What Europe 2031 gets wrong

A viral new essay is full of vivid warnings about Europe's tech dependencies. But its proposals would only make the problem far worse.

“It’s March 2031 and Europe is on its knees. After racking up hundreds of billions of euros in public debt to subsidise Big Tech’s data centres, many of which now sit idle, the region’s economy was hit worse than most when the AI bubble finally burst. Despite soaring unemployment and increasingly violent protests, Europe’s bankrupt governments can do little to ease the pain.

President Vance, in an attempt to distract Americans from a similar economic collapse at home, orders the sudden invasion of Greenland after years of escalating threats. When Europe eventually retaliates with military force and sanctions, the Vance administration immediately shuts off the region’s access to US cloud computing and semiconductors, devastating its economy, crippling critical infrastructure, and leaving its governments in the dark.

It didn’t have to be this way: Europe could have chosen a different future. One where it took the hard but necessary steps to reduce its tech dependencies, including by investing at scale in sovereign alternatives. One where it avoided the worst excesses of the AI bubble by prioritising sustainable data centres and efficient, open-source AI models designed in line with European values and built to complement Europe’s industrial strengths.

Instead, Europe mistook dependence for pragmatism – and paid the ultimate price.”

Like the story above, the viral essay Europe 2031 uses fiction to depict a bleak future in which Europe is subject to the whims of other technological superpowers, having failed to develop its own capabilities. Yet instead of arguing that Europe should urgently build those capabilities, the authors recommend policies – including laying out the red carpet for Big Tech’s data centres – that would make this future far more likely.

To begin with, the style of the essay is problematic in itself. Instead of transparently presenting their analysis, assumptions and conclusions up front, the authors cloak these in a fictional scenario set in the near future, in which an EU official and a German tech entrepreneur exchange texts as Europe sinks into decline.

While this makes for an entertaining read, the emotional tone primes readers to accept at face value the authors’ premises, many of which are highly debatable. As the opening of this article shows, the same technique can be used to draw very different conclusions.

Europe 2031 is right to warn about the grave risks of Europe becoming overly dependent on foreign AI, mirroring the continent’s existing dependencies across other parts of the tech stack. The recent move by the US government to restrict overseas access to Anthropic’s advanced AI models is an early illustration of what could be at stake.

But instead of urging Europeans to double down on their efforts to develop independent capabilities that could mitigate this threat, the story implies that all such efforts will end in failure.

In the world depicted in the story, a 2027 “Digital Sovereignty Regulation” mandating use of European cloud and AI for public sector workloads results in “inferior home-grown alternatives” that undermine Europe’s competitiveness and expose it to devastating cybersecurity attacks. “Public investments, compute subsidies” and “preferential procurement contracts” are also unable to keep Europe in the AI race. What the authors seem to be saying is that government intervention to prevent AI from being dominated by a handful of non-European giants is futile.

What alternative does Europe 2031 point towards then, given its stark warnings about Europe’s dependencies? Paradoxically, the authors’ proposals would not only leave those dependencies untouched, but deepen them considerably.

Instead of urging Europe to go further in its efforts to loosen the hyperscalers’ chokehold on its cloud computing infrastructure, the authors call for “massive investments in compute” led by those very same American providers. Oddly, the authors argue that this would somehow give Europe “leverage” to “secure guaranteed access to frontier AI”, which apparently it should not even try to develop itself.

Deepening Europe’s dependency on US cloud providers – which already control 70% of the region’s computing infrastructure – would give Washington, not Brussels, additional leverage by reinforcing the former’s capacity to cripple the European economy and public sector by withdrawing access to American technology.

The idea that accelerating Europe’s colonisation by foreign data centres would “secure” the region’s access to “frontier” AI, which could be shut off in the same way at any moment, is equally difficult to reconcile with the story’s repeated warnings about the risks of reliance on US technology.

Europe 2031’s solution – that this foreign-owned infrastructure could be kept “under European jurisdiction” – is also implausible. There is no conceivable European law that could prevent the US government from weaponising cloud infrastructure to spy on, manipulate, blackmail or sabotage Europeans – whether through export controls, US laws mandating access to extraterritorial data, or old-fashioned surveillance backdoors.

No matter what the hyperscalers may say, keeping their own government happy will always be a greater priority than complying with European regulation, particularly when that regulation directly contradicts orders from the White House. The recent designation of Anthropic by the Trump administration as a national security risk, limiting the AI startup’s ability to win US public contracts, highlights what is at stake for American companies that try to resist their own government.

One thing Europe 2031 gets right is the need for a “coalition of aligned AI middle powers” with collective leverage vis-à-vis the US and China.

Yet here again, the coalition is conceived primarily as a way to “secure access to frontier AI” and “demand safer, more reliable models”, rather than building those capabilities among coalition members themselves. Moreover, Europe would only be valuable to such a coalition if it could offer allies credible alternatives to technology currently monopolised by US tech giants, a possibility that – as we have seen – the authors advise against.

The broader implication is that not only Europe, but also South Korea, Japan, Canada and the UK – all countries with highly advanced technological capabilities – are incapable of developing cutting-edge AI and should not even try. This is a problematic assumption to say the least, given the often unlikely winners of technological transitions. After all, who would have guessed in 1945 that Japan and Germany, whose economies were devastated in World War II, would later become world-beating producers of cars, consumer electronics, semiconductors and robotics?

Europe 2031 is also deeply pessimistic when it comes to the ability of European governments to counter Big Tech’s dominance and stand up to foreign coercion.

The Digital Services Act, an imperfect but potent tool for limiting the power that tech giants have over Europe’s public square, is dismissed in the story as a sideshow. The Anti-Coercion Instrument – a law empowering the EU to retaliate against foreign threats, including by limiting market access – is never used due to divisions among EU member states. Competition law and the Digital Markets Act, which could be used to tackle the dependencies that the essay’s authors appear so concerned about, are never even mentioned. The overall message is that enforcement is not worth the effort.

That overlooks one of Europe’s strongest sources of actual leverage: access to its large and lucrative market. Greater willingness to restrict access to that market could help Europe secure short-term access to advanced AI while it develops its own capabilities.

As one analysis points out, if AI labs were limited to the US market, they would have to earn an average of $2,400 a month from 70 million American white-collar workers, a highly unlikely scenario. In other words, without access to the European market, the AI bubble would likely pop, bringing the entire American economy and possibly the US government down with it.

More fundamentally, Europe 2031 sees no role for people and democratic governments in reining in powerful corporations and steering technological development towards the public interest.

The story disparages hard-fought employment protections for slowing down job automation, and mocks European workers for daring to use AI productivity gains to “enjoy family time, long lunches, and afternoon walks”. The authors’ preferred approach is so-called “flexicurity”, where the impact of AI on jobs is simply a given and the state alone bears the full cost of supporting and retraining those affected.

Yet as Nobel-prize-winning economist Daron Acemoglu and others have argued, the impact of technology on work – whether it reconfigures, degrades, enhances, or simply eliminates jobs – is not inevitable, but a contingent process shaped by political choices, societal norms and countervailing forces including organised labour.

Throughout the story, contingent pathways for AI development are presented as inevitable, while the central role of corporate power – in promoting certain technologies and business models and suppressing others – is airbrushed out of the picture.

Although most of Europe 2031’s arguments hang on the unquestionable need for Europe to keep up with the US on AI, the essay doesn’t make a convincing case for this belief.

For starters, there is no guarantee that massive investments in compute will remain the key to success in AI. There is growing evidence that AI models are becoming “commoditised” as performance gaps between them narrow. Even if “frontier” AI remains necessary for a limited set of tasks, slightly less performant but cheaper and more efficient models are likely to suffice for most of the day-to-day applications that truly matter for Europe’s productivity.

While the future is inherently hard to predict, any scenario which requires Europe to sacrifice its social model and climate goals to fund a speculative infrastructure buildout deserves intense scrutiny.

The authors further undermine their case by depicting a US political system and society torn apart by the impact of AI. According to the essay, by 2028 “most Americans want little to do with AI” and “increasingly take to the streets” while inequality explodes and corporate executives face assassination attempts. By 2030, although “the protests…have not stopped, the US government has “bought time” with “job guarantees and direct cash transfers”.

If the reward for “keeping up” in AI is mass unemployment and political violence compensated by government handouts and made-up jobs, then perhaps Europe is better off looking elsewhere.

Europe 2031 uses science fiction and gripping prose to promote a policy agenda that is defeatist, dystopian, and anti-democratic. While the diagnosis might be compelling, the cure is likely to be worse than the disease.