Google just locked up half the power from a Finnish nuclear plant for 22 years. That is the deal that makes the $15 billion data center bet in Finland actually work.
The money is real. At least $15 billion, flowing into Finland through 2028. Three new data centers in the north, plus an expansion of the existing site in Hamina on the coast. Google calls this its largest single investment in Europe. I believe them. The scale matches what AI training now demands.
Freezing Nordic air is not a gimmick. It is a cost line that never sleeps. Servers throw off heat all day and all night. In Finland, the outside air does most of the cooling work for free. That cuts power use and keeps the machines stable. Amazon and Microsoft have been chasing the same math across the region. Google is just moving faster here.
The nuclear piece is the part that made me sit up. Google signed a 22-year power purchase agreement with Fortum, the Finnish state-owned utility. The deal covers up to 50 percent of the output from the Loviisa nuclear power plant. Power deliveries start smaller in 2028, then hit the full share from 2030 through 2049. This is Google's first nuclear power contract outside the United States.
Fortum said the plant could not keep running past 2030 without this revenue. Now it can push through 2050. That is a lifetime extension and upgrade paid for by a tech company that needs clean, steady electricity for AI workloads. It is a strange pairing. A search and AI giant propping up a decades-old reactor in a small Nordic country. But the logic is plain. AI training does not tolerate brownouts. Nuclear does not blink when the wind stops.
There is more in the package. Grid improvements. Battery storage. Onshore wind projects. All of it points to the same bottleneck. Power. Everyone talks about chips and models. The real constraint is electrons. Google knows this. That is why the headline number is $15 billion, but the quiet number that matters more is 22 years.
I get excited about this because it shows how serious the AI buildout has become. This is not a pilot. This is not a press release with a ribbon cutting. This is long-term capital tied to long-term power. It changes the shape of the business. You cannot unwind a 22-year nuclear deal if your AI plans stall. You are in it.
And that is the catch. Google is betting that AI demand will stay hot for two decades. That is a long time in tech. We have seen big bets go wrong before. Data centers can sit half empty if the usage does not show up. The power is locked. The buildings will get built. The question is whether the AI services people actually use will grow fast enough to fill all that capacity.
There is another catch hidden in the climate advantage. Cold air helps, but it does not solve everything. The grid still has limits. Construction still takes time. Finland is small. A $15 billion push is huge for a country of 5.5 million people. The project is expected to add about $3.6 billion a year to Finland's GDP during construction and support tens of thousands of jobs. That is good for Finland. It also means the country is now tied to Google's AI timeline. If the tech slows, the local economy feels it.
I also wonder about the message this sends to other governments. Tech giants can now keep old nuclear plants alive with long-term contracts. That is a new kind of leverage. It could speed up clean power in some places. It could also give a few companies outsized say over national energy plans. That is not a conspiracy theory. It is just what happens when one buyer accounts for half a plant's output.
None of this takes away from the genuine thrill of seeing it happen. I love watching big infrastructure get built. I love seeing a company put real money behind its AI talk. Google is doing that here. The $15 billion is not a projection. It is a commitment. The 22-year nuclear deal is not a pilot. It is a bet on the next twenty years of computing.
But wonder without the catch is advertising. The catch here is simple. Google is all in. If AI usage grows the way they hope, this looks brilliant. If it does not, they still owe for the power and the buildings. That is the price of moving first at this scale.