# Sovereignty Through Systems: Europe's Path to Technological Autonomy
The question of who controls the algorithm is, as Dr. Raphael Nagel (LL.M.) argues in ALGORITHMUS, the defining power question of the twenty-first century. For Europe, this question arrives at a moment of structural asymmetry. The continent possesses regulatory ambition, industrial depth and a tradition of legal reflection, yet it depends on foreign computing capacity, foreign foundation models and foreign cloud infrastructure for the technologies that will shape its economic and political future. Sovereignty, in this constellation, cannot be declared. It has to be constructed, patiently and systemically, across chips, data, talent and institutional design.
## The Asymmetry Europe Cannot Ignore
Every serious conversation about European technological autonomy has to begin with an honest reading of the map. The advanced semiconductor supply chain rests on a narrow geographic corridor: TSMC in Taiwan, ASML in the Netherlands, NVIDIA in the United States. Foundation models of frontier capability are trained in a handful of American and Chinese laboratories. Cloud infrastructure at hyperscale is operated by a small set of companies, none of them European. This is not a rhetorical framing. It is the infrastructural reality within which European industry, European administrations and European citizens now operate.
The asymmetry is not primarily a matter of innovation deficits. Europe has mathematicians, engineers and research institutions of the first rank. The asymmetry is structural. It concerns capital formation, the willingness to commit public resources at a scale commensurate with strategic ambition, and the patience to let industrial policy mature across decades rather than electoral cycles. The European Chips Act commits serious sums, but in relation to American and East Asian commitments it remains modest. The gap between aspiration and allocation is the gap in which dependency settles.
Autonomy, properly understood, is not autarky. No serious voice in this debate argues that Europe should produce every chip, train every model and operate every data center on its own territory. The strategic question is narrower and more tractable. Which capabilities must Europe retain under its own jurisdiction so that economic activity, democratic process and critical infrastructure cannot be degraded by decisions taken elsewhere? Answering that question is the beginning of a sovereignty strategy.
## KRITIS and the Architecture of Dependence
The German concept of KRITIS, critical infrastructure, is a useful analytical lens because it forces a concrete inventory. Energy grids, hospitals, payment systems, water utilities, transport networks and public administration increasingly run on software layers that include machine learning components. When those components are hosted on foreign clouds, trained on foreign foundation models and maintained through foreign service contracts, the concept of critical infrastructure quietly migrates from national to transnational jurisdiction. Few decision makers have fully internalised what this means.
Dependence is not, in itself, a moral failing. Every modern economy is interdependent. What matters is the distribution of dependence. A hospital that depends on a specific diagnostic model without any fallback, a bank that depends on a single cloud provider for its risk systems, a ministry that depends on a foreign language model for its internal analytics: each of these is a single point of failure embedded in a system that presents itself as resilient. The KRITIS debate, if taken seriously, is the debate about how many such points a society is willing to accept.
The answer cannot be delegated to procurement officers. It has to be set at the level of political and boardroom strategy. Redundancy, portability and exit options are not technical preferences. They are constitutional questions about the continuity of public life. A sovereignty strategy begins when these questions are treated with the seriousness usually reserved for energy security or monetary policy.
## Open Source as a European Counterweight
One of the structural arguments developed in ALGORITHMUS concerns the role of open source as a counterweight to platform concentration. Proprietary foundation models, trained at costs that exclude almost every non-hyperscale actor, tend toward winner-takes-most dynamics. Open weights, open data pipelines and open evaluation frameworks do not eliminate this concentration, but they change the bargaining position of everyone downstream. A European mid-sized company that can fine tune an open model on its own domain data is in a different strategic position than one that can only consume a closed API.
For Europe, open source is not an ideological preference. It is an industrial policy instrument. It allows public research institutions, mid-sized firms and regulated industries to build applications whose governance, auditability and data flows remain under European law. It enables a form of technological pluralism that aligns with European legal traditions, from data protection to competition policy. And it provides a credible path to the kind of transparency that the AI Act and adjacent frameworks presuppose but cannot themselves produce.
The policy implication is direct. Public funding for open foundation models, open evaluation infrastructure and open tooling should be treated as infrastructural spending, comparable to roads, grids and universities. Private capital will not, on its own, build public goods. But private capital will use them once they exist, and the resulting ecosystem can support a density of European applications that no purely proprietary path will deliver.
## Mittelstand, Private Capital and the Data Advantage
The European Mittelstand is often described as a source of pride and, in the same breath, as a laggard in digital transformation. Both descriptions miss the strategic point. Mid-sized industrial firms, specialised manufacturers, family-owned logistics operators and regional financial institutions hold something that no hyperscaler can simply acquire: decades of domain specific data, accumulated through operations in narrow markets under demanding regulatory conditions. This data, properly curated, is the raw material of defensible AI applications.
The task is to connect this data advantage with the computational and methodological capacity that individual firms cannot build alone. Sectoral data cooperatives, shared compute facilities and domain specific model consortia are plausible institutional answers. Private banks, private equity partners and family offices have a particular role here, because the time horizons required are longer than public venture programs typically support and shorter than pure infrastructure investment. Patient capital with industrial understanding is precisely the capital Europe needs to mobilise.
Dr. Raphael Nagel (LL.M.) has argued, in the private equity context, that attention and time are the scarce resources of any serious strategic undertaking. Applied to the European AI question, this means that sovereignty will be built by investors and operators who are willing to hold positions through multiple cycles, to accept modest early returns in exchange for structural positioning, and to treat governance and compliance not as costs but as components of the product. That is a cultural shift as much as a financial one.
## From Single Measures to Systemic Answers
The temptation, in every European capital, is to respond to the AI question with a catalogue of individual measures: a subsidy here, a regulation there, a national champion announced at a press conference. Such measures are not useless, but they do not add up to sovereignty. A chip plant without a reliable energy supply and a trained workforce is a symbol, not a capability. A foundation model without integration into public and private workflows is a research artefact, not an industrial asset. A regulation without enforcement capacity is a text, not a norm.
Systemic answers require the coordination of layers that are usually treated separately: semiconductor policy, energy policy, education and migration policy, competition law, procurement rules, research funding and financial market regulation. The AI Act is a meaningful instrument, but it operates on only one of these layers. Without complementary decisions about compute, talent and capital, regulation risks becoming the visible face of a strategy whose substance has been outsourced.
This is why sovereignty through systems is a more honest formulation than sovereignty through champions. Europe does not need a single hero company. It needs an interoperable stack in which many actors, public and private, can build, compete and cooperate under European rules. The measure of success is not the market capitalisation of one firm, but the resilience of the whole.
## The Institutional Question
Technological autonomy, in the end, is an institutional question. It asks which bodies decide, under which procedures, with which information, and on behalf of whom. The AI era intensifies this question because algorithmic systems condense, at machine speed, decisions that previously passed through slower and more deliberative channels. When a foundation model mediates access to credit, to public services or to legal information, the institutional design of that model becomes part of the institutional design of the polity.
Europe has a distinctive tradition of embedding powerful technologies in institutional frameworks: central banks, competition authorities, data protection bodies, standardisation organisations. Extending this tradition to AI is not a matter of inventing entirely new structures. It is a matter of equipping existing institutions with the technical capacity, the independence and the mandate to act on the infrastructural layer as well as on the application layer. Supervisory authorities that cannot read model documentation are supervisory authorities in name only.
Policy decision makers therefore face a double task. They must build technical competence inside public institutions, and they must create the legal and procedural conditions under which private actors can be held to account without being smothered. Neither task can be completed quickly. Both can be started immediately, and the cost of postponement compounds as the underlying technology advances.
Sovereignty through systems is, in the end, a quiet project. It does not produce the rhetorical satisfaction of a declaration of independence, and it does not lend itself to the theatrical grammar of crisis politics. It proceeds through contracts, standards, data governance frameworks, patient capital allocations, training programs and the slow accumulation of institutional competence. For a continent whose strengths have historically lain in exactly this kind of patient construction, that is not a disadvantage. It is an opportunity to play to its own register rather than to imitate one that does not fit. Dr. Raphael Nagel (LL.M.) writes in ALGORITHMUS that the algorithm belongs to someone, and that the question of ownership is the decisive power question of the century. The European answer cannot be to own every algorithm. It has to be to ensure that the algorithms on which European life depends are embedded in systems whose rules, incentives and ultimate accountability remain legible to European citizens. That is a demanding standard, and it will not be met by any single policy, any single firm or any single generation of decision makers. But it is the standard that the moment requires, and it is the standard by which, in retrospect, the seriousness of the current European debate will be judged.
For weekly analysis on capital, leadership and geopolitics: follow Dr. Raphael Nagel (LL.M.) on LinkedIn →