Europe's Data Centers Just Fled the Big Cities — and Power Drew the New Map

Europe's Data Centers Just Fled the Big Cities — and Power Drew the New Map

Let me tell you something that's been sitting with me since the numbers crossed my desk this week. For twenty years, the rule of European data centers was simple: build where the people are. Frankfurt. London. Amsterdam. Paris. Dublin. That was the gospel — latency, customers, connectivity, talent. You didn't question it. You picked a hub, you paid the premium, and you got on with it.

That rule just died. The data from CBRE and JLL this week shows 63 percent of Europe's new data center capacity is now landing somewhere other than the big five markets. The industry didn't drift away from the hubs — it got pushed. And the thing that pushed it isn't what most people think. It's not money. It's time. I've been running hosting infrastructure for over a decade, and I can tell you the day you can't get power is the day your location strategy stops being a strategy and starts being a prayer.

The News — The Map Just Changed Under Everyone's Feet

Here's what happened while we were all staring at GPU allocations and chip lead times. Europe's data center capacity is on track to hit 13 gigawatts by the end of 2026 — a fifth more than last year, according to CBRE's European update published this week. But of the roughly 700 megawatts added in the second quarter, only 37 percent went to the five established markets. More than half of all take-up happened outside Frankfurt, London, Amsterdam, Paris, and Dublin. CBRE's own words: "Demand continues to outpace supply, mostly due to a lack of available power causing delays in data center construction."

JLL's EMEA mid-year report tells the same story from the real estate side. Greenfield developments have jumped from 8 percent of the 2026-2028 pipeline to 39 percent. Hyperscale greenfield campuses now sit an average of 175 kilometers from a hub city — up from 46 kilometers. Think about that. The average new AI campus is nearly four times farther from a major city than it was three years ago. This isn't a corner of the map getting crowded. The map itself is being redrawn.

Assad Noori, head of data centers for EMEA at JLL, put it better than I could: "Historically, operators built as close as possible to major population centers, but today's growing AI training infrastructure follows a different logic. The determining factor is increasingly where sufficient power can be secured." Power. Not people. Not latency. Not even customers. Power.

The Numbers — Follow the Megawatts, Not the Hype

Let me give you the numbers that matter, because the headline numbers only tell half the story. Prime powered land in Europe now costs €2.26 million per megawatt — up 82 percent since 2021. Primary markets carry a 2.3-times premium over secondary sites and four times over tertiary ones. So the big hubs aren't cheap. They were never cheap. What changed is that the premium stopped buying anything.

Look at what's happening in the places with actual grid capacity. Neocloud signings — the AI-focused providers — hit 420 megawatts in the second quarter against 89 megawatts a year earlier. That's a 4.7-fold increase, and CBRE says it's concentrated in the Nordics, where renewable power is cheap. Spain installed 439 megawatts by the end of 2025 and the industry association there forecasts roughly 2,537 megawatts by 2030, with Amazon alone committing €33.7 billion in Aragon. In Finland, Nebius is building an €8.5 billion, 310-megawatt site at Lappeenranta. Stargate Norway is starting at 230 megawatts on hydropower near Narvik, with another 290 planned. Even Uniper is committing €5 billion to repurpose old power plant sites — because those sites already have grid connections, and that's now the most valuable real estate in Europe.

Meanwhile, the hyperscalers are just building for themselves. Self-build across Europe now stands at 4.3 gigawatts, up 22 percent — a seventeenth consecutive year of double-digit growth. Roughly 70 percent of that operational self-build sits in Ireland, the Netherlands, Sweden, and Belgium. The old markets still hold the installed base. The new money is going elsewhere.

The Two Readings — A Rational Migration, or a Boom With a Hangover Coming

There are two ways to read this, and the honest answer is that both are true at the same time. This is the dual-aperture moment, and I want you to hold both lenses up before you make any decisions.

The first reading: this is the market behaving rationally. AI training workloads do not need to be in Frankfurt. A training run doesn't care about city latency — it cares about megawatts, water, and land at a price that doesn't bankrupt the project. When the core markets can't deliver power for a decade, developers go where the power is. That's not panic. That's physics. The Nordics, Iberia, Italy, Belgium — these are the places with headroom, and the industry is following the grid. If you're an independent hosting provider, this is the biggest structural opportunity in a decade: a whole new set of markets opening up with cheaper power, cheaper land, and less regulatory baggage.

The second reading: every migration like this ends the same way. JLL says colocation vacancy in the five core markets actually fell slightly last quarter, to 6.4 percent. But vacancy everywhere else in Europe is forecast to hit 19 percent by the end of the year. That's what a building boom looks like when it happens somewhere new — a lot of shovels, a lot of concrete, and a lot of capacity coming online at the same time. When AI demand wobbles — and it will wobble — the first casualty is the newest market, not the oldest. I've watched this movie before. It's the same plot every time: follow the cheap power, build fast, then discover that everybody else followed the same cheap power.

And there's a third thing neither reading mentions: the core markets aren't dying. They're just being redefined. Latency-sensitive enterprise workloads stay in Frankfurt and London. The people who need single-digit milliseconds — trading, real-time applications, the enterprise customers you actually bill monthly — are not moving to Lappeenranta. What's leaving is training. What's staying is the stuff that pays the recurring bill.

The Secondary Bottleneck Nobody's Talking About — You Can't Buy Your Way Out of a Decade

Here's the part of this story that should scare every operator, investor, and policymaker in Europe. The bottleneck was never land. It was never capital. It was never even chips. It's the grid connection queue — and you cannot write a check big enough to fix it.

Connecting a new site in the congested core markets can take ten years or more, according to RaboResearch analysts Owen Thomson and Coen Hutters. Amsterdam is the extreme case at roughly a decade — the Dutch grid operator TenneT has said there is no additional capacity in much of Noord-Holland for the next ten years, and the Netherlands bans hyperscale projects above 70 megawatts of IT capacity or ten hectares across most of the country. Frankfurt's central grid won't see meaningful upgrades until the 2030s. London's West London substation work is unlikely before early next decade — and Nscale's flagship Essex site literally cannot switch on because of it. Denmark's Energinet paused new connections earlier this year after its queue hit 60 gigawatts against a national peak demand of about seven.

Read that last number again. Sixty gigawatts of queue against seven gigawatts of peak demand. That's not a queue. That's a parallel universe. Every one of those projects is someone's business plan sitting in a holding pattern, paying interest, waiting on a transformer and a substation that may not exist this decade. I wrote about phantom gigawatts in the US interconnection queue back in August — this is the same disease on the other side of the Atlantic. The grid is the wall, and the queue is the line of people waiting to run into it.

And Ireland is the cautionary tale for everywhere else. The connection moratorium ended in December 2025, but the price of entry is behind-the-meter generation sized to the full grid connection, siting in unconstrained locations, and 80 percent of annual demand matched by renewable investment. That's not a welcome. That's a toll — in a country where data centers already draw more than 22 percent of national electricity. Every market that opens its arms to this buildout is going to write the same rules in a few years, after the voters notice the bills.

What This Means for Independent Hosting Providers

If you run an independent hosting or colo business, this week's numbers are a roadmap, not a weather report. Here's what I'd be doing.

First — follow the power, but verify it exists. The opportunity is in secondary markets with actual grid headroom: Spain, Portugal, the Nordics, parts of Italy and Belgium. But "available power" in a press release is not the same as a signed interconnection agreement. Do your own due diligence on the queue. If the utility can't tell you a connection date in writing, assume it's a decade.

Second — don't build your business model on the AI training migration alone. That 19 percent vacancy forecast outside the core is the market telling you oversupply is coming to the new markets. AI tenants are fickle — they follow incentives, they renegotiate, they leave when the capex cycle turns. The durable revenue is the latency-sensitive enterprise workload that has to stay in the core. That's your recurring bill. That's the mortgage payment.

Third — treat powered land as the scarce asset it now is. Prices are up 82 percent since 2021 and the premium for primary markets over secondary is 2.3 times. If you're going to build anywhere, secure the power first, then the land — not the other way around. The deal that looks cheap today in a tertiary market is cheap for a reason, and the reason usually involves a substation that doesn't exist.

Fourth — price in the grid, not just the kilowatt. European industrial power already runs at roughly twice US levels on IEA figures. Add the cost of on-site generation, backup, and behind-the-meter infrastructure, and the margin math changes completely. Model your pricing against a world where the queue makes every new build slower and more expensive — because that's the world we're in.

The Bottom Line

The map of European infrastructure just changed, and it changed because of megawatts, not because of any CEO's grand strategy. AI training needs power at a scale the old hubs can't deliver, so the industry is packing up and moving to where the electrons actually are. That's rational. That's also fragile — because when everyone moves to the same new place at the same time, you get a boom, then you get a vacancy report, then you get a correction.

Here's the truth bomb: the companies that win this migration won't be the ones with the biggest capex. They'll be the ones who understood that the grid queue is the real asset — and that you can't buy your way out of a decade. The map is redrawn. The question now is whether you're building on the side of it that has power, or the side that has a line of people waiting for it. Me? I'm checking the queue first. Ent?

— Allan Ali, Founder

This article was produced with AI-assisted research and editorial support. Sources: Reuters (Aug 19, 2026), CBRE European Data Centre update via The Next Web (Aug 19, 2026), JLL EMEA Data Centre Mid-Year 2026 Report, RaboResearch, CNBC International.