JLL’s 175km Shift Is Repricing European Data-Centre Land
Europe’s next AI data-centre sites now sit 175 kilometres from a major hub, up from 46 kilometres. Pay city prices for land without power and you have bought an expensive paddock.
Europe’s next AI data-centre sites now sit 175 kilometres from a major hub, up from 46 kilometres. Pay city prices for land without power and you have bought an expensive paddock.
The property market has moved 175 kilometres away
Reuters reported on August 19 that Europe’s next wave of hyperscale AI data-centre projects is moving dramatically farther from major cities. JLL data show sites due online between 2026 and 2028 sit, on average, 175 kilometres from a major hub. Projects delivered from 2022 to 2025 averaged just 46 kilometres away.
That is not a minor shift in site selection. It is a complete rewrite of the old logic of valuable commercial land.
For decades, the property industry prized proximity: close to CBDs, airports, fibre routes, affluent consumers, corporate headquarters, ports, transport corridors. Those things still matter. But AI training campuses have introduced a much harder constraint: enormous, reliable power capacity. If you cannot get the megawatts, your beautifully located industrial site is just an expensive paddock with a brochure.
JLL’s Assad Noori put it plainly: data centres are going where the power is, rather than power being brought to where demand exists. That is the whole story.
The world’s four biggest hyperscale cloud providers are expected to spend $725 billion in 2026, up from $410 billion in 2025, largely on AI computing and data-centre infrastructure, according to the JLL figures cited by Reuters. Whether every dollar earns a sensible return is a separate question. But the capital is clearly being committed, and property markets are already being rearranged around it.
The expensive old core markets — London, Frankfurt, Amsterdam, Paris and Dublin — will remain important. Enterprise customers are there, fibre ecosystems are there, and nobody is abandoning them. But the next big AI campuses need land, grid access and development timelines that those markets increasingly cannot provide.
That means regional locations, industrial fringes and previously ignored towns are suddenly in the conversation.
The real asset is not the building
Here is the mistake I expect to see repeated: investors will see “data centre” and buy any industrial property with a decent roofline and an optimistic agent.
Don’t.
A data centre is not a warehouse with servers in it. The building matters, but the scarce asset is the package around it: secured power, grid connection timing, planning approval, fibre access, cooling feasibility, water strategy, local political support and a customer willing to sign a long lease.
Take the land-cost figures in the Reuters report. JLL estimates powered land costs an average €2.36 million per megawatt of IT load in Europe’s core markets, versus €978,000 in secondary cities such as Copenhagen, Warsaw and Milan, and €512,000 in tertiary areas such as Bordeaux. In some tertiary locations, it can be as low as €200,000 per megawatt.
That gap is why developers are prepared to look far beyond the familiar hubs. Amsterdam was cited at roughly €2.7 million per megawatt, London at €2.6 million and Frankfurt at €2.5 million. If you need hundreds of megawatts, the difference is not a rounding error. It is the deal.
Tritax Big Box REIT gives us a useful real-world example. In its August 5 half-year report, the UK logistics landlord said it had nearly doubled secured power for its data-centre pipeline to 507MW. Its Manor Farm site near Heathrow has 107MW planned for 2027 and another 40MW for 2029; its Chelmsford project has an initial 125MW scheduled for 2028. The company is not simply buying land and hoping. It is building a power-first development pipeline.
That is the lesson: a conventional property investor asks, “What is this land worth?” A serious digital-infrastructure investor asks, “What is the delivered power worth, when is it available, and who can actually use it?”
Those are very different underwriting models.
Greenfield land is winning because cities are choking
The numbers show how quickly the market is changing. Greenfield projects make up 39% of Europe’s future data-centre pipeline, compared with only 8% of delivered projects, according to the JLL data cited by Reuters. Inner-city sites are expected to make up just 5% of the pipeline, down from 13% of completed projects.
That is a big deal for regional property markets. An area with cheap land alone is not suddenly a winner. Plenty of cheap land is cheap for a reason. But an area with available generation, transmission capacity, council support, fibre routes and fast planning has a chance to become strategically important.
It also means local governments will be tempted to throw the doors open. They see construction jobs, rates revenue, technology investment and a shiny “AI hub” headline. Fair enough. But they need to negotiate like adults.
Data centres can be tremendous economic anchors. They can also consume huge amounts of power and require careful cooling and water planning, while creating fewer permanent jobs than a big manufacturing plant. Residents understand that instinctively, which is why opposition is building in various markets.
The overlooked risk is political, not technological. A developer can model construction costs. It is harder to model a community campaign, a new planning restriction, a grid-priority change or a government deciding that households and factories get first call on constrained electricity.
If you are investing in this theme, never treat a grid connection as a simple line item. Treat it as a moat that can disappear if the politics turn against you.
The $100 billion warning sign
There is another reason to stay clear-eyed: the money rushing into the sector is enormous.
Reuters reported on August 13 that Silver Lake and DigitalBridge-backed Vantage Data Centers was exploring an IPO or sale that could value the company at around $100 billion. A listing could raise roughly $10 billion and, if completed at that scale, would be the largest data-centre IPO to date. Reuters also made clear that talks were preliminary and that no transaction was guaranteed.
That last bit matters. A reported valuation is not a completed deal, and a big valuation is not proof that every data-centre project is sensible.
Still, the Vantage story tells you what capital markets are pricing in: data-centre operators are no longer being viewed as dull landlords. They are being valued as owners of critical AI infrastructure.
Sometimes that will be justified. Long leases to creditworthy hyperscalers, contracted capacity, limited supply and genuine power scarcity can produce excellent assets.
Sometimes it will be nonsense. A mediocre site with speculative capacity, expensive debt and no committed tenant is not “AI infrastructure.” It is development risk with fashionable branding.
I have seen this movie in other sectors. The narrative gets so strong that people stop distinguishing between the best asset and any asset with the right label. That is where the money gets lost.
The contrarian angle: boring utilities may beat glamorous property
Everyone wants to own the data centre because it sounds like owning a piece of the AI future.
But the better risk-adjusted exposure may often sit beside it rather than inside it.
The scarce inputs are power equipment, grid connections, transmission infrastructure, backup systems, cooling, transformers, construction capacity and fibre. Some of those businesses are less exciting at a barbecue. They may also be less exposed to a single tenant, a single building or a single overpaid land acquisition.
The same applies to property. I would rather own a well-bought industrial or regional land position with a credible path to power and planning than a “data-centre themed” asset bought at a silly yield because the seller put AI in the pitch deck.
And I would not dismiss traditional logistics. Tritax’s results are a reminder that good logistics property still has a job to do. But the distinction is now sharper: generic space is a commodity; strategically located, power-enabled land is not.
That is why the next decade will not reward passive property thinking. It will reward operators who understand infrastructure constraints.
What this means for you
If you are a founder, investor, developer or simply someone allocating capital, here is the practical version.
First, stop using “data centre” as an investment thesis. It is a category, not an answer. Ask what the site has that cannot be copied: contracted power, a connection date, planning permission, fibre, customer demand or a development partner with a real track record.
Second, separate secured power from hoped-for power. A press release saying a site is “well positioned” means nothing. Find out the megawatts, the grid operator, the delivery timetable, the conditions attached and who bears the cost if timing slips.
Third, watch secondary and tertiary markets — but do the hard work. The 175-kilometre shift creates opportunity outside the obvious capitals. It also creates traps. Cheap land without power is still cheap land.
Fourth, demand tenant quality and contract quality. A long lease with a serious counterparty is worth more than a speculative development model that assumes AI demand will bail everyone out forever.
Finally, do not confuse a shortage with permanent pricing power. Today’s grid bottleneck is real. But governments, utilities and developers respond to scarcity. The winners will be the people who buy assets with durable advantages, not those who arrive last and pay top dollar for the story.
The property market is being pulled away from city centres and toward electricity. That is not a trend to admire from a distance. It is a change in the map of value. Learn to read the new map before someone sells you the old one.