From Bill Gates to Bernie Sanders, most agree that the AI arms race is disastrous. Only Europe can stop it | Alejandro Hurst

TOA this point, it should be clear that AI is a choose-your-own-adventure book, and all choices end in disaster. Explosive carbon emissions and ecological collapse? Turn to page 37. Other outcomes include some or all: the bursting of the bubble and subsequent financial crisis; mass unemployment through automation; the end of democracy and the rise of authoritarian surveillance states; delusional cults who believe the pattern matching algorithm is a consciousness; terrorists creating devastating biological weapons; nation states deploying autonomous weapons systems that act in non-human time frames; nation states or terrorists hacking into all the countless systems we depend on; or an AI that, consciously or not, follows directives that are at odds with human well-being.

Imagining that the likes of Elon Musk, OpenAI chief Sam Altman, Palantir’s Peter Thiel, and Trump administration technology advisor Marc Andreessen will lead us to a utopian future of flourishing environmentalism, social harmony, and endless leisure requires hallucinations and adulation so great that that The choose-your-own-adventure scenario could only have been written by ChatGPT.

There is a growing, if wildly overdue, realization that we are heading toward calamity, with accompanying calls to slow down, regulate, or even put all of this on pause while we think about it. These calls come not just from activists, but from a set of politicians, experts and big tech insiders too ecumenical to ignore, from Bernie Sanders to Bill Gates, who recently wrote that “if anyone had a credible plan to slow global AI advances, I would probably support it.” He’s just the latest insider to acknowledge the truth that artificial intelligence company Anthropic made earlier this summer, when it called for a “brake pedal.” Industry self-policing will not work. A pause, if there is to be one, must be imposed from outside.

Despite this, an international agreement is unfortunately unlikely to be reached. AI is analogous to nuclear weaponry and follows the same logic: each individual actor is incentivized to run this silly race, with the only justification being that they don’t want to be at the mercy of whoever wins. The best chance of inducing a slowdown, it seems, would be for a single actor to control a bottleneck and have an incentive to close it. Interestingly, the EU has that card, and even has an incentive to play it, if it dares.

Dutch technology company ASML is the most valuable company in Europe thanks to its production of extreme ultraviolet lithography (EUV) machines. These are the foundation of the world’s most advanced chip manufacturing and, as such, are one of the few potential choke points in the entire AI industry. ASML has mastered the technology and intricate supply chain to make the $400 million-per-unit machines needed to etch patterns into silicon, where light is focused at a wavelength of 13.5 nanometers (about 1/10,000 the width of a human hair). Without their machines there are no chips; Without chips, there are no more data centers. That means everything from Taiwanese chipmaker TSMC’s $265 billion investment in future manufacturing facilities (known as “fabs”), to the $7 trillion worth of data centers the industry is expected to build globally (though mostly in the US) by 2030. Without more data centers using these current and next-generation chips, AI’s progression along a growth (and risk) curve exponential would stop.

This is why the US government is imposing extraterritorial export controls on ASML, blocking exports of its most advanced EUV machines to China in an effort to curb its advanced chip manufacturing capabilities. By imposing such restrictions on ASML, the United States has revealed its own vulnerability to this bottleneck: as much as Europe depends on (ultimately replicable) financial and software services from the United States, the future of all chip-based technology depends entirely on ASML. An outright ban on exports of ASML machines would force the global AI industry to tread water, to forge ahead.

The resulting chip shortage and slowdown in AI development would give regulators time to catch up with the technology, give lawmakers time to update everything from copyright laws to tax laws, and give societies time to have real debates about the role we want AI to play in our lives.

Why would the EU do such a thing? Not only for humanistic reasons, but also selfish ones. At best, an outright ban on exports would lead to a reorganization of the supply chain and subject more crucial decisions to the jurisdiction of the EU, where it would be subject to regulatory oversight of the only actor that has demonstrated this. willpower broadly regulate AI. Of course, the best-case scenario is far-fetched: the United States is more likely to retaliate by blocking ASML’s access to US-based suppliers. But even if this were to happen, it would still be in the EU’s interest to provoke such a confrontation, based on the narrow logic of pure economic and geopolitical self-interest, in the sense that the EU would stop falling behind in relative terms in AI models and the “computation” (processing power) of the data centers needed to run them.

Europe’s own AI proponents agree, although what worries them is the relative disadvantage of the EU, rather than the vast panoply of AI risks. The Europe 2031 project is a techno-political forecast – compiled by a group of industry experts – of what could happen to the EU in a scenario where AI radically transforms the economy and the EU falls far behind the United States. They argue that Europe must invest hundreds of billions in the AI arms race before both frontier models and “computing” become what TSMC and ASML are today: beneficiaries of a gap too big for any competitor to close.

In their scenario, at some point in the near future the EU will be forced to put ASML access on the table to ensure access to the US-controlled models and calculations on which its economy and security will depend, only to find that ASML’s supremacy has weakened over time. (The more EUV machines are exported, the more factories will be built elsewhere and the more time China and the United States will have to replicate ASML’s technology.)

Nuclear weapons theorists worry that if the unspeakable were to happen, states would be forced to toward release due to a “use it or lose it” calculation. Exactly the same calculation applies to the only thing that can stop the rapid development of the AI industry: the EU can use the leverage that ASML currently gives it or risk missing the opportunity entirely. It is rare that disparate groups come to a similar conclusion: there is a single European card to play and that must be played. Instead of running towards madness, let’s stop the race. Right now and for everyone.