Europe’s next generation of artificial intelligence infrastructure is spreading far beyond the established data centre districts of Frankfurt, London, Amsterdam, Paris, and Dublin, as developers find that proximity to a major city counts for less when a project requires hundreds of megawatts of electricity. Research from JLL shows that hyperscale campuses planned between 2026 and 2028 are an average of 175 kilometres from a major EMEA hub, compared with 46 kilometres for projects delivered between 2022 and 2025.

The shift does not amount to an abandonment of Europe’s traditional markets, whose combined live capacity has reached about 3.8GW after more than doubling since 2019. Another 1.4GW is under construction and roughly 2GW is planned, while operators continue to compete for scarce space in the largest cities. Instead, the market is splitting according to the physical requirements of different workloads, with AI training increasingly willing to travel towards power while latency-sensitive services remain closer to users.

That distinction is changing the economics of site selection because very large training clusters can require 100MW or more without needing to sit beside a major financial or commercial centre. JLL says greenfield developments account for 39% of the 2026–2028 pipeline, compared with 8% of projects delivered during the preceding three years, as developers seek larger plots where electricity can be secured on terms that make the project viable.

Land costs reinforce the incentive to look further afield, although cheaper property is useful only when power is available. Prime powered land across the five main European markets has risen from €1.24 million per megawatt in 2021 to €2.26 million, while primary locations now command a substantial premium over secondary and tertiary markets. A remote site without a credible grid connection can still be economically useless, which leaves electricity infrastructure doing much of the work once performed by metropolitan geography.

Power divides the market

Frankfurt illustrates the constraint particularly clearly because its colocation vacancy rate sits at about 3.1%, while grid connection lead times can extend for years. Paris, by contrast, added 72.5MW during the first half of 2026, benefiting from France’s comparatively strong electricity position and growing demand for AI capacity. That same infrastructure advantage has already encouraged Orange to turn parts of its French telecoms estate towards AI and cloud capacity.

The newer locations competing for hyperscale projects include parts of the Nordics, Iberia, Italy, and other secondary European markets where grid access, renewable generation, land, and political support can be combined more readily than in mature city clusters. JLL expects more than half of Europe’s AI-related growth to be captured by Nordic and Tier 2 locations, although those projects will still depend on high-capacity fibre links and reliable energy rather than operating as isolated computing islands.

As development spreads, the economic geography of digital infrastructure changes with it. Large campuses can bring construction work, fibre investment, new substations, business rates, and specialist employment into regions that have historically sat outside Europe’s main technology corridors, yet they are capital-intensive assets whose permanent employment footprint is relatively modest compared with their electricity demand.

That imbalance is likely to make local scrutiny more demanding rather than less. Grid investment, water use, land allocation, renewable generation, and the opportunity cost of electricity all become part of negotiations around a project that may serve customers hundreds or thousands of kilometres away. Communities hosting infrastructure for a continental digital economy will increasingly expect more than the argument that computing capacity is strategically important.

The old hubs are not disappearing

Established data centre locations retain advantages that are difficult to reproduce elsewhere because they combine dense connectivity, carrier networks, customers, technical skills, and existing infrastructure. Financial systems, enterprise applications, content delivery, and AI inference services serving large populations can still benefit from being close to major cities, while companies frequently need capacity across several locations rather than choosing one market at the expense of all others.

The distinction therefore lies in which workloads generate the next very large increments of demand. Training a frontier-scale AI system can consume enormous quantities of power without requiring millisecond proximity to an end user, which gives developers more freedom to follow grid capacity into regions where industrial-scale electricity and land can be assembled.

Europe’s data centre map is consequently broadening rather than being replaced. Frankfurt, London, Amsterdam, Paris, and Dublin remain central to the continent’s digital economy, but the largest AI campuses are beginning to follow substations, generation, and transmission capacity beyond those established boundaries. As computing becomes more electricity-intensive, the location of Europe’s AI economy will be shaped increasingly by the physical energy system beneath it.