These drawbacks can no longer be ignored because US policy has shifted from the goal of “coping with China” to the goal of “outrunning everyone” in the world. Along with the export control regime, the Trump administration’s latest 25% duty on some advanced AI chips is designed to attract more investment in domestic high-end semiconductor manufacturing in the US. This should improve manufacturing competitiveness and accelerate the development of AI infrastructure in the US.
The new duty is part of a broader AI strategy that has been taking shape for years and refutes three long-held assumptions: that America favors market efficiency over industrial policy; that China will import computing power rather than build it; and that Europe will be able to regulate the industry without having its own sovereign capacity in it.
The US has long since parted with the belief that markets will optimize production chains. They are imposing export controls in addition to subsidies, tax breaks, and government procurement policies to change the geography of chip design, manufacturing, and deployment. Meanwhile, China is introducing domestic “AI gas pedals” (new chips), expanding manufacturing capacity, and tying AI infrastructure to foreign credit and economic diplomacy. Europe, on the other hand, does not yet see AI as a major topic, clarifying legal definitions and continuing to rely on foreign cloud capacity, chips and models.
Europe is “between two fires”
Entering the AI era over-regulated and under-industrialized, Europe imports almost all advanced semiconductors; pays substantially more than the U.S. for the electricity needed by manufacturers; and continues to rely on U.S. cloud providers for the bulk of computing. If this dangerous dependence was not obvious before, it is now hard to ignore because of open U.S. threats to assert control over sovereign territory belonging to a longtime European ally.
Europe is caught between an aggressive revisionist power (Russia), which is already probing its defenses, and a U.S. administration that is ready to weaponize its industrial, infrastructural, and commercial ties to the continent. If the U.S. uses the leverage of access to AI and advanced computing for coercive purposes, the consequences could be immediate: European defense networks, intelligence systems, hospitals, financial markets, and industrial companies would face a sudden restriction of access to critical cloud computing services, with few local alternatives. In this scenario, the Kremlin would have a chance to intensify hybrid warfare against Europe, knowing that the continent is digitally vulnerable and politically constrained.
Given these risks, Europe must shift its focus from high quality regulation, compliance systems and risk classification. Success on these fronts will become a liability rather than an asset unless Europe makes major breakthroughs toward building the physical and financial infrastructure necessary for a localized European AI industry. European authorities should help create large computing clusters, guarantee cheap and reliable power supply, and commit to supporting capital expenditures in strategic industries.
The bad news is that Europe can’t change course overnight. One advanced data center can easily cost more than €1 billion and consume more energy than the average European city. And a state-of-the-art chip factory today requires more than €20 billion in one-off capex. Prices in Europe’s energy sector are already elevated, venture capital markets lack depth, cloud infrastructure is dominated by foreign providers, and semiconductor targets remain mostly dreams. According to recent estimates, approximately €3 trillion in investment will be needed over the next five years to modernize Europe’s AI industry.
The good news is that Europe isn’t starting from scratch. It controls several key technology bottlenecks. For example, the Dutch company ASML has a monopoly on extreme ultraviolet lithography (EUVL), and its equipment is the basis for the most advanced production lines at TSMC and Samsung. In addition, German and Dutch companies including Zeiss (optics) and Trumpf (high-power lasers) occupy important strategic niches in AI production chains. Together, these domestic hubs give the EU the opportunity to anchor parts of the global AI hardware stack in Europe.
Need incentives and capital
In the AI race, the next scarce resource in Europe after computing power is capital. But unlike computing power, capital can move quickly in response to policy signals and incentives. Europe as a whole lags behind in technology and AI funding (U.S. AI companies raised about $47 billion in 2024, while European companies raised about $11 billion), but it created more fintech startups than the U.S. between 2019 and 2024 (however, it only had $62 billion in deals in 2024, while the U.S. had $209 billion).
In addition, funding for AI proper is already rising from its previous low level: in 2024, European firms raised nearly €3 billion in 137 deals, up about 35% from a year earlier. Venture capital investment in European defense and security technologies has risen to record levels, driven by a reassessment of the continent’s strategic industries. This dynamic is partly due to a gradual shift of private capital toward Europe, as investors respond to uncertainty in U.S. policy and seek long-term exposure to strategic infrastructure and industrial assets in Europe.
The EU, too, needs to start playing hardball. This means building on Europe’s market power to impose conditions on market access, government procurement and regulatory licenses. The conditions could be specific localization obligations, including local packaging, construction of data centers, assembly, and R&D – as the U.S. has already done with TSMC in Arizona under the Chip and Science Act.
In addition, Europe should raise long-term capital through government guarantees and blended finance so that pension funds, insurers, and sovereign entities can support chip factories and computing clusters (such investments don’t really attract venture capital).
Finally, Europe needs to treat power, computing, and data center projects as a single task for planning, not three separate tasks. Prices, licenses, infrastructure all need to be coordinated so that chip factories and computing clusters have predictable capacity, guaranteed purchase agreements, and localization. Positively, all of this is well within Europe’s institutional capabilities.
The lesson of the latest policy changes in the U.S. is not that Europe needs to engage in deregulation, but that regulation without equipment, computing power, and capital makes it dangerously vulnerable in a world of wild competition. Europe can still “catch the train,” but only if it begins to build the industry capacity that will make its regulation meaningful.
Sonia Muzikarova
previously worked as an economist at the European Central Bank, as a diplomat at the OECD,
senior advisor to the Deputy Minister of Foreign Affairs of the Slovak Republic,
now a visiting senior fellow at the Atlantic Council.
© Project Syndicate, 2026.
www.project-syndicate.org
