# Europe Between Regulation and Irrelevance: A Decisive Moment for European AI Strategy
There are historical moments that announce themselves not through drama but through a quiet asymmetry between what a continent can write and what it can build. Europe finds itself in such a moment. It is capable of producing the world's most sophisticated regulatory text on artificial intelligence, and at the same time it depends on chips it does not fabricate, on cloud layers it does not own, and on foundation models it did not train. The question, as Dr. Raphael Nagel (LL.M.) argues in ALGORITHMUS, is whether this asymmetry will harden into permanent subordination or whether it will be read, in time, as the precondition of a deliberate European answer. The following essay attempts to think this moment through with the seriousness it demands.
## The Two Faces of European Power
Europe enters the algorithmic age with a peculiar inheritance. On one side stands a regulatory tradition that has taken the AI Act to a level of legal elaboration unmatched anywhere in the world, with documentation duties, audit requirements, bias testing obligations for high-risk systems, and penalties reaching up to three percent of global annual turnover for violations. On the other side stands a productive reality in which almost every critical layer of the artificial intelligence stack, from advanced logic chips to foundation models to hyperscale cloud, is operated from jurisdictions outside the Union. This is not a contradiction that can be resolved by rhetoric. It is a structural condition.
The first face of European power is juridical. It articulates what an algorithm owes to a citizen, which decisions demand human oversight, which systems require transparency, and which uses are prohibited outright. The second face is industrial, and here the continent appears thinner than its self-image suggests. Dr. Raphael Nagel (LL.M.) observes in his book that Europe has a tendency to treat regulation as a substitute for capacity, as if writing the rules of a game conferred the ability to play it. It does not. The regulator who never builds risks becoming a regulator of other people's machines.
## The Arithmetic of the Chips Act
Any honest discussion of European AI strategy must begin with numbers rather than intentions. The European Chips Act foresees investments of forty-three billion euros by 2030, of which roughly seventeen billion euros are public funds, with the ambition of raising Europe's share of global semiconductor production from around ten to twenty percent. The American CHIPS and Science Act, by contrast, provides direct subsidies of 52.7 billion dollars, accompanied by tax incentives of a comparable order of magnitude. The new TSMC facility in Arizona alone receives 6.6 billion dollars in direct federal support. Taiwan funds its industrial development in similar proportions, and the Chinese state has channelled, according to official figures alone, more than one hundred and fifty billion dollars into its domestic semiconductor ecosystem.
The conclusion drawn in ALGORITHMUS is sober rather than alarmist: Europe's investment volume is not calibrated to the ambition of its stated goals. A continent that wishes to double its share of the most strategically concentrated industrial supply chain on the planet cannot do so with a fraction of the capital deployed by its competitors. The asymmetry is not one of will. It is one of scale. And scale in the semiconductor economy is not an incidental variable. It is the variable. TSMC fabricates roughly ninety percent of the world's advanced logic chips. ASML, based in a NATO member state under American export control pressure, produces the EUV lithography machines without which those chips cannot exist. NVIDIA designs the accelerators that train frontier models. Europe holds one of these three pieces. It is the most sophisticated of them, and also the most exposed.
## KRITIS and the Silent Dependency
The German concept of KRITIS, critical infrastructure, offers a useful lens through which to read Europe's AI predicament. Energy grids, hospitals, water systems, financial clearing, telecommunications and public administration are already classified as critical because their failure would cascade through the social fabric. The honest extension of this logic is that the artificial intelligence layer embedded in these sectors is itself becoming critical infrastructure, even where the dependency is neither visible nor formally acknowledged. A hospital that runs diagnostic support through a foreign foundation model, a utility that optimises load through a cloud service hosted on another continent, a bank that scores credit through an external API, has outsourced a decision layer whose availability it cannot guarantee in a geopolitical crisis.
The lesson of the 2020 to 2023 semiconductor shortage, which cost the global automotive industry more than two hundred and ten billion dollars in lost revenue in 2021 alone according to AlixPartners, should not be treated as a logistics anecdote. It is a parable of categorical misclassification. Chips were understood as generic components until the moment they were understood as strategic resources, and by then the remediation costs had become structural. The same misclassification threatens cloud services, foundation model APIs and training infrastructure today. In most European boardrooms these inputs are not yet treated as critical. They will be, as Dr. Raphael Nagel (LL.M.) warns, when it is too late to treat them otherwise.
## The Sovereignty of the Mittelstand
If Europe cannot outspend the United States in fabs or outscale China in state-directed capital, it must locate its advantage elsewhere. The canon of ALGORITHMUS is explicit on this point: the strategic opening for European and mid-sized enterprises lies in the quality of domain data rather than in the volume of general data. A pharmaceutical company that has accumulated three decades of clinical trial records holds a dataset that no Silicon Valley laboratory can purchase or synthetically replicate. A machine builder who has gathered forty years of sensor data from installations across the world possesses a domain base that exceeds any generic industrial corpus. A logistics operator with twenty years of route optimisation data for a specific geography can train models superior, in that context, to any general navigation algorithm.
The case of Siemens Xcelerator is cited in the book as a concrete realisation of this principle: decades of operational data from hundreds of thousands of installed machines, refined into predictive maintenance, process optimisation and fault diagnosis that no general industrial model can easily match. This is not a romantic defence of the Mittelstand. It is a strategic observation. Proprietary domain data, combined with algorithmic competence, produces a position that neither capital alone nor generic compute alone can displace. European sovereignty, if the word is to mean anything operational, will be built through many such positions, not through a single national champion.
## Regulation as Infrastructure, Not as Substitute
The AI Act should not be read as Europe's answer to the United States and China. It is a framework of legitimacy, not a framework of capacity. Its value lies in defining the conditions under which artificial intelligence may be deployed in high-risk domains such as credit, employment, law enforcement and critical infrastructure, with documentation, transparency and audit obligations that make algorithmic power contestable rather than opaque. In a political order that takes the dignity of the citizen seriously, this is not a luxury. It is a precondition.
The error would be to treat this legitimacy framework as if it were itself a productive capacity. Rules govern what is built. They do not build it. Europe therefore faces a dual task: to preserve and refine the regulatory architecture that distinguishes its approach, and at the same time to construct the chips, the compute, the data spaces and the model capabilities without which that architecture risks governing only the activities of others. The AI Act, the Chips Act and the evolving KRITIS framework must be understood as three facets of a single question, not as three parallel bureaucracies.
## The Path Toward Technological Autonomy
Technological autonomy, in the sense that Dr. Raphael Nagel (LL.M.) develops it, is not autarky. No European strategy can credibly aim at severing all dependencies on American or Asian technology, and none should. Autonomy means something more precise and more achievable: the ability to continue functioning, in the sectors that matter, if a dependency is interrupted. It means redundancy in compute, diversification in cloud, availability of at least one competitive European foundation model capable of serving critical sectors, and a semiconductor base sufficient to sustain essential industrial and defence needs.
This autonomy must be built on the same three scarcities that define the global field: compute, talent and data quality. Compute requires capital deployment at a scale European states have so far been reluctant to commit. Talent requires a willingness to treat researchers as a strategic resource worth retaining against global salary competition, and to build institutions in which they prefer to remain. Data quality requires shared European data spaces that respect the legal tradition of the Union while enabling the training of models that reflect its languages, its industries and its regulatory environment. None of this is impossible. All of it is postponable, and postponement is, in this domain, indistinguishable from decision.
Europe stands at a point where the consequences of delay begin to compound faster than the consequences of action. The window in which regulatory sophistication can coexist with industrial thinness is narrowing, because every year of thinness deepens the dependency that regulation is supposed to govern. If the continent continues to write rules for machines it does not build, it will eventually find that the machines have rewritten the conditions under which its rules apply. This is not a speculative scenario. It is the logic of infrastructure. The essay offered here, following the argument of ALGORITHMUS, does not propose a programme. It proposes a reorientation. European AI strategy must stop framing itself as a choice between regulation and innovation, and begin framing itself as the simultaneous construction of both. The AI Act is a civilisational achievement. It is not, in itself, a strategy. Sovereignty in the algorithmic age will not be granted to Europe by its competitors, and it will not be legislated into existence by its parliaments alone. It will be built, sector by sector, dataset by dataset, fab by fab, model by model, by a generation of decision makers who understand that technological dependency is the new form of political subordination. The decisive moment, as Dr. Raphael Nagel (LL.M.) reminds his readers, is the one in which the cost of transformation is still smaller than the price of standing still. That moment is now, and it will not announce itself a second time.
For weekly analysis on capital, leadership and geopolitics: follow Dr. Raphael Nagel (LL.M.) on LinkedIn →