Link to Grand Continent article in French (and Spanish) here: https://legrandcontinent.eu/fr/2026/08/18/leurope-doit-recommencer-a-construire/
English translation (my original writing, no AI anywhere near - my voice) below:
Europe Pursued US Industrial Policies in Digital. Now It Wakes Up: It’s Not Sovereign
Europe adorned itself with the mantle of global “super-regulator” in digital (pioneer and standard-bearer – remember the “Brussels Effect”?) to police the conduct of a handful of digital giants we did not own or control. The initial antitrust and regulatory flurry around these companies is now petering out, not just because Europe is bending the knee to the Trump Administration to avoid reprisals (in trade or elsewhere, so everything is carefully calibrated to drop slowly, and not be too much of an irritant). But also because the futility of it all is starting to loom large: where’s our own tech infrastructure? We are now major laggards in AI: how did that happen?
The catastrophic answer for Europe is that we fell asleep. Or rather, we fell for a collective incantation: that by pushing forward the painfully “small ball” game of antitrust and regulation, one step at a time, we would civilize the digital giants striding the continent. Yet while we were trying to teach them some manners, these very giants were carpeting Europe with the vast digital infrastructure which today supports our digital life: cloud, datacenters, software, connectivity. We are now operating under existential systemic dependencies, at the mercy of a complex layered external infrastructure (hardware, software, architectures, data environments, authorizations) which we do not own nor control, and yet shapes our economic capability, industrial productivity, geopolitical leverage and the direction of our innovation. We are a digital colony.
What happened was a colossal failure to understand the true nature of the power which was being rolled across Europe. We thought it was bad conduct, it was hard infrastructure. This was not illegal per se: US giants filled the space that Europeans vacated entirely, busy “regulating” instead of “building”. What is truly uncomfortable is that in doing so, we de facto seconded US industrial policy in digital – move forth and multiply – while oblivious to Europe’s predicament. We thought all we had to do was to containment by “enforcing the law” and a few rules of behaviour - when developing our own digital infrastructure would have required muscular, decisive industrial policy. Yet we still somehow think that “industrial policy is separate from digital policy” and in Brussels we assign them to separate, siloed competences. And worse yet: at this point, when it is clear the failure has been our own failure to exercise agency and use our capabilities, there are Europeans going around today preaching that Europe is just a “middle power”, and our only hope is to hold hands with other “middle powers” (Canada, India, Japan, Brazil) because we do not stand a chance.
How did this happen? How has the regulatory mindset Europe unfortunately inhabits so undermined our own sense of agency and possibilities? There is a lot to do, but ownership and control of a greater share of our own tech infrastructure are imperatives for Europe. This is not protectionism, or autarky as naysayers suggest. Europe is an economic superpower and we have the means, the funds and the talent to do better.
Europe’s “Conduct” Fixation Created Cover for the Occupation of Our Infrastructure
Tech giants who strode into Europe and established themselves in search, social media, app stores, ecommerce, software and more, drew legal complaints from small European websites upset they were not getting “fair” distribution and discoverability online. As always with hugely asymmetric bargaining power the terms were unfair, there was “self preferencing” (favouring one’s own version of the service), exploitation, stealing entrant’s ideas, overcharging. All of this was seen in Europe just through the familiar legal construct of “abuse of dominance”: companies with market power extracting an unfair bargain, need to be taught how to behave. A huge amount of intellectual effort went into rationalising how this conduct could be characterized as “anticompetitive”. We developed “economic theories of harm” – rationalising how it was “rational and profitable” for these companies to marginalise, exclude or exploit competitors. The EC started the painful process of opening cases, gathering evidence, developing “theories of harm”, with years of to-and-fro, submissions, information requests, timing extensions, hearings, decisions, appeals, judgments, more appeals. All lasting at least a decade. Then we pivoted to regulation: purposing the same antitrust rules as conduct principles (“thou shall not self-preference”), on the assumption that setting the rules of the road was going to be somehow easier than proving misconduct ex post. All looked like a lot of effort. It attracted bicep emojis from regulators (“we are taming them!”), amplified through a cottage industry of academics and civil society hopeful this was Europe’s resistance to “digital monopolies”. Except that none of the giants this meant to “tame” took any of it very seriously: with endless opportunities for appeal and infinite resources, the process can be strung along indefinitely. And by an large, these companies won: none of these cases moved the dial in terms of remotely shifting the business models that motivated the conduct (a point I had made for years).
But more fundamentally, this was like shooting at an avalanche with peas. What was really happening is that while the intellectual energy of the European elite was reassured by these conduct cases (“we are doing something!”), these same targets were asking privately: “what’s up with Europe? It suits us you keep yourselves busy with these cases, but where’s your real assets? a few websites? keep at it while we build it all out ourselves”.
And so they did. We now operate our digital lives through a complex layer of stack components involving hardware, software, data, architecture, controls, permissions, authentications - all of which we do not have the key to. This is all “at their majesties’ pleasure”. From chips all the way to applications, Europe is a digital vassal. And it is no surprise we “find ourselves” in a laggard position vis-à-vis the US and China on AI. I push back every time a European politician utters the words: “Europe finds itself behind in tech infrastructure”. No we don’t “find ourselves”: it did not suddenly happen, it was not an act of God. We led ourselves there.
It’s not that regulation is useless per se. But it is totally unequal to the task
The truly extraordinary thing, with hindsight, is the astonishing fact that Europe allowed a handful of American companies to become the basic digital infrastructure of the continent. The antitrust/regulatory language of “gatekeepers” and “self-preferencing” was inapt. The problem is much more fundamental: we have effectively outsourced large parts of our cloud, operating systems, app distribution, business software, search, advertising infrastructure, communications, data storage and now increasingly AI infrastructure to a tiny number of foreign private companies. And while we would not tolerate anything remotely comparable in other infrastructures (energy, banking, defence, telecoms), we somehow normalized it in digital because each step arrived packaged as a product choice rather than as a transfer of strategic power.
It is not that antitrust and regulation were completely pointless - some of the cases were necessary and intellectually important. For instance, it was important to figure out how Google leveraged power by inducing Android OEMs to place Search as the default search engine on new devices as a condition for being able to pre-load the Google Play Store; or how being on both sides of advertising auctions gave Google inordinate power to distort the advertising market to its advantage. But this work[1] was never remotely commensurate to the entrenchment we were beginning to see. We spent years proving particular instances of tying, self-preferencing, exclusion and discrimination while the underlying architecture of dependency continued to deepen. We spent years explaining to each other how conventional antitrust thinking should update to account for network effects and zero price economics. Years of economic papers at academic conferences, with variants of economic models on “foreclosure” and “exclusion”.
Even on its own terms, the record is weak: cases against Google ended up in infringement decisions in Brussels and US Federal Courts, but no remedies scratched the surface. Let’s name the others for completeness: EC cases against Amazon for self-preferencing in the Buy Box and use of seller data ended in settlements and a nothing burger for Amazon; cases against Apple ended in settlements, some fines, some minor obligations that don’t remotely affect the business model; Microsoft Teams – ditto, a clever settlement on some disaggregated price schedule which does nothing to really power entry[2]. Regulation then stepped in, reformulating and repurposing the same rules under the “DMA” – but pace has been slow, the EC knows it is under the Damocles’ sword of tariffs and other retaliations, only Apple and Google have received some form of final decision over the past 3 years - all very feeble.
But much worse: the real issue was not market conduct towards small rivals; rather the fact that these players had become de facto privately owned market infrastructure. If you control the cloud, the operating system, the app store, the default, the identity layer and increasingly the AI interface, telling you not to discriminate at one particular point in the stack hardly changes who owns the terrain.
Worse: an uncomfortable truth
Worse still. We are now finally understanding we need to think of digital infrastructure in terms of industrial policy. But Europe already had an industrial policy in digital: it was largely American industrial policy. Every time a ministry standardised on Microsoft, a university moved its data to Google, a hospital became dependent on AWS, or a public agency built around one proprietary stack, European public money helped finance the scale, learning effects, data advantages and switching costs of the very firms Brussels later complained were too powerful. European businesses building on this infrastructure are just funding the mechanisms through which incumbents perpetuate dominance. So European companies remain dependent on American cloud, American app stores, American distribution and eventually American acquirers.
A more muscular approach could have tried to push for structural separation, limits on bundling across cloud/software/identity/AI, tough restrictions on acquisitions of strategically important European companies, real interoperability and switching obligations, procurement diversification requirements, and perhaps even a presumption that critical European public institutions cannot become technologically captive to a single foreign hyperscaler. But in the political economy of the 2010s, Europe just did not believe it could seriously pursue American giants. Meantime, the infrastructure rollout was complete.
Moving away from the antitrust/regulation blind spot, and building, is foundational to Europe’s economic future
The failure has not simply been weak enforcement, not even primarily bad regulatory design. It was failure to understand the nature of power in digital markets. What matters is ownership of infrastructure, control of bottlenecks, access to capital and ability to scale. Europe spent fifteen years trying to civilise five private digital empires while doing far too little to reduce their territory, build alternatives and shift power away from them. At this point the objective cannot simply be to continue to pursue better behaviour by these incumbents. It should be a deliberate reduction in European dependence on them, combined with the creation of European technological capacity.
The reason this is such an existential issue for Europe is not “just” the observation we are dependent, at the receiving end of potential geopolitical leveraging and weaponization of these services. It is the much more profound issue that what is at stake is the future of Europe’s economic model. Mario Draghi in his famous September 2024 “Competitiveness Report” made a central point of his analysis of Europe’s decline in productivity growth relative to the US over the last 20 years. The true extent of this gap has been the topic of much excitable economists’ debate over the last few months (between Luis Garicano, Philippe Aghion, Antonin Bergeaud on one side, and Paul Krugman, Benjamin Wolf, Seth Ackerman, Brad de Long and a few more on the other, as summarised at this DG Grow event). But no matter where one stands on the measurement, it is broadly accepted that a large share of the widening gap reflects differences in IT creation, adoption and diffusion. The US lead on IT and innovation helped America, unsurprisingly, much more than Europe, leading to higher US wages and profits. For Europe, owning a larger share of our infrastructure and building on it would allow us to capture more of the value we create, as opposed to leaking it to external actors. It is imperative for Europe to recapture at least a portion of the infrastructure we have subcontracted, and to which we do not own the keys. This is not protectionism or autarky: we just cannot be building our future entirely on someone else’s infrastructure.
This is painfully manifest now in AI. Europe is watching the US sink extraordinary sums into training frontier models. Whether that bubble of circular finance will implode we cannot yet tell, but we do not have frontier labs. One faction of “AI maximalists” in Europe are so distressed by this that they are churning out pleas for Europe to bend the knee further to US hyperscalers, begging to be promised access to these models in exchange for favourable terms to build more of their datacentres on European soil. This is such a non-starter I will not discuss it here further (though I discussed it elsewhere). We also do not have the resources and the apparent exceptional drive of China, which seems to be catching up in impressive ways. Europe is telling itself “we will be champions at industrial AI, physical AI, we don’t have frontier models but we are domain kings in mechanics and machinery”. True we can. But beware: this is still not “sovereign AI”. We will be building industrial AI on someone else’s infrastructure.
What’s the prescription?
It is hard. Europe’s current predicament would require an aggressive, single-minded drive to treat digital infrastructure the way countries once treated railways, electricity grids, aerospace or defence - as an area where public capacity, procurement, capital and strategic direction are legitimate and necessary. Let’s stop the noise on antitrust and regulation, they are a distraction and a waste of energy – have at it in your corner, no one really cares.
We would need very large direct investment in compute, cloud, chips, data centres, energy, foundational software and open models, governments becoming anchor customers rather than merely writing subsidy cheques, public institutions deliberately creating demand for European suppliers, and perhaps public or quasi-public investment vehicles willing to take long-duration equity risk at a scale that normal European venture capital simply cannot absorb. But European institutions have demonstrated they cannot do this. Time and again Europeans look hopefully at Brussels with the standard “infantilised” posture “the Commission must be doing something” – but Europe fails every single time. The regulatory instinct prevails, Brussels produces endless consultation papers and draft laws, fails to deliver anything timely or radical enough, and sprinkles little money on small initiatives, but no more. The Member States have shown they can do a little of their own initiative (e.g. parts of the public administration in France, Germany and elsewhere have shifted to European providers without any Brussels mandate). A useful signal but not enough. Behaviour is slow to adjust to aspirations.
There are two areas where private initiative can make a difference. Movements like EuroStack are seeking to animate not just the discussion, but the action.
The critical area is demand activation. Unless private sector demand moves away from hyperscalers’ defaults, there is no market and no investment in European solutions. Here we need to overcome significant inertia, as the entrenchment of hyperscalers with European businesses runs deep: procurement choices, investment patterns and incentives were all seen in the past as manageable trade-offs, no one thought of dependencies in the face of convenience, scale and short-term efficiency: ready-made infrastructures, integrated service stacks and immediate scalability. Moving involves detailed effort and engagement with Boards to motivate an assessment of dependencies, evaluation of the business case for alternatives, and managing the shift. The hardest part is this is a classic “collective action” problem where the benefits of acting together are clear, but there is perceived risk in “going it alone”. Collective procurement and buy clubs are getting on the agenda.
The second strand is directing investment to European solutions. Of course, complex digital infrastructure ecosystems are not primarily created through one-off capital injections. They emerge when sustained demand generates revenue streams that justify investment and ecosystem development. This is why demand (again) is key: the issue is how European private procurement decisions (which are vast) can be reorganised to create viable markets for sovereign infrastructure. Without such demand signals, no investment programme can materialise. But that said, we need to see the extraordinarily rich pools of European capital (especially institutional capital: pension funds sit on over €3T of assets under management, and insurance companies €11T) liquify and direct themselves towards European scale ups. Pension funds and insurance companies elsewhere are more conversant with these deployments – in Europe there are still too many regulatory constraints and as a result institutional funders remain reticent, quote “fiduciary duties” and repeat “pension funds do not speak to VCs”. On the positive side, we are seeing signs that the “investible window” (what these actors may consider “investible”) may be slowly shifting and orienting more to support European tech.
Plainly, it is hard to think of a centrally-organized digital sovereignty initiative in Europe. Which institution would lead this, realistically? Building infrastructure should be unambiguously thought of as industrial policy. Yet in Brussels we separate industrial policy from digital policy because “we don’t want to step on toes” - instead of working together. We cannot continue to wait for Godot and wring our hands hoping that “the Commission will do something”: get us a Capital Market Union, get us a Single Market - can we wait decades?
Ultimately, Europe is better than its institutions. We have a dozen amazing tech ecosystems dotted across the continent – while there is value in ecosystem aggregation, we will never have a single Silicon Valley but can have a network of ecosystems. China works on this model. This is a feature, not a bug. We have extraordinary talent. We have enormous capital pools that must be activated. Antitrust and regulation were simply unequal to the task, it is high time we abandon the delusion they are doing much for Europe, treat digital as our industrial policy and focus all our intellectual and positive energy on BUILDING. Europe is an economic “superpower”, not a “middle power” – let’s start acting like one.
[1] I was involved in these cases adverse to Google
[2] I was involved in these cases for defendants