The ‘Sovereignty’ Narratives That Don’t Help Europe

European “digital sovereignty” and “digital autonomy” have become so fashionable that the terminology is being appropriated from all sides as the jumping off point for all sorts of claims and agendas.

While fewer people had the first idea what “digital sovereignty” was only a couple of years ago (I first mentioned it and the concept of a “EuroStack” on Tech Policy Press in August 2024), we are now in “sovereignty overload” as everyone drags their favorite topics into the discussion—digital regulation, access to frontier models, talking down Europe, “AI maximalism,” you name it. “The missing ingredient for digital sovereignty is X!” (where X is the writer’s pet passion); or “Digital sovereignty can never be achieved because of Y!” (where Y is the writer’s own special angle); or the best: “Digital sovereignty is delusional so we need to do deals with hyperscalers who will help Europe towards greater sovereignty!”

If we want the project of digital sovereignty to deliver for Europe and beyond, we need to stick to a narrow plan that produces real results, and not roll in all directions. Here’s a simple rendition of the plan from the perspective of the EuroStack Initiative Foundation—which I regard as the main pragmatic voice for European digital sovereignty today, supported by nearly 500 companies across the tech stack:

• The objective is to strengthen European digital assets and capabilities so that Europe is less of a “digital colony”. This does not mean protectionism, autarky, or exclusion of non-European vendors. There is no general adversarial sentiment towards US tech firms. There is no scenario in which they can all be supplanted. The aim is nonetheless to grow the share of European demand supplied by European solutions. The motivation is value capture and growth, not “democracy” and “European values.” Europe is currently in a position of grave dependency on non-European suppliers for multiple layers of its digital stack, and this undermines its ability to choose its own technology path, the direction of innovation, productivity growth, industrial strategy and ultimately geopolitical leverage.
• Sovereignty should be concerned with the “full stack,” not just chips or cloud or hardware or software or connectivity or AI. We must care about growing European capabilities along the stack. This does not mean that the current configuration of the stack needs to be supplanted by an equivalent “European stack” built as an alternative monolith—this would be an unachievable plan, and the Bertelsmann Report suggestion this would require €300 billion was both an understatement and a measure of how much such a plan would be a non starter. We cannot be other than organic, opportunistic, incremental. No one has “a comprehensive plan” (least of all the European Commission). We will make it up as we go along, as economic incentives shape demand and funding. This is the nature of capitalism. Because while this should be an industrial strategy project, the reality is that Europe is too fragmented and scattered, and the EU’s authority and speed are too feeble and slow to make a difference. This must be a private initiative or bust. The geopolitical environment is favorable; funding is turning to European assets.
• “Sovereignty” is not specifically “AI sovereignty.” If one means that, one should be explicit. But no sensible discussion of digital sovereignty or autonomy begins and ends with AI. AI is a use case: cloud + software + data. And as was well explained by Frederike Kaltheuner and Leevi Saari on Tech Policy Press, AI that is built on hyperscaler cloud infrastructure (not just for model training, but also for inference, i.e. compute for model deployment) is not “sovereign.” We should be clear about limitations and magical thinking.
• There is no clear, agreed roadmap because there is no single entity in control. So again, there cannot be a single plan. There will be different patches of initiatives at the national level as institutions have started doing their thing. The Commission’s June Sovereignty Package disappointed, but Member States are increasingly moving off US solutions independently. Enterprises and businesses are starting to come off the fence: Airbus has gone for Scaleway, the Dutch National Bank and multiple others for StackIt, and many more are experimenting and planning migrations. There is no central planning, and as there is no regulatory obligation to “buy European,” there needs to be a vision and a business case. But there is. And uncertainty around the adequacy of protections for European data transferred to the US (on which over 3,600 US companies processing European data currently rely) after the Trump v. Slaughter SCOTUS ruling hangs as a further risk on European boards and motivates the search for alternatives.
• There is growing interest from capital allocators. The European meme has forever been that the continent doesn’t have “innovation,” and that’s why we are where we are. Not so! Europe has tons of innovation—including research and startups—which it doesn’t scale because notoriously it sends billions of savings a year to US venture capitalists to invest in US tech, or in European startups after they migrate. This is changing: as European demand for European products increases, returns improve and capital allocators are turning to scaling up as a worthwhile endeavor.
• This is what Europe should focus on, single mindedly and aggressively. Buy, fund, build. Do not punt to the Commission and the institutions all the time. That venture funds depend for a third of their funding on government entities in Europe, vs 4% in the US, reflects the reality that the European mindset remains utterly dependent and deferential to Brussels. But this need not be the way. It breeds a different type of dependency—from the bureaucracy. There is a vast lobby industry which makes up the Brussels Bubble and wastes energy in useless meetings and pursuits. Events. Roundtables. Summits. Dinners. This is all a distraction, which perpetuates Europe’s lack of actual agency.
• Will it work? Maybe, with effort. Not as a top-down Commission plan, which will always be inadequate as it is not informed enough by business and capital, as well as underfunded. Private initiative needs to move and gain agency. European elites are complacent. Life is too sweet still in Europe. Citizens are much more alert and patriotic: but the elites are massively complacent. We will see.

As if all of this wasn’t difficult enough, a few narratives have sprouted around “digital sovereignty” lately that are either unhelpful, or at best a distraction:

1. A distraction: ‘The missing ingredient to digital sovereignty is enforcement of digital regulation’

This is the preserve of the regulation fans—those who invested massively in pushing and advocating for regulation like the Digital Markets Act (DMA) and the Digital Services Act (DSA), and cannot quite let go of this hope. The reality is that the regulation of the tech giants had failed to deliver anything worthwhile in Europe even before the Trump administration changed the context. But there is a hardcore set of experts for whom the task of developing and inspecting regulation is an occupation, who fundraise on this basis for white papers and symposia and gatherings among themselves and regulators, and who cannot let go of the idea despite the lack of any clear achievements of European digital regulation other than very marginal ones.

Did chasing Google Search give Europe a cloud? Ditto Android. Ditto Apple’s App Store, or Amazon e-commerce. I am all for powerful monopolies being reminded there is a limit to throwing their weight around and extracting (and I was involved in enforcement against Google directly in multiple jurisdictions). But alas experience has shown it does not majorly change their conduct, nor does it change market structure on the ground. European entrants fleetingly hoped they would benefit from the DMA, deluded by legions of advisors who stood to profit. But real results? They are difficult to discern.

The refusal of Europeans to face up to this as they still waste energy on the task of “regulate” instead of “build”—which is the most important thing we should have spent effort on for the last 15 years—is frustrating, because it funnels a lot of attention into this ultimately pointless effort.

Now the pivot. “Sovereignty” was nowhere in the minds of regulation fans when the regulation came about. It was good old fashioned utility-style regulation, with a couple of twists to accommodate network effects. Now that sovereignty is fashionable, regulation is being repurposed as “if we had proper enforcement of digital regulation, sovereignty would happen much faster because we could actually switch to alternatives!” This logic assumes there actually are alternatives, but they are not there because Europe forgot to build.

2. Wrong headed and dangerous: ‘Europe will never be a sovereign AI superpower’ or ‘Europe can never win the sovereign AI race’

It’s bad enough if uninformed Americans write this stuff. But Europeans writing this are truly puzzling. This was, for instance the title of a recent piece on Foreign Policy, written by Europeans, but the same sentiment is widely shared among many who write with what appears to be a mixture of schadenfreude about the state Europe is in (as if it wasn’t a common problem all over the world), plus a dose of knowing, savvy, and Silicon Valley cool aid thrown in.

First, what does the claim even mean? Who ever said Europe will be an “AI superpower,” or a sovereign one at that? Europe is an economic superpower, that much is true, and not a “middle power” as many would like it to resign itself to. And as such, it can mobilize resources—it is not poor. But it can do a lot with more agency and urgency. In AI, it is true Europe cannot currently create a frontier AI lab to match the capabilities of the top three US labs, and most of the value will be captured in compute and core intelligence. But there are multiple alternative architectures to the current LLM scaling laws, and Europeans can build something even if they are a few steps behind. For inference, smaller specialized models do not necessarily need to be stripped down versions of LLMs—more players are serving that segment, and there is a real opportunity.

3. Truly dystopian: ‘Europe needs to do deals with hyperscalers to enhance its future sovereignty’

Another version of the above narrative takes it to the next level and argues the answer is a wartime-speed buildout of compute on European soil, which includes cutting deals with US hyperscalers to site their data centers in Europe (thus formally under European jurisdiction) in exchange for access to frontier LLM models that are ultimately at the whims of Washington (see here). And what do hyperscalers get for it? The talk is to give them inducements to build the compute needed to serve incoming European demand—permits, access to cheaper surplus French nuclear energy, etc. This is horrifying and unnecessary. Europe knows how to build data centers—many Europeans can and do. The constraint isn’t funding—there is plenty of funding to build in Europe if there is demand. Without demand materializing, European builders take risks that US hyperscalers do not face, since they have already secured forests of GPUs and can always deploy capacity to serve their massive captive customer base in the US and abroad. But if there are data centers to be built in Europe, they should be built by Europeans. Not to do so is rolling back sovereignty further.

***

There is much to do for Europe. The answer is agency: European enterprises and capital allocators taking the initiative without waiting for Godot—the institutions—forever. Stop the chattering and essay writing, and do. Europe has the talent, the money and the capability to rise. There is no “race” to be “won”—let’s do serious work and grow.