The current trajectory of AI calls for the most ambitious political agenda in the history of post-war Europe. Unless we embark on it now, Europe will lose the ability to shape its own future. We will end up economically and politically sidelined, with values we cannot defend, social welfare systems we can no longer fund, risks we cannot address, and a Union that cannot hold.
Europe 2031 is a five-year scenario about Europe's impending slide into irrelevance: how AI is driving it, and what can still be done to change course.
To understand how Europe is at risk of squandering the coming AI revolution, the story first loops back to 2025 - and to three things it got wrong. It misjudged how fast AI would move. It misjudged how much it would change. And it misjudged its own ability to catch up.
How we got here — January 2025 to June 2026
Europe misreads the speed and scale of AI, and a series of reasonable-looking decisions deepen its dependence.
- DeepSeek approaches the frontier cheaply, and Europe draws the wrong lesson. In early 2025, the Chinese model R1 is taken as proof that catching up is cheap and that compute barely matters - a comforting idea. But efficiency and compute compound rather than substitute: more chips make the clever shortcuts easier to find, and DeepSeek's own progress is capped by the AI chips China cannot import.
- The AI Action Summit mostly deals in rhetoric. At the Paris AI Action Summit, the EU announces a €200 billion AI fund that is mostly repackaged money and hoped-for private investment, dwarfed by what the US is actually spending. ‘Sovereignty’ becomes the rallying cry, but actual measures lack teeth and avoid hard trade-offs.
- GPT-5 disappoints, and the bubble narrative takes hold. When OpenAI's GPT-5 underwhelms, European sceptics read it as confirmation of an ‘AI bubble,’ and momentum stalls. Out of view, the opposite is happening: coding agents in Silicon Valley begin automating software engineering, and the leading labs start using their own models to build the next generation.
- The governors don't use the tools. Most European civil servants are barred from frontier systems on data-protection grounds, and few of them know how to code. As a result, those meant to regulate and govern the technology often do not properly understand it.
- Access becomes a favour, not a right. By mid-2026, some European leaders revisit their earlier scepticism. Anthropic's Claude Mythos, withheld from public release, turns out to be capable enough to reshape cybersecurity. Europe is initially left out of the defensive coalition formed around it. Soon after, a US executive order routes new frontier models through a classified review, letting Washington choose which ‘trusted partners’ get them first.
- Europe realises it holds few of the cards. Controlling just five per cent of the world's AI compute against America's eighty, Europe has little leverage to demand access to frontier AI. Its answer is a positive-sounding tech-sovereignty package that remains too little, too late.
What could happen next — August 2026 to March 2031
Europe doubles down on sovereignty but forgets to build leverage, while the AI race between the US and China escalates.
- 2027: a Mythos-level open-source model triggers a ransomware wave, and sovereignty policies backfire. Germany and France have just proposed a bill mandating European-only AI for critical public-sector workloads. So when offensive capabilities spread to anyone who wants them, the organisations that switched to European providers in advance - and are thus running defences well behind the frontier - are the ones locked out of their systems and paying the ransoms. The wave only eases when the US and China both ban open-source frontier models, which leaves Europe more dependent on closed American ones than ever.
- 2028: AI stops reasoning in language humans can read, and Washington forces the Dutch to cut ASML's exports to China. The capability jump breaks the oversight tools regulators were relying on, and the EU AI Office - already locked in proceedings against two American developers - has no room to respond. When the Dutch are pressured to halt exports of ASML's older DUV machines to China, other Member States offer little support. The Netherlands caves, and Europe negotiates nothing in return.
- 2029: the US begins rationing frontier AI by country, and economic divergence accelerates. The growing compute shortage reaches a breaking point, and the US rations frontier inference through a tiered, country-based system that reserves most capacity for itself and a few selected allies. Most of Europe lands in Tier 2 and sees its compute allocation from US cloud companies halved. When the EU reaches for the "trade bazooka" to win Tier 1 status, the vote falls short of a qualified majority. European GDP growth begins to diverge sharply from America's: Europe owns little of the AI stack, adopts it more slowly, and gets only limited access to the frontier models now running large parts of the economy.
- 2030: Europe is hollowed out from abroad as firms are outcompeted and industries bought up. By 2030, the US and China are locked in a race both sides increasingly see as existential. To keep China from winning robotics, the leading US AI company buys up Europe's distressed carmakers and toolmakers for their floor space and industrial data, converting car plants into robot factories. Unemployment rises as automation spreads and better-equipped foreign firms outcompete European ones. French debt spirals as welfare costs climb and the tax base erodes; southern Europe follows, the euro comes under sustained pressure, and the Union begins to fragment. Chinese credit lines appear across the continent, buying goodwill and trying to prise Europe away from Washington.
- 2031: Washington moves to seize ASML, and Europe is left with three terrible options. By 2031, power is more concentrated than ever before in history. A handful of people in San Francisco, Washington, D.C., and Beijing are deciding humanity's future. The only card Europe still holds is ASML - the one bottleneck the entire AI race runs through. Watching Europe drift towards China, the White House decides it needs direct control of the company and issues an ultimatum. Having failed to build any leverage of its own, Europe is left choosing between becoming an American protectorate, handing the future to China, or withering away in isolation.
Why the European model collapses under business-as-usual
AI's impact will match or exceed that of the industrial revolution, but it will arrive in years rather than decades. Europe's current response is ten to a hundred times too small, and aimed at the wrong target. Too often, sovereignty is treated as settling for inferior European solutions while hoping that worthwhile but unlikely moonshots pay off. In reality, what it demands is leverage and a willingness to accept ugly trade-offs. Leverage comes from being indispensable, not half-heartedly self-sufficient - which also means deciding which habits to let go of in order to protect the principles that are non-negotiable: human dignity, equality, and the freedom to shape the continent's own future.
The failure that Europe 2031 describes is one of incentives and institutions, not of individuals. None of the story requires our leaders to act in bad faith. Instead, the very things that served Europe well in calmer times work against it now. Consensus and careful procedure are how it built a Union of twenty-seven; under time pressure, however, they become the reasons hard truths get deferred, acting early looks career-ending, and institutions cannot keep pace with the technology. Every decision seems to make sense on its own, but the sum leaves a Europe that keeps its procedures and loses its principles.
What Europe can still do
Time is short, but we believe Europe’s course can still be changed. As a start, we offer the following five recommendations:
- Enable massive investments in compute and the supply chains beneath it. Mobilise public and private capital at a scale Europe has not attempted in peacetime, directed at the foundations of the AI economy: energy, semiconductors and data centres. Bringing tens of gigawatts of compute onto European soil will require dedicated economic zones, targeted energy policy and radically streamlined permitting. Europe cannot build this alone, and should partner with American providers on terms that keep the infrastructure under European jurisdiction and secure guaranteed access to frontier AI.
- Build a coalition of aligned AI middle powers. European countries are not alone; many other middle powers face similar challenges. Alongside EU cooperation, the Netherlands, Germany and France should build a small, nimble coalition with countries like Norway, the UK, Canada, Japan and South Korea. Each holds a real position in the AI supply chain - talent, compute, semiconductor bottlenecks - that can become joint leverage to secure access to frontier AI or to demand safer, more reliable models. A strong coalition can also mediate between the US and China, which could turn out to be its most important role.
- Reform labour markets for AI adoption. A flexicurity model, like the Danish example, allows firms to adopt AI more deeply while protecting the workers it displaces through retraining and income support. Attempting to preserve jobs unchanged risks losing them to faster-adopting competitors abroad; the more durable course is to steer - rather than block - the diffusion of AI and to distribute gains fairly.
- Expand European strengths in robotics and industrial AI. While it seems unlikely that Europe can still meaningfully compete in LLMs, it can play a key role in the upcoming physical AI revolution. That requires screening foreign investment into European manufacturers, opening industrial data and process knowledge to domestic AI developers, removing the bottlenecks that prevent promising European companies from scaling, and forming partnerships with American companies that yield lasting gains rather than one-off windfalls.
- Offer a positive vision of what AI can do for society. A story about what Europe stands to lose will not carry the required reforms on its own. Many voters already dislike AI, and will not absorb years of AI-driven disruption merely to avoid something abstractly worse. While we haven't attempted to set out a positive vision here, we believe Europe urgently needs one. Both bottom-up social movements and political leaders have a role to play in building it.