Europe wants to become digitally sovereign. Across the continent, governments and industry are investing billions of euros in data centres, cloud infrastructure, semiconductors and artificial intelligence. Hardly any digital strategy today is complete without the ambition of developing a “European AI”.

The underlying assumption appears straightforward: if the infrastructure is located in Europe and the AI is developed in Europe, digital sovereignty will naturally follow.

But this is precisely the assumption Europe should challenge.

Digital sovereignty neither begins with the location of a data centre nor ends with the development of a European AI model. It is ultimately determined by a far more fundamental question: Who remains in control of technology, data and decision-making?

The current debate around artificial intelligence illustrates how easily digital sovereignty is reduced to infrastructure. Europe undoubtedly needs competitive computing capacity, cloud services and advanced AI capabilities. Without technological excellence, Europe will struggle to remain economically competitive and strategically resilient.

Yet infrastructure alone does not create sovereignty.

That is why Europe should begin its AI debate with five fundamental propositions.

Thesis 1: European hosting alone does not create digital sovereignty

An AI system does not become sovereign simply because its servers are located in Europe.

The location of infrastructure may reduce certain legal and geopolitical dependencies, but it says little about who controls the underlying models, how they are updated, which data they have been trained on or whether organisations can realistically switch to alternative providers.

Digital sovereignty therefore begins not with infrastructure, but with control over technology, data and decision-making.

Thesis 2: Dependencies cannot be eliminated—but they must be understood

No country or organisation will ever become completely independent of global technologies or international supply chains.

The real challenge is therefore not to eliminate dependencies altogether, but to identify, understand and manage them.

Organisations that cannot assess their reliance on critical technologies, interfaces or AI models cannot claim to be digitally sovereign. Transparency about dependencies is not an administrative exercise; it is a prerequisite for strategic autonomy.

Thesis 3: Sovereignty requires the ability to question AI

Organisations should never be satisfied simply because an AI system produces convincing answers.

They must understand how those answers were generated, which data informed them, which assumptions shaped the outcome and how reliable the results actually are.

This is particularly important for critical infrastructure operators, industrial organisations, public authorities, healthcare providers and security professionals. In these environments, AI must not only perform—it must also be explainable.

Sovereignty belongs not to those who simply use AI, but to those who can challenge, verify and, when necessary, override its conclusions.

Thesis 4: Europe’s strength does not lie in copying Silicon Valley

Europe is unlikely to win the global AI race simply by building larger foundation models or investing ever more heavily in computing power.

The United States benefits from unmatched capital markets, hyperscale cloud providers and globally dominant digital platforms. China is pursuing its own state-backed AI strategy at enormous scale.

Europe should resist the temptation to imitate either model.

Its competitive advantage lies elsewhere: in trust, transparency, security, interoperability and regulatory certainty. Europe does not necessarily need to build the world’s largest AI models. It needs AI that governments, industry and operators of critical infrastructure are willing to trust.

Thesis 5: Regulation becomes an advantage only when it creates trust

The EU AI Act is often criticised as a potential obstacle to innovation. Some of these concerns are legitimate. Excessive bureaucracy can slow technological progress.

But regulation can also create markets.

If European AI systems become demonstrably secure, auditable and transparent, these characteristics may evolve into a genuine competitive advantage. Highly regulated industries are rarely looking for the most spectacular AI—they are looking for the most reliable one.

The AI Act will therefore succeed not because it imposes additional compliance obligations, but because it helps establish measurable trust.

The real question

Europe unquestionably needs world-class AI capabilities and competitive digital infrastructure. But digital sovereignty cannot be achieved through geography alone. It exists only where organisations understand the technologies they rely upon, recognise their dependencies and retain control over their data, processes and decisions.

The debate should therefore move beyond the question of whether Europe needs another frontier AI model. The more important question is whether Europe can develop AI systems that are transparent, verifiable and trustworthy enough to support critical infrastructure, industrial operations and public administration.

Perhaps Europe is asking the wrong AI question. The future of digital sovereignty will not be determined by the size of a model, but by the degree of trust it can inspire. That may well become Europe’s most distinctive competitive advantage.