You cannot achieve AI sovereignty through regulation alone.

      For years, critics of Europe’s AI approach have consistently pointed out the same issue: while the United States focuses on developing artificial intelligence, Europe is busy creating regulations for it. The EU AI Act stands as the clearest illustration of this divide—a significant attempt to regulate AI that many viewed as evidence that Europe prioritized rules over innovation.

      Recent developments indicate that this discussion may be evolving. In June, U.S. restrictions temporarily barred access to Anthropic’s advanced Fable 5 model, causing disruptions for businesses and governments across Europe until the ban was lifted just weeks later. Simultaneously, OpenAI faced increasing pressure to restrict access to some of its most sophisticated features, and Anthropic reportedly dealt with political scrutiny regarding its limitations on the use of its models for autonomous weapons and mass surveillance.

      Earlier this year, the Trump administration allegedly threatened to reconsider government contracts with Anthropic after the company declined to permit its models for certain military and surveillance purposes. Whether these events are short-lived or long-lasting, they indicate a larger shift: frontier AI is no longer viewed merely as commercial software; it is evolving into a strategic infrastructure.

      This context helps explain why Austria reportedly encouraged the European Union to consider attracting Anthropic to establish a stronger presence in Europe. The proposal flipped one of the continent’s major criticisms on its head. For years, Europe’s regulatory framework was seen as a deterrent for frontier AI companies. Now, legal predictability, regulatory stability, and the rule of law are being highlighted as competitive strengths.

      The irony is striking. The very AI Act that was criticized for potentially driving innovation away may, in fact, become increasingly beneficial for attracting AI firms. However, there remains a deeper inquiry lurking beneath the surface: even if Europe succeeds in becoming a safe haven for frontier AI companies, how secure can that haven truly be?

      A company relocating to Europe cannot simply sever ties with the ecosystem essential for frontier AI to function and grow. It still relies on advanced semiconductors, hyperscale cloud providers, vast computing power, and affordable electricity—elements that are not peripheral but central to the infrastructure of modern AI. These resources remain predominantly concentrated in the United States or dominated by American companies.

      This reliance is significant as it highlights the limitations of what Europe can actually provide. A company may shift its offices and employees, yet it maintains a deep reliance on an American ecosystem composed of cloud providers, semiconductor supply chains, and computation infrastructure, as well as the energy systems that support them.

      Furthermore, relocating a business does not remove potential political influence. Governments do not need to control every aspect of the AI value chain to impact its trajectory; they merely need to control a critical link. By restricting access to advanced chips, cloud infrastructure, computing capacity, or other essential necessities, the entire strategy can begin to unravel.

      In essence, relocating a business does not necessarily displace its dependencies. This highlights a broader vulnerability in Europe’s AI approach. For more than twenty years, Europe has exerted influence through what legal scholar Anu Bradford famously termed the “Brussels Effect”: the ability to shape global markets by establishing compliance rules for companies.

      This strategy was particularly effective in areas such as privacy, consumer protection, and competition policy because access to Europe’s market was a substantial incentive. However, artificial intelligence disrupts this complex equation. Unlike past digital innovations, frontier AI is constrained by sophisticated, hard-to-replicate physical infrastructure.

      High computing capacity, advanced semiconductors, massive data centers, plentiful electricity, and significant capital reserves are now the critical factors of production. These elements cannot simply be created through regulation.

      The statistics underline the scale of the challenge. Presently, the U.S. controls approximately three-quarters of the world’s frontier AI computing resources, with Europe comprising only a small portion. American hyperscalers also account for around 70% of Europe’s cloud market, while private AI investment in the U.S. is nearly ten times that of the European Union.

      Additionally, the U.S. consumes nearly half of the global data-center electricity and continues to expand its AI infrastructure at a speed that Europe has yet to achieve. Leadership in AI is increasingly defined not only by who creates the rules but also by who owns the infrastructure.

      This is not the first instance in which economic power has shifted due to control over production factors. During the Industrial Revolution, governments could regulate trade, yet true economic authority rested with those who controlled coal, steel, railroads, and factories. Similarly, the AI revolution is instigating a comparable shift, with today’s strategic assets encompassing GPUs in place of steam engines, hyperscale data centers instead of railroads, and massive electricity supplies instead of coal.

      None of this diminishes the significance of the AI Act. Recent events suggest that Europe’s focus on legal certainty may emerge as a true competitive advantage. In an era of increasing geopolitical unpredictability, reliable institutions are crucial. However, predictability is just one part of a much more intricate equation. The principal risk for