
One key advantage for European labs is a willingness to develop AI openly. By publishing models for anyone to use and refine, breakthroughs can compound through collaborative efforts. “You are multiplying the power of these models,” explains Wolfgang Nejdl, director of the L3S Research Center in Germany, part of a consortium building a large language model for Europe.
This contrasts sharply with the closed-shop approach of many US AI giants, which guard their training data and model details closely.
Geopolitical Urgency
The urgency is heightened by strained relations between Europe and the Trump administration. Disputes over Greenland sovereignty, tariffs, immigration, and tech regulation have raised concerns about the future of the NATO alliance.
Recent clashes – including a $140 million fine levied against X (formerly Twitter) by the European Commission and retaliatory threats from US officials – underscore the growing tension. European leaders recognize that reliance on American AI is increasingly a liability.
Onshoring AI: Funding, Deregulation, and Native Models
European nations are responding with funding programs, targeted deregulation, and partnerships with academic institutions. Efforts are underway to develop competitive large language models in European languages, like Apertus and GPT-NL. However, as long as models like ChatGPT and Claude outperform European alternatives, the US lead will likely persist.
“These domains are very often winner-takes-all,” notes Nejdl. “Not being able to produce state-of-the-art technology means you will not catch up.”
The Path Forward: Sovereignty or Choice?
The precise scope of Europe’s “digital sovereignty” remains unclear. Does it require complete self-sufficiency, or merely improved capabilities in select areas? Should US-based providers be excluded, or simply offered alongside domestic alternatives?
Some advocate for policies that incentivize or require European businesses to buy from homegrown AI firms – a strategy reportedly employed by China. Others warn that such measures could disadvantage European companies compared to global peers.
Despite the disagreements, most agree that catching up with the US is possible, even for resource-constrained labs. The SOOFI project, led by Nejdl, aims to release a competitive 100 billion parameter language model within the next year, proving that progress doesn’t solely depend on the largest GPU clusters.
“Progress in this field will not to the larger part depend anymore on the biggest GPU clusters. We will be the European DeepSeek.”
Ultimately, Europe’s success hinges on its ability to innovate strategically, collaborate openly, and reduce its dependence on American dominance. The race is on, not just for technological supremacy, but for geopolitical leverage in the age of AI.
