Europe 2031 Warns of AI Compute Gap and US Seizure of ASML

A Europe that controls just five per cent of global AI compute, against America's eighty, finds itself by 2031 choosing between American protectorate status and handing the future to China: that is the scenario a coalition of AI researchers, think-tankers, and investors has put before policymakers, covered by The Guardian. The document, called Europe 2031, traces how three misjudgements made in 2025 set the decline in motion: underestimating how fast AI would move, underestimating how much it would change, and overestimating Europe's ability to close the gap.

The second of those misjudgements is examined with particular care. In early 2025, the Chinese model R1 was read as proof that catching up was cheap and that compute barely mattered. The scenario's authors argue this was wishful reading. Efficiency and compute compound rather than substitute, and Europe drew comfort from the wrong lesson. Three individual US AI labs already command more compute capacity than the entire continent combined, including UK sites and US-run facilities in Europe.

The scenario's depiction of 2026 to 2031 is specific enough to sting. A ransomware wave triggered by freely available frontier models, US rationing of compute on a tiered country-by-country system, American corporations buying distressed European carmakers and machine-tool manufacturers, French debt spiraling as welfare costs climb and the tax base erodes, and the euro under sustained pressure as southern Europe follows. The only remaining card in Europe's hand, according to the document, is ASML, the one bottleneck the entire AI race runs through. By 2031, the scenario has Washington issuing an ultimatum and moving to seize direct control of the company, leaving Europe to choose between becoming an American protectorate, handing the future to China, or withering away in isolation.

The authoring coalition is broad: Michiel Bakker from MIT and Google DeepMind, Philip Fox of the KIRA Center and co-author of the International AI Safety Report, Lily Stelling from the Oxford Martin AI Governance Initiative, and several Brussels-based researchers and investors, with Tom Chivers helping shape the narrative form. The document identifies summer 2026 as the last actionable window before the gap becomes self-reinforcing.

Scenario documents are built to be alarming, and this one earns its alarm from specificity rather than hyperbole. The honest caveat is that what the reporting does not give you is a clear account of what acting in that window would actually require: which compute investment thresholds, which policy instruments, which industrial levers would change the trajectory. The scenario names the deadline but not the key. For practitioners, the thing worth watching is whether any European government can convert ASML's genuine chokehold position into a domestic AI investment mandate before that leverage is claimed by someone else.