All summer long, Le Grand Continent will stay in motion. Each day, wherever you are, we will bring you ideas you won’t find anywhere else (but will be everywhere come the new season), texts that are hard to find or refreshing with our Sundays. To receive them directly in your inbox and support this momentum, consider subscribing to the magazine
Europe has assumed the title of the world’s “super-regulator” in the digital realm, positioning itself as a pioneer and leading force in the field, as evidenced by the famous “Brussels effect.” The approach is well known: it involves regulating the practices of a handful of tech giants without owning or controlling them. Currently, the initial wave of antitrust and regulatory measures aimed at these companies is waning, not only because Europe tends to bow quietly before the Trump administration to avoid trade retaliation, but also because the futility of this process is becoming apparent. But, in the meantime, where does our own technological infrastructure stand? We are massively lagging in artificial intelligence. How did we get here?
The answer — admittedly not flattering to Europe — is that we fell asleep. Or rather, we let ourselves be carried along by a collective refrain: the idea that, step by step in the tedious game of competition and regulation “at a small scale,” we would manage to civilize the digital giants that have invaded the continent. While we labored to teach them manners, these same companies were busy covering Europe with a vast infrastructure that underpins our digital life today: the cloud, data centers, software, and connectivity. We now operate in a context of existential systemic dependencies, at the mercy of an external, complex infrastructure composed of several levels (hardware, software, architectures, data environments, authorizations), of which we are neither owners nor masters, and which nonetheless shapes our economic capacity, our industrial productivity, and the direction of our innovation. We have become a digital colony.
It is a major failure to understand the true nature of the power that unfolded across Europe. We thought it was a misstep: it was actually a matter of physical infrastructures. They are not illegal in themselves: the American giants simply occupied the space left vacant by Europeans, too busy “regulating” rather than “building.” We have thus de facto supported the American digital industrial policy, while failing to see that the situation was critical for Europe in the same domain. We thought all we had to do was contain the situation by “enforcing the law” and a few codes of conduct, whereas building our own digital infrastructure would have required a robust and proactive industrial policy. We still think that “industrial policy is separate from digital policy,” and the Commission ties their management to distinct, siloed competences. Worse still: while it is now obvious that our failure stems from our inability to show initiative and call on our capacities, some Europeans claim that Europe is a “mid-power,” and that, as such, our only hope is to ally with other “mid-powers” such as Canada, India, Japan or Brazil. We would have no chance otherwise.
How could this regulatory mindset, so characteristic of Europe, have so entirely undermined our sense of autonomy and our future prospects? There is much to be done: it is imperative that Europe take control of a larger share of its technological infrastructure. It is not about protectionism or autarky, as some detractors suggest. Yet Europe is a European economic superpower that has all the means, funds, and talent needed to do better.
Europe’s obsession with “good conduct” has served as a pretext for the alienation of our infrastructures
The giants of the digital world, which have established themselves in Europe across research, social networks, app stores, e-commerce, software, etc., have faced legal complaints from smaller European websites unhappy about not receiving equitable distribution and visibility. As always when the balance of power is highly asymmetrical, the conditions were unfair: self-preferencing practices (favoring one’s own version of a service), exploitation, theft of ideas from newcomers, and excessive pricing. All of this was perceived in Europe through the familiar legal prism of “abuse of dominance”: companies with substantial market power that profit from an inequitable situation must learn to behave properly. Considerable intellectual effort was devoted to explaining how such behavior could be characterized as anti-competitive.
We now operate in a context of existential systemic dependencies, at the mercy of an external, complex infrastructure composed of several levels, of which we are neither owners nor masters.
Cristina Caffarra
We developed “economic theories of harm” explaining how it was “rational and profitable” for these firms to marginalize, exclude, or exploit their competitors. The European Commission embarked on a slow process of opening dossiers, gathering evidence, and formulating “theories of harm,” with years of back-and-forth, information requests, deadline extensions, hearings, decisions, appeals, judgments, and more appeals. All this lasted at least a decade. Then we turned to regulation, proposing to apply the same antitrust rules as codes of conduct (“You shall not favor yourself”), assuming that setting the rules of the game would be easier than proving fault a posteriori. This enterprise looked colossal. Regulators wore “bicep emojis” — “We tame them!” — amplified by a small cohort of academics and civil-society actors who saw Europe’s resistance to “digital monopolies.” Except that none of the giants we sought to tame took these measures very seriously: with endless appeal possibilities and infinite resources, the process can stretch indefinitely. Overall, these companies won even: none of these cases managed to move the needle, not even a little, to alter the business models behind these practices.
In reality, it was like trying to put out a fire with a glass of water. While Europe’s intellectual energy felt reassured by these good-conduct cases (“We are doing something!”), the targeted companies were quietly rubbing their hands: “What’s going on in Europe? We’re glad you’re dealing with these matters, but what are your real assets? A few websites? Keep pushing with your regulations: in the meantime, we will build everything ourselves.”
That is precisely what they did. Now, our digital life is governed by a complex structure of components, from software to data, including apps, chips, architectures, control devices, as well as authorizations and authentications, of which we hold no keys: Europe has become a digital vassal. It is therefore not surprising that we lag behind the United States and China in artificial intelligence, a term now used by almost all European policymakers. But Europe is not lagging there by accident. It is not a sudden or force-major phenomenon. It is something we have caused ourselves.
Without being useless, regulation is a tool largely insufficient to meet our needs
The vocabulary of antitrust and regulation, with its “guardians” and its “self-preferencing,” was not the right one. The problem is far more fundamental: we have outsourced a large portion of our cloud, our operating systems, our app distribution, our corporate software, our search engines, our advertising infrastructure, our communications, our data storage, and increasingly our AI infrastructure, to a handful of foreign private companies. We could not imagine acting the same way toward other infrastructures, such as energy, banking, defense, or telecommunications. Yet we have normalized it in the digital domain, because each step presented itself as a product choice rather than a transfer of strategic power.
It’s not that antitrust legislation and regulation are wholly useless: some litigation was necessary and intellectually important. For example, it was important to understand how Google leveraged its power by nudging Android device makers to designate its search engine as the default on new devices, in exchange for the ability to pre-install the Google Play Store, or how its presence on both sides of the ad auctions gave it disproportionate power to skew the ad market in its favor. But this work never matched the consolidation of this dominant position we were beginning to observe. We spent years detailing, item by item, cases of tied selling, self-preferencing, exclusion, and discrimination, while the architecture underpinning this dependency continued to take root deeper. Years were spent trying to explain how conventional antitrust thinking should evolve to account for network effects and the zero-price economy. Finally, years were burned publishing summaries of academic conferences on economic-model variants, market closure, and exclusion.
The tally of all these crusades is rather meager: the proceedings against Google led to findings of infringement in Brussels and in American federal courts, but no remedial measure could have any real impact. For the sake of completeness, here are the other proceedings: the European Commission’s actions against Amazon for favoritism in the “Buy Box” and for the use of sellers’ data ended in amicable settlements and a negligible measure for Amazon; proceedings against Apple likewise ended in amicable settlements, with a few fines and minor obligations that do not affect its business model; the same for Microsoft Teams, where the case ended with a clever agreement on an unbundled pricing grid that does not facilitate market entry. Regulation then intervened, reformulating and redirecting the same rules within the framework of the DMA, but the pace was slow. The European Commission knows it is under the sword of Damocles of tariffs and other retaliatory measures. Only Apple and Google have had a final decision in the past three years, all fairly timid.
The real problem did not lie in the relations between large platforms and small competitors in the market, but in the fact that these players had become, de facto, market infrastructures backed by private capital. If you control the cloud, the operating system, the app store, the default settings, the identity layer, and, increasingly, the AI interface, then insisting on non-discrimination at a particular level of the stack does not change the nature of land ownership.
A Disconcerting Truth
Finally, we are beginning to understand that we must view digital infrastructures through the lens of industrial policy.
In the digital domain, European industrial policy was the same as American industrial policy. Every time a ministry chose Microsoft, a university moved its data to Google, a hospital became dependent on AWS, or a public body organized around a proprietary software stack, European public funds helped finance the expansion, learning effects, data advantages, and switching costs of these same companies that the Commission later deemed too powerful. European firms relying on this infrastructure merely finance the mechanisms by which incumbents perpetuate their dominance. European companies remain dependent on American cloud, American app stores, American distribution, and, eventually, American buyers.
Europe has become a digital vassal.
Cristina Caffarra
A tougher approach could have aimed to enforce structural separation, to limit consolidations among cloud, software, identity-management, and AI sectors, to impose strict interoperability and provider-switch obligations, to require diversification of public procurement, and perhaps even to establish the principle that critical European public institutions cannot become technologically dependent on a single foreign hyperscaler. But, in the political-economic climate of the 2010s, Europe simply did not believe it could seriously challenge the American giants. Since then, the deployment of these infrastructures has been completed.
Breaking the deadlock in competition law and regulation is essential for Europe’s economic future
The failure does not merely lie in insufficient enforcement of regulation, nor even in a flawed conception of it as a whole. It is a deeper misunderstanding of the very nature of the power that operates over digital markets. What matters are the ownership of infrastructures, the control of bottlenecks, access to capital, and the ability to scale. Europe spent fifteen years trying to “civilize” five private digital empires, without doing enough to reduce their grip, develop alternatives, and take away part of their power. At this stage, the objective can no longer be limited to demanding better behavior from these operators: we must deliberately reduce European dependence on them, while building a genuine technological capability of our own.
If this question is existential, it is not merely because we are dependent and exposed to potential geopolitical instrumentalization or strategic use of these services. It is a much deeper problem: the very future of Europe’s economic model is at stake. In his September 2024 competitiveness report, Mario Draghi placed at the heart of his analysis the slowdown in Europe’s productivity growth relative to the United States over the past two decades. The actual magnitude of this gap has sparked lively debates among economists in recent months (between Luis Garicano, Philippe Aghion and Antonin Bergeaud on one side, and Paul Krugman, Benjamin Wolf, Seth Ackerman, Brad De Long and a few others on the other). But whatever stance one takes on the methods and metrics to be used, there is broad agreement that much of this gap reflects differences in creation, adoption, and diffusion of information technologies. The United States’ lead in digital and innovation has, unsurprisingly, benefited America more than Europe, translating into higher wages and profits in the United States. For Europe, owning a larger share of our infrastructures and relying on them would allow us to capture a larger portion of the value we create rather than letting it flow to external actors. It is imperative that Europe reclaim at least part of the infrastructures we have outsourced and do not control. This is not protectionism or autarky: we simply cannot build our future entirely on someone else’s infrastructure.
That now appears clearly in the field of artificial intelligence. Europe notes that the United States is investing colossal sums in training advanced models. We cannot yet say whether this circle of funding will burst, but one thing is certain: we do not currently have a frontier lab. A faction of “AI maximalists” in Europe is so flustered that they are multiplying calls for Europe to bow further to American hyperscalers. They plead for access to these models in exchange for favorable terms to build more American data centers on European soil. This stance is bound to fail. Since it also lacks the resources or the seemingly exceptional momentum of China, which is catching up at an impressive pace, Europe seems to have promised itself to become the champion of “industrial and physical AI.” In the absence of cutting-edge models, it could become the queen of mechanics and machines. Indeed we can do this, but it remains far from a solution to achieve a “sovereign AI,” and the problem will persist: we would have built an industrial AI on someone else’s infrastructure.
What to do?
The difficult situation Europe currently faces requires a determined approach to regard digital infrastructures as nations once regarded railways, electrical networks, aerospace, or defense — a realm where public capabilities, public procurement, capital, and strategic orientation are legitimate and necessary. Let us end the uproar about competition law and regulation: they distract us and waste our energy.
We must instead undertake direct and very significant investments in computing, cloud, chips, data centers, energy, core software, and open models. It would also require governments to become reference customers rather than simply signing checks for subsidies, and for public institutions to deliberately create demand for European suppliers. Public or semi-public investment vehicles should be ready to take long-term equity risk at a scale that traditional European venture capital simply cannot absorb. European institutions have shown they are not up to the task. Time and again, Europeans turn to Brussels with Hope, adopting this infantilizing stance, as if saying: “The Commission must do something.” Europe fails each time because the regulatory impulse inevitably wins: it produces endless consultation documents and bills, without delivering measures fast or bold enough, and it hands out a few euros to initiatives too cautious and too modest to go further. Member states have shown they can act on their own initiative (for example, parts of the public administration in France, Germany, and elsewhere have contracted European providers without waiting for a specific mandate). This is a useful signal, but not enough. Attitudes lag behind aspirations.
Owning a larger share of our infrastructures and relying on them would allow us to capture a larger portion of the value we create.
Cristina Caffarra
There are two areas where private initiative can make a difference. Movements such as EuroStack seek not only to broaden the debate but also to spur new actions.
The key point lies in stimulating demand. As long as the private sector does not turn away from the default solutions offered by the hyperscalers, there will be neither a market nor investment in European solutions. We must overcome this inertia, because the entrenchment of hyperscalers with European companies is deep: procurement choices, investment models, and incentives have long been treated as acceptable compromises; no one cared about the dependencies created by ease of use, which came with short-term efficiency, turnkey infrastructures, integrated service suites, and immediate scalability. Changing course requires careful work and engagement with boards to persuade them to evaluate these dependencies, assess the profitability of alternatives, and orchestrate the transition. The hardest part is that this is a classic “collective action” problem: the benefits of acting together are obvious, but going it alone is seen as a risk. That is why group purchases and “buy clubs” are now appearing.
The second facet involves directing investments toward European solutions. The complex ecosystems of digital infrastructures do not develop solely through occasional injections of capital. They emerge when sustained demand generates revenue streams that justify investment and the development of an ecosystem. This is why demand is essential: it is about reorganizing the European private sector’s procurement decisions (which account for a substantial volume) to create viable markets for sovereign infrastructures. Without these demand signals, no investment program can come to fruition. That said, the enormous reserves of European capital (especially institutional capital: pension funds manage more than €3,000 billion of assets and insurance companies €11,000 billion) must loosen and direct toward growing European companies. Elsewhere, pension funds and insurance companies are more accustomed to this type of investment. In Europe, regulatory constraints remain too numerous and, consequently, institutional investors remain hesitant, citing fiduciary duties and repeating that “pension funds don’t deal with venture capital firms.” Yet there are signs that the “investment window” (what these players consider investable) could slowly shift toward supporting European technologies.
It is clear that imagining a centralized European initiative for digital sovereignty remains difficult. Which institution could realistically lead it? Building infrastructures should undoubtedly fall under industrial policy. Yet the Commission continues to separate industrial policy from digital policy, because “we do not want to tread on anyone’s toes” — instead of working together. We cannot keep waiting, gnawing our nails, hoping that “the Commission will do something”: give us a Capital Markets Union, give us a single market.
Tant que le secteur privé ne se détournera pas des solutions par défaut proposées par les hyperscalers, il n’y aura ni marché ni investissement dans les solutions européennes.
Cristina Caffarra
In the end, Europe is better than its institutions.
We possess about a dozen exceptional technology ecosystems scattered across the continent. While clustering them is valuable, we will never have a single Silicon Valley. Instead, we can have a network. China operates according to this model, and that is a strength, not a weakness. We have extraordinary talent. We also have enormous reserves of capital that must be mobilized.
Antitrust law and regulation have simply proven insufficient; it is time to abandon the illusion that they will deliver something substantial to Europe and to treat the digital realm as our industrial policy, concentrating all our intellectual and positive energy on the construction.
Europe is a “superpower” economy, not a “mid-power”: let us start acting accordingly.