To gear up for your return, try Le Grand Continent. The ideas and analyses you won’t find anywhere else, but you’ll soon see them everywhere. Delivered to your inbox from 8 euros per month
On June 12, 2026, the White House imposed export restrictions that drastically curtailed foreign access to the most advanced artificial intelligence models from the giant Anthropic. Overnight, Europe lost access to one of the world’s most powerful AI systems, on which many of its companies were already beginning to rely. It only took a sudden decision by the Trump administration to reveal the staggering cost of Europe’s dependence on this American frontier technology. Access has since been restored, but the incident offered yet another proof that the United States is neither reliable as an ally nor as a partner, and it made Europe’s urgent aspiration for greater technological sovereignty—now a central element of the continent’s competitiveness and security—even more pressing.
Yet the debate on this issue is often marked by unrealistic and unattainable aspirations. Europe cannot aim for total technological sovereignty, but rather for selective sovereignty. The strategic challenge is therefore not to eliminate all dependencies, but to distinguish those that are most critical, those that can be reduced, those that must be mitigated, and those that must be managed astutely so that Europe can defend its interests and its values.
The Emergence of a Categorical Imperative
The European quest for strategic autonomy, including in the realm of technological sovereignty, predates the AI explosion or Anthropic’s disruption. Concerns about Europe’s excessive reliance on the United States and China have grown over the past decade. While strategic autonomy began as a slogan promoted mainly by France — which other European capitals dismissed as Parisian dirigisme — the importance of this political objective now enjoys broad consensus among Europe’s leaders. When Emmanuel Macron highlighted its importance at the AI Action Summit in Paris in February 2025, he no longer needed to persuade his counterparts. The only question was how long it would take to achieve it.
This ambition has become all the more urgent as Europeans are squeezed between the two powers that dominate the AI technology frontier. Across key indicators, Beijing and Washington are rightfully seen as the undisputed leaders, spanning the entire technology stack—from AI models to semiconductors, cloud services, and data centers. They concentrate most of the talent and attract enormous sums of investment. Europe, far behind, is particularly vulnerable. Yet its companies do possess certain strengths, notably world-class robotics skills. ASML—the Dutch maker of sophisticated lithography machines used in cutting-edge chip manufacturing—is thus one of the most strategic players in the AI race. But the gap between European capabilities and those of the United States or China is already too large and continues to widen.
Because of these relative weaknesses, Europe is generally portrayed, at best, as a spectator and, at worst, as a victim. European governments and some firms have long chosen to ignore this innovation lag, relying on American technologies without sustained efforts to reduce dependence. For years, the Union bet on its expansive regulatory powers, assuming they would steer technological development in line with its values by regulating for rights protection. But faith in the normative power of the Union has given way to the realization that Europe cannot simply be an arbiter in a world marked by escalating trade and tech wars, deteriorating transatlantic relations, and rising tensions. It must also be an active player, defending its interests with robust offensive and defensive strategies. Sovereign control over key technologies such as AI has become essential to safeguard the continent’s economy, security, and democracy.
In 2024, Mario Draghi’s competitiveness report mobilized minds across Europe. In it he laid out Europe’s existential challenges and showed how a thriving tech sector was vital to regaining competitiveness.
This objective is also crucial for defense, as there can be no security without prosperity. Russia’s invasion of Ukraine and the waning confidence in American security guarantees forced Europeans to rebuild their own military capabilities. This demands substantial resources for economies already grappling with debt-laden public finances and anaemic growth. It is therefore fitting that Draghi’s report placed competitiveness and technological sovereignty at the forefront of the political agenda. The acceleration of the AI race makes this goal even more urgent and decisive.
Why Europe Should Not Seek to Be Totally Sovereign
A complete AI sovereignty would imply total—indeed absolute—control over the entire stack. And yet the stack is not limited to the model layer; it includes sophisticated infrastructure—data centers, cloud services—as well as semiconductors, which are essential to training and deploying state-of-the-art AI models. Such control would also demand substantial energy resources to power data centers, and mastery of the supply of rare earths critical to semiconductors and other key hardware components.
Let us first state the obvious: the AI race exists precisely because neither the United States nor China enjoys total sovereignty over this technology, and it is likely that neither can claim it. AI relies on global, complex supply chains that are impossible to fully control or nationalize. China remains dependent on foreign semiconductors and the equipment needed to fabricate them. The United States has refused to supply the most advanced chips to China to preserve its lead in training the most powerful AI models. In return, Beijing has learned to leverage America’s dependence on rare earths, compelling the Trump administration to ease export controls on semiconductors in exchange for access to these essential minerals. Chips illustrate the scale of these interdependencies: the United States designs them, Taiwan manufactures them, the Netherlands supplies key equipment, Japan provides chemical inputs for their fabrication, Korea contributes memory semiconductors, and China controls the raw materials. Each country could weaponize the chokepoints it controls to influence the AI race. But even by acting forcefully, each will remain vulnerable to being leveraged by other players further along the supply chain at another chokepoint.
If neither the United States nor China can claim total sovereignty over AI, Europe can claim even less. The continent accumulates far more dependencies and has far fewer resources to disentangle itself. It would be dangerous for Europe to pursue a maximalist approach to sovereignty by cutting itself off from American and Chinese technologies. If European companies do not have access to the best technologies—which in AI are often American or Chinese—the adoption of AI in Europe risks slowing down. The real productivity gains from AI depend on how we integrate it across sectors. And many of the productivity gains promised by AI have not yet materialized. It is precisely at this crucial stack level—the application—that Europe currently has no inherent structural disadvantage, provided it preserves reliable access to leading models necessary for transforming its economy.
The objective, therefore, must be selective technological sovereignty. Europe should be strategic while remaining realistic about which dependencies it can accommodate and which it must mitigate risk for, in the short, medium, and long term.
It should prioritize reducing foreign dependencies that are most sensitive to sovereignty and therefore most damaging if used for hostile purposes, especially with regard to essential infrastructure and critical public services. At the same time, it should pursue the ambitious development of intrinsic capabilities across the entire stack, carefully prioritizing its investments.
A Strategy for Middle Powers
Like other powers lacking the resources of the United States and China, Europe faces major strategic choices on how to reduce its tech dependencies. Experts at Chatham House recently proposed a typology of alternative strategies for middle powers: alignment, hedging, pooling, or specialization.
Alignment involves deliberately siding with either the United States or China—and thus accepting dependence on one, but not the other—in the hope of obtaining privileged access and other advantages. Hedging means leaning on a variety of foreign suppliers while selectively developing domestic capabilities. Pooling aims to build partnerships with like-minded countries to amplify the collective influence of all participants. Finally, specialization entails identifying narrow segments of the global AI supply chain where the country can develop autonomous capabilities, or even become a global leader, thereby strengthening its bargaining power.
If these four strategies are presented as exclusive, Europe is well placed to combine them: alignment and hedging have clear limitations, but pooling and specialization could be productively leveraged.
Risky Alignments
Europe should first avoid becoming overly dependent on the United States or China. In the past, aligning with Washington could have been a plausible strategy, but it has never been optimal. For decades Europeans relied on many American technologies, reassured by close economic, political, and military ties underpinning the transatlantic partnership. Under Donald Trump’s second term, the United States became an increasingly unpredictable and unreliable partner: tariffs, attacks on laws and democratic institutions, threats to Greenland, and withdrawal of security guarantees. This turn of events forced many Europeans to conclude that alignment with a power increasingly hostile to the continent would be not only imprudent but dangerous. The recent Anthropic outage episode shows that the United States is capable and willing to deprive Europeans of access to American technologies without prior warning or negotiation.
At the same time, rapprochement with China is hardly an attractive alternative. Beijing openly seeks to profit from the United States’ unpredictability by presenting itself as a more stable and reliable supplier of technology. But China has shown it is willing to use coercive economic measures. Its grip on essential raw materials gives it significant power and leverage. It has already used this by imposing export controls, limiting Europe’s access to rare earths—critical for the continent’s automotive, green, digital, and defense industries. Finally, the People’s Republic remains an authoritarian regime, raising further questions about Europe’s dependence on actors whose political model differs so fundamentally.
Hedging: A Strategy of Diversification
Europe could instead adopt a hedging strategy between the United States and China, using their technologies selectively while deliberately avoiding dependence on a single supply source. This would allow it to play the superpowers off against each other.
When negotiating access to American and Chinese technologies, Europe should use its vast market as leverage. China and the United States both desire and need access to Europe’s more than 450 million affluent consumers. European firms and consumers thus represent a major revenue stream for American and Chinese AI developers, strongly incentivizing them to serve this market. This access could, for example, be used to negotiate a “trusted partner” status that would preserve Europe’s access to frontier American AI. Simultaneously, Europe could negotiate technology transfer agreements as a precondition for allowing Chinese tech firms to invest or operate in Europe. The aim of this hedging strategy is to avoid excessive dependence on either side while strengthening Europe’s bargaining position.
Pooling Capabilities Through New Partnerships
Pooling capabilities with other middle powers would further strengthen Europe’s bargaining power against the United States and China.
Europe has many allies and partners with whom to collaborate. A closer partnership with like-minded countries such as Australia, Canada, South Korea, and Japan would amplify the collective weight of a coalition of willing players. None of these middle powers wishes to align fully with either the United States or China. They know that individually they could become targets of economic coercion by one side or the other. Not only do they face common threats, but they also share values that enable them to forge trust-based cooperation in their mutual interests. Moreover, these countries possess key technologies that could be shared to reduce their individual vulnerabilities—whether access to energy, semiconductors, open-source AI models, or research and public markets. Such collaboration would allow them to pool resources, gain economies of scale, and acquire complementary capabilities, while offering greater resilience to coercive strategies from Beijing and Washington.
Specializing Europe: From Industrial AI to R&D and Defense
The Union should identify areas of specialization to become a leading industrial player and a pivotal partner in certain segments of the AI value chain.
Companies like ASML already demonstrate that a European enterprise can play a crucial role in pushing the frontier. Europe should now proactively seek other areas where its advantages could be developed, extended, and converted into global leadership positions. These areas should receive additional investments through targeted industrial policy, but also by easing private capital inflows. Beyond these offensive strategies, Europe should safeguard these sectors against acquisitions by external actors, by limiting foreign direct investment and even blocking certain acquisitions by non-European players in strategic sectors to ensure that these capabilities remain in Europe. Ownership of these key assets provides Europe with obvious economic gains, but above all a strategic edge to resist coercion and negotiate preferential access to AI value chain segments it does not control.
Industrial AI is a natural domain for Europe’s specialization. Unlike consumer AI, it yields near-immediate productivity gains in robotics, manufacturing, chemistry and materials, logistics, healthcare, energy systems, and industrial operations—precisely the sectors where Europe remains globally competitive. While attention until recently focused on language models and agents, industrial AI serving the real economy offers vast opportunities that Europe must seize.
In this area, the continent already benefits from a strong industrial tradition with global manufacturing giants such as Siemens, Schneider Electric, Bosch, BASF, and Airbus. It remains a leading actor in sectors requiring precision engineering, including aerospace components, industrial robots, and medical devices. Europe’s deep engineering expertise and its vast reservoir of high-quality industrial data could be leveraged to achieve breakthroughs in industrial AI. While China has made significant strides in applying AI to physical industries and achieved impressive advances in sectors such as robotics, it has not yet won the entire competition. We would be mistaken to abandon the race for “embodied AI” to China. Instead, we should ambitiously integrate AI into sectors where Europe already has concrete industry players and globally recognized expertise. In this regard, there are now more AI-focused industrial startups in Europe than in the United States, indicating that we have already begun to reap these advantages. These firms must now be able to grow in Europe and become world leaders.
The scientific use of AI, which integrates this technology into R&D processes, offers another promising avenue for Europe’s global leadership in AI. Europe’s excellence in research aligns with breakthroughs in chemistry, biology, and pharmaceuticals. In particular, by leveraging a highly developed regulatory framework, Europe could focus on developing AI solutions for tightly regulated sectors that require very reliable technology. Thus, in areas such as banking and insurance, compliance and the explainability and verifiability of AI models and applications are crucial. European firms have long operated within this constraint framework and could lead the way by using AI to develop solutions for efficient and reliable compliance processes.
Finally, geopolitical shifts have spurred significant investments in European defense firms. AI-driven physical technologies such as drones could form a privileged specialization, as could AI-enabled systems for intelligence gathering and battlefield planning.
Overall, Europe does indeed have strategic options. But its pursuit of AI sovereignty will require a blend of hedging, pooling, and specialization. The right mix of these strategies can reduce Europe’s vulnerabilities by managing existing dependencies while developing its own AI capabilities in parallel.
Taking Advantage of American Mistakes
To address the strategic dilemmas it faces, Europe needs a clear view of its strengths and weaknesses. It must urgently tackle structural barriers to growth, foremost among them market fragmentation. The Union can selectively draw on certain American and Chinese successes—such as the successful integration of venture-capital markets in the United States and China’s focus on industrial AI applications—yet it must also have the confidence to develop AI in a way that aligns with Europe’s values and strategic priorities.
The EU’s current chief weakness lies in the fragmentation of its market. Despite decades of legislative effort, the single market remains incomplete, hindering European firms from growing and fully leveraging this vast economic space of 450 million consumers. This is especially true for digital services. Too preoccupied with tariffs imposed by Donald Trump, the Union sometimes forgets that its own market is its largest partner. It must urgently complete the integration of capital markets so that AI startups and other European firms can fund growth. If these reforms are implemented boldly, they offer the most promising path to greater technological sovereignty.
Beyond fragmentation, this virtuous trajectory is often seen as compromised by two other real weaknesses. First, a lack of capital to fund AI infrastructure needs would limit Europe’s ability to build a top-tier ecosystem. From this perspective, Europe’s failure to exploit the Internet revolution continues to haunt a continent without tech giants generating substantial revenue to finance infrastructure. Second, Europe’s propensity to regulate technology is often seen as a brake on European innovation. Critics of the European regulatory state often point out that American companies are not subject to such regulations, leaving them room to innovate.
An Efficient AI
Nevertheless, these two perceived weaknesses can also become advantages for Europe. First, if Europe must grow its AI infrastructure investments and build data centers, it should not attempt to match the mass-scale development of U.S. infrastructure. The leading Silicon Valley hyperscalers are expected to invest around $800 billion in AI infrastructure just in 2026, as they race to keep the United States ahead in frontier AI. Much of these investments go toward building data centers where the most powerful models are trained. But this energy- and capital-intensive expansion model is meeting growing resistance in the United States and elsewhere. Public sentiment toward AI in the United States has turned against it, and there is reluctance to host gigantic data centers within communities. These data centers consume enormous amounts of energy, sometimes saturating the power grid and driving up energy prices for consumers already facing inflation and a cost-of-living crisis. In response to this public sentiment shift, many data-center projects have been reevaluated, paused, or canceled. Beyond environmental effects and energy prices, AI users and some companies are beginning to question the costs of relying on high-capital, resource-intensive models. With the value proposition of AI still limited, many businesses are turning to Chinese AI, which is cheaper and “sufficiently capable,” recognizing that most tasks do not require top-tier costly models and can be performed with more resource-efficient AI.
This market reaction, combined with public opinion, offers Europe a valuable lesson. The vast wave of American investment may not bear fruit—and could even waste substantial resources. Growing anti-AI sentiment could also curb American ambitions in this field; Europe could still avoid this if it deploys AI more prudently and with regard to the common good. The best course, therefore, is to invest cautiously, aiming to create AI models that are energy- and capital-efficient, which, while not matching frontier AI capabilities, meet growing demand both domestically and abroad.
Governing Technology and Establishing Trust
A second lesson for Europe is that it must not forego its propensity to regulate AI. American technological dominance is not explained by attachment to techno-libertarianism or unregulated AI. On the contrary, American success stems from its thriving tech ecosystem—bold entrepreneurial culture, a large domestic market, highly integrated capital markets, and access to global talent. These fundamental characteristics have enabled U.S. tech firms to reach global leadership—an ensemble of qualities that the Union has not yet managed to replicate. China also regulates AI and has demonstrated how regulation and innovation can coexist. This suggests that Europe’s regulatory affinity is not the reason the continent has fallen behind the United States and China in the tech race.
Moreover, as AI model capabilities advance, it is neither strategically desirable nor politically feasible to grant AI-sector firms unlimited power to steer the evolution of this technology. Public resentment and mistrust of AI show that we need predictable and protective governance. Regulatory pressures are intensifying in the United States as well. AI populism is gaining ground, which could limit the development and adoption of this technology in the United States. Large American firms are increasingly recognizing that they cannot continue to develop this technology without societal buy-in. It is also telling that 145 AI-related bills were enacted at the U.S. state level last year.
Regulation builds trust—and trust in AI is essential for society to broadly adopt this technology. Trust is itself a strategic capability. Far from being a brake on Europe’s AI ambitions, a stable regulatory framework is essential to realizing them. Better regulation can thus facilitate AI adoption in Europe and spur the economic growth the continent needs. It can also serve as a meaningful comparative advantage for Europe as it seeks to position itself as a leader in trusted AI.
An AI Race Europe Can Win
Those who seek full, total technological sovereignty for Europe are as unconstructive as those who resign themselves to vassalage. The uncomfortable truth is that Europe will not achieve this sovereignty in the near future, if ever. The earlier Europeans acknowledge this reality, the sooner they can begin crafting realistic and effective strategies that make the continent less dependent and more resilient. Europe cannot eliminate all dependencies, but it must manage them wisely.
Recognizing that technological sovereignty is not within immediate reach does not mean we are out of options. Europe is a wealthy continent, brimming with immense talent and global companies. It has a vast internal market, stable institutions, and valuable allies and partners around the world. What Europe needs today is a new mindset. We should not behave as if we are “already convinced of our own decline,” bracing for defeat as an inevitable outcome of the AI race. Instead of such resignation, Europe must summon the courage, confidence, and conviction to rely on its many strengths, playing not only defense but also offense in this race. Such renewed ambition will allow it to capitalize on these advantages and become a more autonomous power, capable of writing its own future in AI.
Finally, Europe must decide which AI race it wants to lead. This choice in itself is an exercise in technological sovereignty. European firms are unlikely to be the first to reach general artificial intelligence frontier or to persuade the world to adopt a European AI stack. Yet Europe can still win the race that matters most: the wide-scale deployment of AI across its economy and public institutions, to boost Europe’s prosperity, strengthen its security, and preserve its democracy, while managing the risks this technology poses for its citizens. To succeed, political leaders will have to make tough compromises between competing policy objectives. But if they approach these compromises with pragmatic ambition, they can deliver to Europe a form of selective and meaningful technological sovereignty.