For seven years, US journalist Karen Hao investigated the world surrounding OpenAI and other industry giants. In her book Empire of AI, she draws a radical parallel: the structures of Silicon Valley resemble those of the former colonial powers. They lay claim to land, water, energy, data, and intellectual property. They concentrate knowledge, capital, and infrastructure in the hands of a few. And like the colonisers of past centuries, they package their extraction in a moral narrative of progress. “If we continue down this path and we return to an age of empire,” Karen Hao warned during a parliamentary breakfast event in May, “democracy will not survive.”

Alongside the event at the German Bundestag – hosted by digital policy expert Rebecca Lenhard of Alliance 90/The Greens – Karen Hao also contributed her in-depth analysis to another event within the SCRIPTS Forum “Emergent Digital (Dis-)Orders”: an evening discussion at the DGAP villa near Tiergarten as part of the “Remaking America” series, together with Professor Anita Gohdes.

At both events, the audience was quickly confronted with the same seemingly sobering question: what can Europe realistically still do? A continent lagging far behind in the global AI race, accounting for barely five per cent of worldwide computing capacity, and simply lacking the astronomical financial resources of either the United States or China.

Europe should not even try to compete in the global race, Hao argued bluntly. In Silicon Valley, the dominant dogma is one of maximum scale: ever more data, ever more computing power, ever larger models. Between GPT-2 and GPT-4 alone, the computing power deployed increased by a factor of 10,000. The reason is simple: it is easier to throw unimaginable amounts of computational power at a problem than to develop leaner, smarter methods. Anyone today making a simple AI query that prompts the system to churn through the entire internet in the background is, in sustainability terms, behaving roughly as if they were taking a space rocket to the bakery.

“Bicycles of AI”

If Europe were to embark on a blunt catch-up race, European taxpayers’ money would ultimately finance new infrastructure from which US corporations, cloud providers, and chip manufacturers would benefit most. “But there’s plenty of research that shows that you don’t actually need scale to reach comparable performance,” Hao explained further. “You can also simply use different methods to get these same exact capabilities.”

If Silicon Valley is building the rocket, Hao concluded, Europe should focus on developing the “bicycles of AI”. Specialised, energy-efficient, and trustworthy applications could pursue concrete societal goals such as cancer detection, pharmaceutical research, climate protection, or education. Such systems would not only be cheaper and more resource-efficient – a clear, values-based vision could also attract top global talent.

That the market may be ready for such a shift is reflected in growing resistance in the homeland of AI itself. In the United States, 80 per cent of people are now sceptical of the technology. Grassroots initiatives are protesting against the enormous water and electricity consumption of new data centres, creatives are filing lawsuits over the unlawful exploitation of their intellectual property, and whistleblowers from within the tech companies are exposing abuses through open letters.

The experts in the Berlin discussions made clear just how profoundly AI is shaping the global order. During the parliamentary breakfast, Dr Katya Munoz expanded the economic perspective by adding a security dimension: digital sovereignty is not merely a question of economic policy, but of national security. If Europe neither owns nor controls its critical infrastructure, then in a crisis – such as a blackout or a coordinated wave of disinformation – it may not even be able to determine whether a disruption is technical, commercial, or political in nature.

That evening at the DGAP, Professor Anita Gohdes, whose research focuses on digital repression and authoritarianism, sharpened the audience’s awareness of the industry’s geopolitical myths. In Silicon Valley, the so-called “China card” is frequently played: “Don’t regulate us, or Beijing will win.” According to Karen Hao, however, this argument is increasingly losing its force. Developers in Silicon Valley themselves are now making widespread use of Chinese open-source models because they are often cheaper and more efficient than ChatGPT and similar systems.

Both Berlin discussions highlighted how deeply politicised the issue of AI has become. The debate is about data centres driving up local electricity prices and draining communities’ water supplies. It is about copyright law and precarious click-work. Above all, however, it is about the ethical question of why democratic societies should unconditionally embrace a technology whose infrastructure and decision-making logics largely escape democratic oversight.