Why Europe’s fragmented AI geography may be an advantage

June 2026

In Europe, we often discuss artificial intelligence from a position of anxiety. The usual comparison is with the United States and China, and the usual benchmark is the race for frontier models. From this perspective, Europe appears fragmented, undercapitalised and slow. However, this may only tell part of the story. If we look beyond frontier models and focus on applied AI, Europe's fragmented geography could be a strength and strategic advantage.

A recent report ‘European AI Competitiveness Beyond Frontier Models (Abre numa nova janela)’ by interface, a Berlin-based think tank, offers an interesting perspective on this debate. Rather than focusing on frontier AI models alone, the report analyses more than 24,000 AI startups across the EU, UK, Switzerland and United States to understand where applied AI ecosystems are emerging and how they relate to existing industrial strengths. This introduces a territorial perspective to a debate that is often dominated by discussions about computing power, foundation models and hyperscalers.

The report also acknowledges important caveats. For example, startup databases tend to overrepresent sectors that rely heavily on venture capital and larger metropolitan ecosystems, while smaller regional ecosystems and consumer-oriented AI activity may remain partially invisible. Nevertheless, it offers interesting insights into the geography of Europe’s emerging AI economy.

AI start-ups are geographically dispersed across the continent. In the EU, Paris, Berlin, Amsterdam, Munich and Stockholm are important hubs, but together they account for only around a quarter of analysed EU AI start-ups. The remaining three quarters are spread across 95 cities.

At first glance, this appears to be fragmentation. Compared with the Bay Area or London, AI ecosystems in the EU may lack critical mass, dense capital networks, and speed of agglomeration. However, from a territorial perspective, this pattern may also reflect Europe’s underlying economic structure, which is polycentric, sectorally diverse, and rooted in existing regional strengths. The question is how to transform this distributed geography into a new model of AI competitiveness.

Here are a few key messages from the report, viewed through a territorial lens:

Europe’s AI landscape is fundamentally polycentric

This matters because applied AI does not develop in isolation. It often requires proximity to users, industries, data, regulatory environments, and domain-specific expertise. Therefore, a dispersed geography can support the adoption of AI across different types of regional economies. Rather than one dominant hub exporting generic solutions, Europe may develop multiple specialised ecosystems linked to manufacturing, healthcare, finance, logistics, research or public services.

Existing industrial structures still matter

AI is often portrayed as a disruptive technology that overrides established economic geographies. Yet the report paints a more nuanced picture. Germany, for example, shows a strong specialisation in manufacturing and supply-chain AI. Munich emerges as an important hub in this field. Meanwhile, Barcelona shows signs of specialising in healthcare and life sciences. Meanwhile, Paris and Milan stand out in finance and insurance. These patterns suggest that AI development often builds on existing capabilities, networks, and industrial knowledge.

This means that Europe’s future AI competitiveness may depend less on creating isolated technology hubs and more on integrating AI with existing regional strengths. The relevant question is: which regions have the industrial, institutional and knowledge bases that will allow AI to become meaningful locally? This is directly linked to the logic of smart specialisation, but with a new urgency. AI has the potential to enhance existing capabilities in some regions while bypassing others.

Europe may have particular advantages in applied industrial AI

The report finds that manufacturing and supply chain is the only analysed market where the EU slightly outperforms the United States in startup numbers. Europe has strong engineering traditions, advanced manufacturing ecosystems, industrial firms and high-quality domain-specific data. These are assets, which may become central to the next phase of AI, especially in robotics, embodied AI, industrial optimisation and specialised models.

This points to a different competitiveness narrative. Europe’s AI future may lie in embedding AI in the real economy: factories, logistics systems, medical technologies, energy systems, mobility networks and public infrastructures. In such fields, territorial proximity, trust, regulation, domain knowledge and industrial data matter. These are areas in which Europe’s fragmented economic geography could be productive.

Another related territorial dimension that is still largely absent from Europe’s AI debate is energy geography. AI competitiveness is increasingly dependent on access to abundant, stable and affordable electricity, ideally from renewable sources. As data centres, AI factories and computing infrastructure expand, locations with significant renewable energy potential and favourable climatic conditions could become strategically important. Northern Sweden and Finland, for instance, already offer cheap renewable electricity, industrial capabilities and developing digital infrastructure (see our previous blog post on re‑industrialising for Europe’s green transition (Abre numa nova janela)). Europe’s future AI geography may therefore depend not only on startup ecosystems and capital markets, but also on the geography of energy transition and green industrialisation.

The geography of AI raises a cohesion dilemma

While distributed AI ecosystems can widen participation and support regional transformation, they can also lead to dispersion. However, dispersion can also result in weak economies of scale, limited access to capital, fragmented markets, and difficulties in attracting talent. Therefore, the same geography can be viewed in two ways: as an asset for place-based embedding, or as an obstacle to global competitiveness. The challenge lies in connecting specialised regional ecosystems to European value chains, data spaces, investment networks, and applied AI markets.

If Europe’s AI geography is polycentric by nature, its competitiveness cannot rely solely on a few metropolitan growth poles. Competitiveness also depends on Europe’s capacity to connect industrial regions, energy systems, research ecosystems and digital infrastructures across borders. In this context, the future of European AI competitiveness may depend as much on territorial integration as on technological innovation itself. In other words, Europe needs more than just stronger hubs. It needs better-connected hubs.

Europe’s AI challenge is not only technological

This issue mirrors the concerns raised in both Mario Draghi’s report on European competitiveness (Abre numa nova janela) and Enrico Letta’s report, 'Much More Than a Market' (Abre numa nova janela). Draghi argues that Europe’s core weakness is not a lack of innovation per se, but its difficulty in scaling it up and transforming it into productivity growth. Letta stresses that Europe’s future competitiveness depends on stronger integration of infrastructure, energy, finance, knowledge, and innovation systems.

The geography of AI makes these challenges highly visible. Whether regional AI ecosystems can scale, whether startups remain anchored in Europe and whether applied AI becomes a source of territorial development or another driver of spatial concentration depends on access to growth capital, compute, talent, data and cross-border market integration.

European policy needs a more territorial understanding of AI competitiveness

Current initiatives on AI factories, data spaces, supporting start-ups and scale-ups, computing infrastructure, and market integration are important. However, their effectiveness should also be judged by how they shape Europe’s geography of opportunity. Do they only reinforce the largest hubs? Do they connect smaller, specialised ecosystems? Do they help industrial regions access AI capabilities? Do they enable regions outside the usual innovation centres to participate in applied AI markets?

Europe’s fragmented AI geography can become strength if it is organised as polycentric specialisation: different places developing different AI capabilities, connected through European markets, infrastructures and institutions. Therefore, the report's key message is that the future of European AI competitiveness may depend not on creating one Silicon Valley, but on fostering many specialised places.

by Kai Böhme

(Abre numa nova janela)