Finland's data centre boom and the queue for the grid

At the end of March 2026, Nebius announced it would build a 310-megawatt AI factory in Lappeenranta, Finland. The investment was put at over ten billion dollars, and on completion it would be among Europe's largest. Capacity comes into use in stages from 2027.

The news sounds like an isolated event. It is not. Finland is in the middle of a build-out whose scale is hard to grasp from press coverage, because that coverage contains three kinds of number that mean entirely different things.

Three numbers that get mixed up

Enquiries. Fingrid, which operates Finland's transmission grid (the backbone of electricity transmission) has received enquiries about connecting data centres amounting to over 50,000 megawatts. That is many times Finland's entire current peak load. An enquiry costs nothing and commits to nothing, and the same project can enquire about several locations.

Connection agreements. By the end of June 2026, the planned final total scale of data centre projects that had signed a connection agreement was, on the project developers' own figures, over three gigawatts. An agreement is a commitment and it costs money. Fingrid signed around 50 new connection agreements during the year, roughly a third of them for data centres.

Built capacity. This is the smallest number and the only one that consumes electricity.

Jussi Jyrinsalo, Fingrid's director responsible for transmission grid planning, has estimated to Yle that around a tenth of the enquiry volume is realised, on the order of five gigawatts. Even that is a lot: Finland's peak electricity consumption is currently around 15 gigawatts.

When you read a story mentioning tens of gigawatts, ask first which number it is. The difference between an enquiry and an agreement is roughly the difference between attending a viewing and signing the deed.

Where the real bottleneck is

Not the availability of electricity, but time and the grid connection.

A data centre is built in roughly 18 to 36 months. A new high-voltage line or substation takes years, because it involves design, permitting, land acquisition and appeal periods. The appetite to invest therefore outruns the pace at which plants can be connected.

Two consequences are already visible. Where the connection points are starts to matter more than the price of the site, and projects move to where the grid is. Yle has reported that southern Finland's grid cannot carry all the planned projects and that Fingrid has had to restrict the connection of large consumers. The electricity exists, but in the wrong places.

Second, being in the queue is itself a competitive advantage. Whoever signed a connection agreement two years ago is building now, and whoever enquires today will not be building on that spot this decade.

Finland's electricity consumption in the first half of 2026 totalled 46.1 terawatt-hours, 6.3 per cent more than the year before. The growth is explained particularly by data centre and electric boiler projects.

Why Finland

The reasons are prosaic and have nothing to do with AI.

• A cool climate reduces the share taken by cooling, which is a data centre's second-largest consumer of electricity.
• Low-emission and relatively stable generation, with nuclear, hydro and wind covering a large part of it.
• Space and industrial sites that already have electricity infrastructure left over from heavy industry.
• A stable legal system, which matters a great deal in a ten-year investment.

The growing share of wind brings the opposite phenomenon too: generation varies, whereas a data centre consumes steadily around the clock. A large steady consumer is both a problem and a solution for an electricity system. It requires firm generation, but it also brings the grid a load that makes generation investments worthwhile.

What the region actually gets

This is where hopes and reality diverge most.

Jobs. The construction phase employs considerably more people than the operating phase, and the difference is large rather than small. A data centre is a highly automated building, not a factory. This is not a criticism but a fact worth knowing before a project's employment promises are added up. The precise figures are worth asking the project for, separately for each phase.

Property tax and land. These are real and continuing income for the municipality.

Waste heat. A data centre's heat can be fed into a district heating network, and in Finland this is unusually often possible because the district heating network exists. The conditions are that the network is close enough and that a buyer for the heat can be found. If not, the heat goes into the air.

The price of electricity. A large new consumer raises demand. The effect depends on whether generation is built at the same pace. This is the most contested part of the subject, and reasoned estimates are offered in both directions.

What to watch

If you want to know where this is heading, three measures say more than the press releases:

1. The combined capacity under connection agreements in Fingrid's interim reports. It is the only public figure that measures commitments rather than intentions.
2. The pace of building permits and commissionings. An agreement is not concrete.
3. Transmission grid investment decisions. They say where capacity will actually appear, and they are made years before anything is visible.

In one sentence

Data centres are coming to Finland, but the pace of building is set not by the appetite to invest but by how fast you can get connected to the transmission grid, which is why the gigawatts of enquiries and the gigawatts of agreements are worth keeping apart.