It’s 2031. The U.S. vice president calls the leaders of the Netherlands, France and Germany on a secure line. The offer? Give the U.S. control over the Dutch lithography machine maker ASML, or else Washington will prevent Europe from accessing much-needed U.S. computing power — in effect threatening to turn off Europe’s digital lights overnight.
Under pressure from the U.S. and China, the EU caves. Under this fictional AI doomsday scenario, the EU loses the one company that has helped it maintain some tech sovereignty in the global AI transformation and becomes a permanent vassal to the U.S. in the 21st century’s tech wars.
That’s the final step in a sequence of events laid out by a group of European AI researchers and released in mid-June. The scenario, called Europe 2031, has gone viral on social media and is the talk of the town in Brussels-based tech circles and at AI conferences around the continent. Across nearly 20,000 words, the online opus explores how Europe could lose technological sovereignty during the AI boom and how EU policymakers are erring.
However, the scenario is not just another rehashing of the perennial debate over competitiveness and regulation. Europe 2031’s release, and the reaction to it, also reflects a split within Europe’s AI community between those who believe Europe needs to accelerate in AI to stay relevant globally and shape policy, versus those who think that would sacrifice Europe’s safety and sovereignty.
Europe 2031 also reflects a belief among many European techies that the bureaucratic, risk-averse culture of Brussels is hopelessly out of touch with the earth-shattering changes going on in Silicon Valley — a frustration that’s struck a chord across Europe.
“The scenario plays into an existing sentiment” that some in Europe have with Brussels, said Daan Juijn, the director of research at the Arq Foundation and one of the people behind the project. “For many people, we said the quiet part out loud.”
What is Europe 2031?
Europe 2031 begins in January 2025 with the release of the Chinese model DeepSeek, and in an epilogue stretches all the way to June 2034. The scenario follows the stories of two Europeans: one works for the European Commission on AI policy, the other has gone to Silicon Valley to make it big with an AI start-up. In the years to come, the Commission staffer struggles to get her superiors to take seriously the threat that AI poses to Europe’s way of life, while the Silicon Valley start-up founder becomes wildly successful.
The scenario is rich in technical details that speak to the extent of Europe’s AI challenge. Europe 2031’s writers report a huge gap between Europe and the rest of the world in data center capacity. At the start of the scenario, Europe is behind the U.S. by around 16 gigawatts of compute power. By the end, that’s over 200 gigawatts. With such a disparity, Europe is unable to run key systems on its own, leading to the loss of ASML forced on it by Washington in the final chapter.
Surprising also is how the fictitious timeline for Europe’s loss of sovereignty has matched reality — and already underestimated it. Only two days after Europe 2031’s release, the U.S. government demanded Anthropic prevent any non-U.S. citizens or those based outside of the U.S. from accessing its latest Fable model. Frontier models being blocked from EU release is something Europe 2031 had only forecast for 2029.
Still, the exact numbers and details in the scenario matter less than what the narrative reveals about the essence of reality — that Europe’s challenge with AI is cultural, not technical, and how it’s being lost in translation between Silicon Valley and Brussels.
After a visit to Silicon Valley, the Commission staffer marvels at the huge gap in understanding of AI between the people building and regulating it. “It is 9,000 kilometres from Brussels, but it feels further,” thinks the Commission official in the scenario. “The idea that AI is sparking a new industrial revolution is a truism in California; in the European Commission offices, it is bordering on science fiction.”
With some caveats, that cultural divide still holds water, according to conversations with around a dozen people working closely on European AI policy. Silicon Valley and Brussels continue to approach AI very differently.
“What rings true in the scenario is that Europeans can be very quick to judge American ways of working and innovating,” said Nicole Lemke, an AI researcher at the European tech think tank Interface. “There are a lot of assumptions that our system is somehow ‘better.’ The European policy discourse was too quick to cast aside the competition at the AI frontier without acknowledging how big and risky of a bet that would be.”
Holes in the Europe 2031 plot
However, Europe 2031 has also come in for its fair share of criticism, especially from those who say the storyline ignores AI safety. Instead of directly emphasizing AI safety, the Europe 2031 authors argue that the only way to build a safer global AI is for Europe to compete at the frontier of the technology — a nuance they only stick in the FAQ.
To do so, the Europe 2031 authors propose that Europe deregulate everything tied to data center construction to try and catch up to the U.S., and even work with companies such as Amazon and Microsoft to build U.S.-owned data centers on European soil.
That’s a controversial suggestion, and European AI researchers at the Large-Scale AI Risks Conference in Leuven, Belgium on June 24 were quick to dismiss it as counter to Europe’s actual points of leverage and sovereignty interests.
“Copying the American model with hyperscalers and huge, autonomous models in Europe should not be followed because of the safety consequences,” said Risto Uuk, the head of European policy and research at the Future of Life Institute. He also said that the negative scenario ignored bright spots for Europe, such as the success of the Code of Practice to regulate large AI models, or recent progress on the 28th Regime that could create a pan-European legal framework for businesses.
“The recommendation to spend public capital on local AI investments in partnership with American companies is misguided, and would provide Europe with a false sense of security,” said Zach Meyers, director of research at the Centre on Regulation in Europe. Instead, he suggested doubling-down on Europe’s advantages in lithography, chip packaging and specialized AI models to stay globally relevant.
Defending the recommendation to partner with U.S. companies, Juijn said that there are only so many European firms that can build data centers at scale. In a worst case scenario, he explained, Europe could cut the power to U.S.-owned data centers, but only if they were located on European soil.
Giorgos Verdi, policy and advocacy lead at the Open Markets Institute Europe, dismissed the depiction of EU policymakers being completely naive to the threat that AI poses. While there was “a small drop of truth” in the lack of communication between Silicon Valley and Brussels that the scenario poses, Verdi says the Commission’s portrayal was off.
“It’s unfair to have this characterization that EU policymakers are out of touch with AI,” Verdi said. “These people in the Commission may not be engineers themselves, but they have a strong understanding of the technology.”
Ultimately, the fascination around Europe 2031 reaches beyond questions about Europe’s future. It also speaks to present-day concerns about the state of Europe’s approach to AI — and a growing urgency for Europe to do more now while there’s still a window to make changes.
Sign up to The Parliament's weekly newsletter
Every Friday our editorial team goes behind the headlines to offer insight and analysis on the key stories driving the EU agenda. Subscribe for free here.