Europe risks irrelevance in the frontier AI race

The continent can regain clout by controlling strategic links of the AI supply chain and strengthening state capacity, argue two researchers from RAND

Afek Shamir
Fellow, RAND Center on AI, Security, and Technology
Max Negele
Fellow, RAND Center on AI, Security, and Technology

23 Feb 2026

Europe grid on globe with a network in blue

Europe’s leaders are learning to live with a harsher world. They are rethinking the transatlantic relationship and facing geostrategic competition with China, all while Russia’s invasion of Ukraine grinds on at Europe’s eastern flank. There is another urgent concern, too. Artificial intelligence systems capable of matching humans on nearly all economically valuable cognitive tasks may arrive within just a few years. And on its present trajectory, Europe will have little say over how this technology is built or governed.

The capabilities of leading AI systems have advanced quickly since the launch of ChatGPT in 2023. Cutting-edge models now achieve gold-medal scores in elite mathematics Olympiads, outperform humans on PhD-level science benchmarks and have even won their inventors Nobel prizes. Some AI systems can reason across domains and execute complex tasks with minimal human oversight. The adoption of AI is also happening faster than almost any technology in history. This includes its use across businesses, but also in laboratories and security agencies.

As a result, the revenues of leading AI firms have soared to more than $10bn. Some argue that AI-related investment has created a financial bubble. Even if this proves true, the impact of the technology on the global balance of power is profound and hard to deny.

The capabilities of AI will not remain confined to the civilian realm. Two security consequences stand out. The first is diffusion. AI can lower the cost of acquiring dangerous offensive capabilities. Current models can already help design biological agents, and as they improve, they may make it easier for non-state actors to develop chemical or biological weapons. The second is military transformation. For states, more capable AI systems promise to transform how warfare is conducted through automated intelligence analysis, faster operational planning, more sophisticated cyber-attacks and, ultimately, autonomous weapons. The war in Ukraine offers early glimpses of this shift. While AI does not yet confer a decisive advantage on the battlefield, AI-powered drone operations and automated intelligence analysis are playing a growing role.

Countries that develop and control the most advanced AI systems stand to capture a disproportionate share of their potential gains. It comes as no surprise, then, that having an edge in AI is a strategic priority for great powers. Under the administrations of both President Joe Biden and President Donald Trump, America has pursued AI leadership as a national-security imperative, championing domestic firms and imposing semiconductor export controls on China (some now controversially revoked). Beijing, for its part, aims to become the world’s AI leader by 2030 and is racing to indigenise its supply chain. Its reported progress on extreme-ultraviolet lithography, a crucial step in producing advanced semiconductor chips, does not mean China will match the West imminently. But it does signal the intensity of a whole-of-state effort.

If Europe wants a seat at the table, it must become competitive. This matters not merely for safeguarding European values, but for ensuring that the development of a potentially civilisation-shaping technology is not left entirely to great-power rivalry. Europe could be the continent where AI is built with safety, security and the preservation of fundamental rights and democracy at its core.

Unfortunately, Europe is not well-positioned to play this role. A recent report by RAND and leading academics identifies three key reasons.

The first is Europe’s dependence. On its current trajectory, transformative AI will be built almost entirely outside the continent. The development of cutting-edge models is concentrated in a small group of firms, overwhelmingly American and Chinese ones. Control over computing power, the critical resource underlying AI progress, is similarly skewed. The European Union hosts less than 5% of the world’s supercomputing performance, compared with nearly 75% in America. Talent is another aspect. Many leading European AI researchers have accordingly relocated to the US to access more plentiful resources.

This dependence has broader implications. Reliance on others for advanced models, cloud capacity and cutting-edge chips leaves Europe exposed to coercion.

A competitive homegrown frontier model would help, but it is not the only path to relevance. Previous general-purpose technologies offer a clue. For example, Britain invented electricity, but America and Germany benefited just as much by deploying it effectively. Europe could pursue a similar strategy with AI and focus on where value is created along the technology’s value chain. Three areas stand out.

Start with applications. Europe still holds large troves of specialised data in sectors like manufacturing and health care. Such data are often proprietary, hard to replicate and closely tied to real-world processes, which can prove more valuable in building useful AI innovation than a marginally better model. Next comes Europe’s leverage upstream. In semiconductor equipment, ASML, a Dutch lithography giant, remains indispensable to advanced chipmaking and thus occupies a critical position in the AI supply chain. Energy presents a harder challenge. Europe’s high electricity costs and slow permitting deter data-centre investment. But accelerating nuclear capacity and streamlining approvals could reverse this weakness.

The second reason for Europe’s lag is institutional capacity. The majority of European governments are not able to track frontier AI and grasp its policy implications the way leading nations do. Staying abreast of frontier developments typically requires embedding top technical talent in government and securing access to leading models for evaluation.

Britain has been an exception, setting a high bar. Its AI Security Institute (UK AISI) has been treated as a “startup in government” and has been endowed with a cumulative £340m since its inception in 2023. This backing has enabled the country to maintain testing access to leading American AI models and gain insight into frontier US AI companies. Britain also hosted the influential Bletchley summit in 2023, which brought together America and China, along with 26 other nations, to co-operate on preventing the most catastrophic AI safety risks.

The EU’s equivalent centre for expertise is the AI Office, which has recruited capable researchers but commands a starting budget of just €46.5m, less than half the £100m starting budget of its British counterpart. It mandates evaluations of frontier models, but focuses on enforcing the EU’s AI Act rather than building strategic awareness across European governments. The result is a double risk. Europe may struggle to grasp what frontier AI can already do and its implications, let alone where it will be in five years. And it may fail to build the right types of indigenous assets—including talent, compute, advanced models and applications—needed to reduce its dependence on unreliable partners.

The third reason is a mismatch of tempo. Existing EU strategies are a start, but they are not yet calibrated to the pace and scale of the competition. For example, EU plans for AI infrastructure and compute buildout, while substantial, do not close the gap with America and China. The planned European AI “gigafactories”, together comprising several hundreds of thousands of advanced chips, are dwarfed by comparable American infrastructure. OpenAI, one of America’s leading AI firms, expected to control well over one million GPUs by the end of 2025.

There are signs of growing ambition. The Franco-German Digital Sovereignty Summit in November 2025 announced a Frontier AI Initiative that marks a potentially significant shift. Unlike earlier EU programmes focused on infrastructure and sectoral adoption, this initiative explicitly targets capability competition at the frontier and aims to attract world-class talent capable of developing state-of-the-art systems.

Yet the implementation path remains uncertain. No concrete budget has been published for the initiative, and a timing gap between its planned launch in early 2026 and gigafactory availability in 2027-28 means any recruited teams would initially lack the frontier-scale compute that leading AI companies already use.

If Europe is to address these shortcomings, it needs to ramp up speed and ambition. The window for securing AI competitiveness is a matter of years, not decades, and the stakes extend well beyond Europe itself. If the development of transformative AI is dictated entirely by great-power competition between America and China, there is no guarantee that the interests of broader humanity will prevail. The decisions Europe takes in the next few years will determine whether it has a voice in shaping what comes next, or merely lives with the consequences.

Views or opinions expressed are those of the author and any individuals cited, and do not necessarily reflect those of Economist Enterprise or any other member of The Economist Group.