Europe’s AI Data Centres Are Moving Toward Power, Not Population Centres

European AI data-centre development is becoming power-led rather than city-led. Hyperscale AI campuses need unusually large and dependable electricity connections, substantial land, cooling infrastructure and high-capacity fibre. In Europe’s established hubs, those inputs are increasingly scarce, expensive or slow to secure, making secondary and greenfield locations more attractive. 14

The shift is measurable

Projects planned for delivery between 2026 and 2028 are expected to sit an average of 175 kilometres from a major EMEA hub, compared with 46 kilometres for projects delivered between 2022 and 2025. That is almost four times farther. Greenfield projects are also projected to account for 39% of the 2026–28 pipeline, compared with just 8% of completed projects. 17

The move outward does not mean Europe’s traditional FLAP-D markets—Frankfurt, London, Amsterdam, Paris and Dublin—are being abandoned. Their combined live capacity has reached 3.8 GW, and they remain important for connectivity, enterprise customers and latency-sensitive workloads. The emerging pattern is a split: core hubs retain proximity-dependent activity, while large AI training and cloud campuses seek room and power elsewhere. 47

Europe’s wider data-centre market is still expanding rapidly. CBRE expects total capacity to reach 13 GW by the end of 2026, a 20% year-on-year increase, while new colocation signings intended for AI more than quadrupled during the first half of 2026. 12

Why power is outweighing proximity

Electricity availability is the critical constraint

AI workloads rely on dense GPU clusters that require large, continuous power supplies. In metropolitan markets, grid congestion, land shortages and planning restrictions can delay new capacity. A site with a credible, earlier connection to high-voltage infrastructure may therefore be more valuable than one located close to a large customer base. 14

This is changing the order of priorities in site selection. Developers are increasingly looking first for usable electrical capacity and connection speed, then assessing fibre routes, transport, workforce availability and customer access.

Peripheral sites offer more land and flexibility

Greenfield and secondary locations can provide larger parcels, lower powered-land costs and more room for phased expansion. One reported JLL estimate puts powered land at an average of €2.36 million per megawatt of IT load in Europe’s core markets, falling to about €978,000 in secondary cities. The figures are estimates rather than universal prices, but they illustrate why the economics can favour sites beyond the traditional hubs. 11

Less-constrained locations may also offer more flexibility around permitting and campus design. That matters for AI facilities, which need space for electrical equipment, cooling systems, backup capacity and future expansion—not simply a building that can fit on an urban plot.

AI campuses have heavier physical requirements

The shift is not just a real-estate decision. High-density AI facilities need substantial power delivery, cooling capacity, electrical redundancy and resilient connectivity. A remote site is viable only if it can combine those elements with reliable fibre, construction capacity and a credible approach to renewable-energy procurement. 14

Cheap rural land alone is not enough. If a site lacks transmission capacity, diverse network routes, water or cooling options, and a workable permitting path, its apparent cost advantage can quickly disappear.

Where Europe’s next campuses are spreading

Development is broadening toward Iberia, Milan and the Nordic countries, while Madrid, Milan, Marseille and Athens are among the southern European alternatives cited as having more room for growth. Scandinavia is also attractive in some cases because of its renewable generation and cooler climate. 67

The result is a more distributed European data-centre map:

• FLAP-D markets continue to serve enterprise demand, dense network interconnection and workloads that benefit from proximity to users.
• Secondary cities can offer better access to land, power and permitting capacity.
• Large greenfield campuses are increasingly suited to AI training and other workloads that can tolerate greater distance from population centres.

This is better understood as an expansion of Europe’s infrastructure footprint than a simple migration out of cities. The core hubs remain active, but new capacity is being added where the physical inputs are available. 47

Investment and regional economic opportunity

The investment cycle is large. The European Data Centre Association projects €176 billion in cumulative data-centre investment between 2026 and 2031, with commercial colocation and hyperscale facilities accounting for more than two-thirds of Europe’s IT power. 1416

Host regions could benefit from:

• Construction and civil-engineering activity
• Grid and substation upgrades
• Fibre and telecommunications deployment
• Renewable-energy development and procurement
• Specialist operations and maintenance
• Cooling, electrical and engineering suppliers
• Waste-heat recovery and district-heating projects
• Training, education and wider digital-services investment

The broader economic contribution is also significant. An industry estimate cited by Data Center Knowledge places European colocation’s GDP contribution at approximately €30 billion in 2023, with forecasts approaching €100 billion by 2030. 3

However, local employment claims need to be handled carefully. Hyperscale facilities are highly capital- and power-intensive, and the number of permanent jobs after construction may be modest relative to the scale of investment. The strongest regional benefits are more likely where a data-centre project attracts associated energy, engineering, supplier, cloud and skills investment rather than operating as an isolated facility.

The environmental and community trade-off

Moving campuses away from dense cities can reduce pressure on urban land and metropolitan grids. It may also make it easier to connect facilities directly with renewable generation or develop waste-heat projects. But relocation does not remove environmental impacts; it can redistribute them to rural areas and smaller communities.

The European Commission identifies rising electricity demand, cooling-water requirements and emissions from non-decarbonised power supplies as key challenges for data centres. The EU is also developing a common data-centre rating scheme and minimum energy-efficiency standards as the sector expands.

Local communities may face competition for:

• Grid capacity needed by homes and existing businesses
• Water resources and cooling infrastructure
• Land and transport capacity
• Renewable-energy projects
• Local visual, noise and environmental amenities

For that reason, social licence will become part of site selection. Developers and public authorities will need transparent assessments of grid impacts, clear water and cooling plans, credible clean-power arrangements, controls for noise and backup generators, and enforceable local benefits.

What the 2026–28 pattern means

Europe’s AI infrastructure is not simply leaving its major cities. Instead, the market is separating into two complementary models: connected urban hubs for latency-sensitive and enterprise workloads, and large power-oriented campuses in locations with more land and faster access to electricity.

The decisive competitive advantage is increasingly timely, dependable and increasingly clean power. Regions that can combine grid capacity with fibre, permitting, cooling, skilled suppliers and community support are best positioned to capture the next wave of AI investment. Those that secure only the building, without addressing water, emissions and local infrastructure, may find that the economic opportunity comes with lasting political and environmental costs.