European Digital SovereigntyA Layered Approach

Picture a European policymaker staring at a map. One pin for every place the world's semiconductors are manufactured. Another set of pins for rare earth mineral deposits. A third for where the major cloud data centers sit. Almost none of those pins are in Europe. The question that map raises — how dependent is Europe really, and on what, and where — is exactly what Haroon Sheikh's paper "European Digital Sovereignty: A Layered Approach" sets out to answer. The way he answers it is the interesting part. The problem Sheikh starts with is methodological. Digital power is genuinely hard to measure because it is not one thing. It sprawls across rare earth mining, chip fabrication, satellite networks, cloud infrastructure, artificial intelligence, consumer applications, and billions of connected devices. Existing indices capture slices of this: the International Institute for Management Development World Competitiveness Center ranked Sweden third and Denmark fourth in digital competitiveness. The European Union's own I-Digital Economy and Society Index shows European strengths in connectivity and digital skills, but weaker performance in digital public services. None of these single metrics pulls the picture together. Sheikh's point is that attempting to create one headline number would mask the crucial differences between technologies and supply chains. You need a different kind of tool. That tool is the stack. Sheikh adopts Benjamin Bratton's concept from "The Stack: On Software and Sovereignty," which argues that contemporary digital infrastructure is best understood as vertically layered architecture — not as a flat geographic territory. Think of it as a building. Each floor rests on the one below. Different floors host different kinds of activity. And crucially, power looks very different depending on which floor you are standing on. By organizing the digital world this way, Sheikh can analytically separate the dynamics of chip manufacturing from those of consumer applications, while still tracing how the layers connect and depend on each other. That is the methodological bet the paper makes, and the rest of the paper tests whether it pays off. Starting from the ground up: the resource layer. This includes lithium, cobalt, and rare earth elements — the geological base from which all digital technology ultimately comes. Sheikh's mapping names China as the dominant actor here, citing companies like Jiangxi Ganfeng Lithium and China Minmetals Rare Earth. The European Union is highly dependent on mines in the developing world, and while some deposits exist within Europe, opening new mines would conflict with European environmental and labor standards. In Sheikh's framework, the European Union's position on the resource layer is classified as a "Threat." Moving one floor up, you encounter semiconductors. Taiwan's TSMC dominates advanced chip fabrication. US firms like Intel matter, as do South Korean and Japanese producers. China has Semiconductor Manufacturing International Corporation and HiSilicon. Europe's role here is mostly through ASML in the Netherlands, which produces the specialized lithography equipment that chip manufacturers need — a meaningful niche, but not a manufacturing position. The European Union's chip layer is also classified as a "Threat," which is what motivated the European Union Chips Act. One floor higher and the picture changes. The network layer — telecommunications equipment, satellites, undersea cables — is where the European Union is genuinely strong. Nokia and Ericsson are competitive globally. Airbus plays a role in satellites. Sheikh explicitly states that the European Union "has proven capabilities in the field of telecommunications infrastructure, satellite technology, and undersea internet cables." This is the layer where Europe is classified as "Strong," and it matters because it shows that the stack method is not just a list of European weaknesses — it identifies real industrial assets too. Then comes cloud. Here the assessment is stark. Sheikh's paper uses the word "Dire." Amazon, Microsoft, and Google dominate global cloud infrastructure. The paper states flatly that "on the global scale, there are no European firms." The European Union's response is Gaia-X, a Germany and France-led initiative to build a common European data-sharing infrastructure with shared standards. But Sheikh records the controversy: Gaia-X explicitly aimed not to create a European Amazon or Microsoft — it aimed to set standards and enable future European providers. In April 2021, non-European firms including Microsoft, Google, Amazon, Palantir, Huawei, and Alibaba became full Gaia-X members, raising real questions about whether the initiative could deliver on its sovereignty ambitions. Higher still is the intelligence layer — artificial intelligence algorithms and the data needed to train them. This layer depends on specialized hardware like Nvidia graphics processing units and on the cloud infrastructure beneath it. The paper surveys the global debate: Kai-Fu Lee argues China has an edge due to the sheer volume of application data; Jeffrey Ding counters that the United States retains advantages in fundamental science and semiconductor hardware. Where does Europe sit? Sheikh's assessment is that the European Union "is world leading" in fundamental artificial intelligence science, "on par with the United States and still ahead of China." But on data, the European Union is weaker than both, and the Union lacks large integrated technology firms. European artificial intelligence companies like Spotify and Booking exist but are, in Sheikh's words, "more specialized niche players" dependent on US capital. The Artificial Intelligence Act and national artificial intelligence strategies are the European Union's instruments here, but "the scale of ambitions and investments remain far below those of the Chinese and American artificial intelligence strategies." Above that is the applications layer — the consumer-facing platforms where Google, Amazon, Meta, Apple, and Microsoft dominate globally, and where Baidu, Alibaba, Tencent, and Bytedance are growing fast. The European Union is classified as very weak here. Then comes the connected-device or Internet of Things layer at the top: Sheikh notes that Internet of Things devices are expected to nearly triple from nine point seven billion in 2020 to twenty-nine billion in 2030. China is strong in sensors and surveillance infrastructure; the United States fields Tesla and Boston Dynamics. The European Union has industrial strengths through automotive companies, Siemens, Philips, and Alstom, and the paper classifies this as an "Opportunity" — a layer where smart policy could still yield real gains. The pattern the stack reveals is consistent and important. The European Union tends to be stronger at lower, more material layers — networks, physical infrastructure — and weaker at upper, winner-takes-all layers — cloud platforms, artificial intelligence services, applications. That vertical diagnosis then drives different policy prescriptions at different floors. Which brings us to what the European Union actually does. Sheikh maps the European Union's policy toolkit onto the stack. The General Data Protection Regulation, or GDPR, operates primarily at the cloud and intelligence layers, shaping how data is collected, stored, and used. The Digital Markets Act and the Digital Services Act target the cloud and applications layers — precisely the two layers where the European Union is weakest. The proposed Artificial Intelligence Act targets the intelligence layer, including an explicit ban on applications like social credit scoring. The Chips Act targets the semiconductor layer. Gaia-X and the COREnect network program target cloud and network layers respectively. The paper invokes Bradford's "Brussels Effect" to explain how regulation can punch above its weight. If the European Union market is large enough, has regulatory will, and the product in question is not easily split into different versions for different markets — data collection being a prime example — then the European Union can effectively set global standards unilaterally. Regulation becomes an instrument of sovereignty even when industrial capacity is absent. But the stack also shows the limits of that logic. Regulation can reshape behavior at higher layers, shaping what dominant foreign firms must do within Europe. It cannot create chip factories, cloud data centers, or the platform ecosystems that took decades and hundreds of billions of dollars to build. Gaia-X illustrates the tension: conceived as a standards and infrastructure initiative rather than a direct industrial champion, it absorbed non-European members and remains contested. The stack framework makes clear that regulation and industrial policy are complements, not substitutes — and which one the European Union needs depends entirely on which layer you are talking about. Sheikh closes with three dilemmas the stack surfaces but cannot resolve. First: sovereignty versus openness. Europe's environmental and labor standards create supply chain dependencies at the resource layer, and its strict data protections have arguably constrained European business at the cloud and intelligence layers. Second: regulation versus innovation. The European Union argues that ethical artificial intelligence standards can become a competitive advantage, but it is genuinely unclear whether that logic survives the winner-takes-all dynamics of digital platforms. Third: industrial policy versus internal market rules. Efforts to build European champions through industrial policy can collide with competition rules — the Siemens and Alstom merger dispute is one illustration, and the Gaia-X membership controversy is another. The paper presents the stack as a diagnostic tool, not a prescription. It does not tell you what to do. What it does is provide policymakers a shared vocabulary — a common map — so that when they argue about digital sovereignty, they are at least arguing about the same territory and the same floors of the same building. In a debate this complex and consequential, that shared map turns out to be a genuinely useful thing to have. This lecture was created by ennepō. Go to https://ennepo.ai to Discover, Create and Follow the latest research in your field. Read when you can. Listen when you want to.