A dramatic geographic shift is reshaping Europe’s data centre landscape. Hyperscale facilities scheduled for completion between 2026 and 2028 will sit an average of 175 kilometres from major EMEA cities, nearly quadruple the 46-kilometre average seen in projects delivered from 2022 to 2025, according to research from JLL.
The catalyst? Surging power requirements from AI infrastructure. Grid bottlenecks, planning hurdles, and vanishing land availability in established markets now push developers toward regions where electricity actually exists. The four largest hyperscale cloud providers plan to spend $725 billion in capital expenditure during 2026, a 77% jump from $410 billion this year, with much of that funding targeting AI compute and supporting facilities. By 2030, AI workloads could consume roughly half of global data centre capacity.
**Power now dictates location**
Assad Noori, Head of Data Centres at JLL for EMEA Work Dynamics, points to a fundamental planning reversal. Where proximity to population hubs once ruled site selection, securing reliable power now drives decisions for AI training infrastructure. Builders increasingly follow the electricity, not the demand.
Eva Sóley Guðbjörnsdóttir, CFO and Deputy CEO at atNorth, echoes this observation, noting how site selection criteria have evolved from enterprise computing into the AI era. Training workloads tolerate latency far better than traditional enterprise applications, opening doors to locations previously considered impractical.
Greenfield sites now represent 39% of Europe’s future pipeline, a steep climb from just 8% for projects delivered in the previous three years.
**Core markets still expand**
Meanwhile, the FLAP-D markets of Frankfurt, London, Amsterdam, Paris, and Dublin continue their growth trajectory. Combined live capacity across these five cities has reached 3.8 GW, more than doubling since 2019. Another 1.4 GW sits under construction, with 2 GW in planning. Annual deliveries should hit 453 MW in 2026, nearly triple 2020 levels. Vacancy remains tight at 6.4%, with Frankfurt tightest at 3.1%.
Martin Jensen, EMEA Division President for Data Centres at JLL, says core markets will stay essential for latency-sensitive enterprise demand. However, hyperscale AI infrastructure requires power and land that urban hubs cannot provide, accelerating investment into secondary markets and entirely new regions.
**Middle East follows similar path**
The Middle East mirrors this dispersion trend. JLL reports 2.6 GW under construction and 13.8 GW in planning there, with total capacity projected to quadruple by 2030. Sovereign-backed projects continue moving forward despite regional tensions, including HUMAIN’s planned 100 MW AI campus in Dammam and the first 200 MW phase of Stargate UAE. Timelines may shift, according to Daniel Thorpe, Head of Data Centre Research for EMEA, but intentions remain firm.
