On August 19, Reuters published JLL pipeline data on where Europe’s hyperscale data centres will land over the next two years.
Sites coming online between 2026 and 2028 sit an average of 175 km (109 miles) from a major hub. Sites delivered between 2022 and 2025 averaged 46 km.
More than three times further out.
Assad Noori, who runs JLL’s data centre business across EMEA, named the mechanism without hedging it: the determining factor is increasingly where sufficient power can be secured, rather than simply where demand exists. Data centres are being brought to where the power is, not the other way around.
The supporting numbers
Greenfield accounts for 39% of Europe’s forward pipeline against 8% of delivered projects.
Inner-city locations drop to 5% of the pipeline from 13%, with the balance in industrial or edge-of-city sites.
DC Byte, tracking early-stage projects, counts nine proposed gigawatt-plus campuses in Europe. One sits near a major city, Paris. The rest run from rural Spain to northern Sweden.
The capital behind it is in the same dataset. JLL estimates the four largest hyperscale cloud providers will spend $725B in 2026, up 77% from $410B in 2025, mostly on AI compute and data centre infrastructure. By 2030 the firm expects AI workloads to account for roughly half of global data centre capacity.
Land prices explain the exodus. Powered land runs €2.36M per megawatt of IT load in core markets: Amsterdam around €2.7M, London €2.6M, Frankfurt €2.5M. Secondary cities including Copenhagen, Warsaw and Milan drop to €978K. Tertiary areas such as Bordeaux come in at €512K, and can go as low as €200K.
A tenfold spread on the same megawatt.
The core markets do not empty out
Nothing in the data says London and Frankfurt stop mattering. Martin Jensen, who heads JLL’s EMEA data centre division, drew the line between two different businesses: Europe’s core markets remain critical because enterprise demand is not going anywhere, while hyperscale AI infrastructure requires a completely different scale of power and land.
That is the split worth holding onto. Enterprise colocation stays in the city, close to the customers and the network. AI training campuses leave, because a training run cares about megawatts and cooling water and nothing else.
Rupert Duckworth of Savills described the London constraint in physical terms: the market has already absorbed heavy cloud-driven build-out, other asset classes compete for the same land at high prices, and power is now constrained across the key cloud locations. Grid connection queues do the rest. A developer choosing between a multi-year wait for a London connection and a shovel-ready site in tertiary France is not making a close call.
What sits inside the 129 km
Translate this into labour-market terms.
This capex cycle is doing two things at once. It is subtracting: back-office, service and junior analytical roles in London, Frankfurt and Amsterdam are being thinned a layer at a time by models and agents. It is also adding: $725B has to be dug, poured, wired, cooled and connected, and every one of those is real work for real trades.
Those two things now happen 175 km apart.
The roles being removed are in the city. The roles being created are in rural Spain and northern Sweden. Someone who has spent eight years reconciling accounts in Canary Wharf does not change commute because a campus opened in the Nordics. There is no transfer path between those two labour markets.
Reuters puts the upside carefully: the trend can mean investment in underprivileged areas where governments are trying to stimulate jobs and growth, though developers also risk opposition from residents concerned about dwindling habitats and competition for power and water.
“Stimulate jobs” is the phrase worth opening up.
How data centre employment is actually shaped
Data centres are capital-intensive, not labour-intensive. Investment announcements tend to skip that line.
Headcount on a hyperscale campus splits into two phases. Construction is dense: civils, steel, mechanical and electrical, high-voltage, fire systems. A large project peaks in the hundreds to low thousands and runs two to four years. Operations is thin: a campus of a few dozen megawatts typically carries a permanent team in the tens, concentrated in HV electricians, mechanical and cooling technicians, infrastructure management, security and site operations.
The employment curve is a spike with a long thin tail. What a region durably keeps is mostly tax base and a grid upgrade, not a payroll.
None of which argues the projects are not worth doing. It argues that when an announcement puts “creates X,000 jobs” in the headline, the first question is whether that is peak construction or permanent establishment. Those two numbers differ by an order of magnitude.
Where this fits
We have been tracking this thread.
Microsoft cut 4,800 roles inside a record AI capex quarter on July 8. Amazon posted $62B of profit while pushing capex to $220B on July 31. Both describe capital moving from payroll to plant inside a single company.
Korea’s AI levy, which bills the employer rather than the vendor, passed August 14 as one fiscal attempt to claw some of it back.
The JLL data adds the spatial axis. Money moves off the payroll and into the building, and the building then moves out of the city. Every step widens the distance between where the capital lands and where the displaced worker lives.
What to watch, by role
Urban back-office and junior white-collar. This cycle does not turn back into hiring in your city. Power and land in London, Frankfurt and Amsterdam are constrained enough that JLL wrote it into the report, and the new capacity is not going there. Hedging your own role against “AI investment will lift local hiring” is a bet this dataset closes off in Europe’s core markets.
HV electricians, cooling and mechanical technicians, M&E construction management. This is the genuine demand side in the data, and it prices mobility. The work is in Bordeaux, inland Spain, northern Sweden. It is not where you currently live. For anyone willing to relocate, it is one of the most legible skills demand curves in Europe over the next three years.
Regional employment forecasting. When a campus is announced, separate peak construction headcount from permanent establishment. The first disappears in two to four years. The second is what the region keeps.
Timing. JLL’s window is 2026 to 2028 for delivery, which puts the construction employment peak in the next two years and permanent operations roles opening progressively from 2027.
External source: Reuters (Simon Jessop, Iain Withers), “Europe AI data centres seek cheaper, quicker energy and land,” August 19, 2026
