Europe is considering pulling out of the AI race

Brussels. The heads of state and government of the European Union will meet in Dublin on Friday for their first discussion on a topic that has long since become strategic for the future prosperity and security of Europe: artificial intelligence. “Given the rapid pace at which artificial intelligence is developing, I believe the time has come to launch an initial discussion at leader level on the opportunities offered by AI and its transformative impact on our economies and societies,” wrote António Costa, President of the European Council, in his invitation letter. The leaders will be asked three questions. However, having adopted the world’s first legislation to regulate the risks of artificial intelligence, the central question the 27 member states must answer is this: should the EU abandon the race for cutting-edge AI and focus on industrial applications, or is there still a chance to keep pace with the United States and China?

The fact that EU leaders are holding their first discussion on AI in 2026 speaks volumes about their ability to keep up with the times, moving away from the permanent crisis management mode that has characterised the EU for over a quarter of a century now. With the AI Act, the EU institutions have focused on the risks of artificial intelligence. According to critics, too many rules have hampered development in Europe. According to its supporters, the AI Act was ahead of its time by setting out guidelines for the global regulation of a technology that poses risks to humanity. Mario Draghi’s report, published in September 2024, set out a path forward by creating a European innovation ecosystem: competing on cutting-edge models (though not necessarily all of them), achieving infrastructure autonomy (computing power, cloud and semiconductors) and integrating AI into European businesses and services through the industrial applications of existing models. There was a plan, but there was no sense of urgency to put it into action.

The three questions posed to the leaders by Costa do not directly address the dilemma between cutting-edge models and the industrial application of AI. “How can Europe become a leading and sovereign player in strategic areas of the AI value chain (computing, cloud, models and chips), as well as in the integration of AI across the entire economic system?”, asks the President of the European Council. “How can European businesses – particularly SMEs – and public administrations adopt AI on a large scale? How can we mobilise public and private investment on a large scale to build a competitive European AI ecosystem, and what role should public demand play?” The choice of whether or not to abandon state-of-the-art models may become a fait accompli. And not only because Europe is failing to mobilise the private and public investment enjoyed by major American and Chinese players. The President of the Commission, Ursula von der Leyen, appears to want to focus almost exclusively on industrial applications.

Von der Leyen’s stance on AI keeps pace with current events. In September, following a month of debate on the potentially catastrophic effects of certain cutting-edge models on humanity, von der Leyen acknowledged in her State of the Union address that Europe “will also need its own capabilities, particularly to safeguard our national security. ‘We must stay in the race, as this is crucial for our independence,’ she added. But the key part of the Commission President’s speech lies elsewhere. ‘We do not necessarily have to develop cutting-edge technologies ourselves to be able to derive maximum value from them,’ said von der Leyen.

Sources within the Commission have confirmed to us that the President’s strategy on AI is to focus on industrial applications, abandoning the race for cutting-edge models. The gap with the United States and China is considered too wide. The costs are too high. Europe has only one major player, Mistral, which is struggling to keep up. Von der Leyen has appointed a special envoy for industrial AI, the Dane Jim Hagemann Snabe, chairman of Siemens, who could present his initial recommendations ahead of the informal European Council in Dublin. But von der Leyen’s underlying strategy is already summarised in the part of her State of the Union address that was not influenced by current media events.

“The key (for the EU) lies in creating the value that artificial intelligence has not yet realised: bringing AI from the screen into the real economy and society. In manufacturing plants, hospital wards, precision agriculture, our energy networks – the so-called ‘smart grids’ – autonomous driving and the defence sector,” said the Commission President in September. “This is where the real value lies. And it is precisely here that Europe enjoys significant advantages. We have world-class industries that possess data of the very highest quality. This data represents a precious European treasure: it is the resource capable of fuelling bespoke industrial artificial intelligence models.” Several observers expect a deregulation initiative to facilitate the use of data in industrial AI applications.

António Costa has invited two guests to Dublin to stimulate discussion amongst the leaders. Nobel laureate Geoffrey Hinton will speak about the wider impact of AI on societies, as well as issues relating to security and trust. Mistral’s chief executive, Arthur Mensch, will focus on Europe’s position in the global AI landscape and on the strategic issues the continent must address to strengthen its competitiveness and sovereignty. Mensch will also be asked to present his vision on the potential of AI for industrial transformation, promoting its wider adoption across the economy. However, in an interview with Politico, Mensch has already issued a warning: “The idea that we can build something meaningful without owning the foundations – that is, the reasoning models, the ‘brains’ capable of running factories and helping to design better systems – is, in my view, a real trap”.

Over the past six years, the EU has discovered the damage that dependence on others can cause. Russia has used gas as a weapon against Europe in the war against Ukraine. China has cut off supplies of critical raw materials and chips in response to trade wars. The United States uses security as a weapon of coercion. This has become the era of ‘kill switches’: the buttons that can shut everything down, from F-35s to connected electric buses. Von der Leyen’s decision to focus on industrial applications, without investing in cutting-edge models, reflects a strategic mindset that is now outdated: the EU is too large a market for Anthropic or OpenAI to decide to exclude it from their models. The Trump Administration’s decision in June to impose export controls on Anthropic’s two models, Mythos 5 and Fable 5, should serve as a reminder to everyone that tomorrow, cutting-edge AI models could become the new Russian gas.