Radio ExpressTV
Live
Europe Cannot Avoid Solving the AI Dilemmas
Sonya Muzikarova: Former economic expert at the European Central Bank and a former diplomat at the OECD, she is a Senior Fellow at the Atlantic Council.
When it comes to artificial intelligence, Europe’s biggest challenges are not the sudden arrival of advanced foreign models or the spread of American and Chinese platforms within its markets. Rather, the challenge lies in the political economy governing AI, which fundamentally reflects Europe’s own shortcomings: industrial scalability, computational power (data centers and chips), and a genuine unified market that facilitates strategic expansion.
These shortcomings can no longer be ignored, as the aim of U.S. policy is shifting from “managing China” to “surpassing everyone” globally. Alongside its export control systems, the recent 25% tariff on a select group of advanced AI chips under the Trump administration is designed to attract more investments into domestic semiconductor production, making local manufacturing in the U.S. more competitive while accelerating the development of its AI infrastructure.
This tariff expresses a broader strategy in managing AI that has been taking shape for years, challenging three long-held assumptions: that the U.S. would prioritize market efficiency over industrial policy; that China would import computing capabilities rather than build its own; and that Europe could regulate the industry without developing its own sovereign capabilities.
The United States has surpassed the belief that markets will optimize supply chains. It imposes export controls alongside subsidies, tax incentives, and procurement policies to reshape where chips are designed, manufactured, and deployed. Meanwhile, China is pushing to launch local “AI accelerators” (new chips), enhance its manufacturing capabilities, and link its core AI infrastructure to external lending and economic diplomacy. In contrast, Europe has primarily treated AI as a secondary issue, working to refine legal definitions while remaining dependent on foreign clouds, chips, and models.
Entering the AI era under excessive regulation and insufficient manufacturing, Europe now imports the vast majority of advanced semiconductors, pays significantly higher industrial electricity prices compared to the U.S., and continues to rely on American cloud service providers for most computing needs. If this dangerous dependency was not clear before, it has become increasingly difficult to ignore in light of explicit U.S. threats to assert control over the sovereign territories of a long-standing European ally.
Europe now finds itself caught between an aggressive, disruptive power (Russia) that is already testing its defenses and a U.S. administration willing to use its ties in industry, infrastructure, and trade with the continent as leverage. If the U.S. exploits its ability to access AI and advanced computing coercively, the effects could be immediate: European defense networks, intelligence systems, hospitals, financial markets, and industrial companies may face sudden restrictions on vital cloud services, with few local alternatives. In this scenario, the Kremlin could escalate its hybrid war against Europe, knowing the continent is digitally vulnerable and politically constrained.
Given these risks, Europe must move beyond its focus on regulatory excellence, risk ratings, and compliance systems. Without making serious progress in building the physical and financial infrastructure required for a domestic AI industry, these concerns will emerge as more of a burden than an advantage. Specifically, European policymakers must support the creation of large computing clusters, ensure a reliable supply of affordable electricity, and commit to sustained capital expenditure in strategic sectors.
The bad news here is that Europe cannot change its course overnight. The reality is that the cost of building a single advanced data center could easily exceed one billion euros ($1.2 billion) and consume more energy than that used by a medium-sized European city. Moreover, building a sophisticated semiconductor facility now requires over 20 billion euros in initial capital expenditures. Yet energy prices in Europe are already high, venture capital markets are shallow, foreign companies dominate cloud infrastructure, and its semiconductor goals remain largely ambitious. Recent estimates suggest that updating Europe’s AI industry will require approximately 3 trillion euros in investments over the next five years.
The good news is that Europe does not start from scratch. It holds a significant share of vital technologies. For instance, Dutch company ASML dominates the ultraviolet photolithography field, supporting the most advanced production lines for TSMC and Samsung. Similarly, German and Dutch suppliers such as Zeiss (optics) and Trumpf (high-power lasers) hold strategic positions in the AI production chain. Collectively, these local contracts provide the EU with the means necessary to establish parts of the global AI hardware ecosystem in Europe.
After computing, capital is the most scarce input in Europe’s AI race. However, unlike computing, capital can quickly shift in response to political signals and incentives. Despite Europe’s lag in comprehensive technology and AI funding — American AI companies attracted around $47 billion in 2024, whereas European firms garnered approximately $11 billion — high-tech startups have produced more than those in the U.S. between 2019 and 2024 (albeit the value of deals did not exceed $62 billion in 2024 compared to about $209 billion in the U.S.).
Furthermore, funding earmarked for AI has already seen growth from a low base, as European companies raised nearly 3 billion euros across 137 deals in 2024, marking an increase of nearly 35% compared to the previous year. Investment in European defense and security technologies has reached record levels, reflecting a reassessment of the strategic industries on the continent. This momentum partially reflects a gradual shift in directing private capital toward Europe, as investors respond to U.S. political uncertainty and seek long-term exposure to strategic infrastructure and industrial assets in Europe.
The European Union must start playing aggressively as well. This means leveraging the strength of the European market by tying access to markets, procurement, and regulatory approvals to concrete local commitments — local mobilization, constructing data centers, assembly, or research and development projects — just as the U.S. has done under the CHIPS and Science Act with TSMC in Arizona. Europe must also actively mobilize long-term capital through public guarantees and blended financing, enabling pension funds, insurance companies, and sovereign entities to secure semiconductor factories and computing clusters (investments that venture capital will likely avoid).
Finally, Europe needs to treat energy, computing, and data center deployment as a single planning challenge, rather than three separate issues. It should work to align pricing, licensing, and infrastructure so that factories and computing clusters have reliable electricity, purchase agreements, and predictable locations. Encouragingly, ensuring such coherence falls within the capacity of European institutions.
The lesson from recent changes in U.S. policy is not that Europe should eliminate regulatory controls, but that regulation without the means, computing power, and capital leaves it vulnerable in a highly competitive world. Europe can still catch up, but only if it begins building the capacities that make regulation meaningful.
