Google commits €13 billion to Finland for AI data centers and signs a 22-year nuclear power deal — its largest single European investment

Let's look at where the money actually goes. Combining Reuters and Google's official materials, the €13 billion program is assembled from several parts: the core is a data center cluster — Google plans three new data centers in northern Finland, with sites across Hamina, Muhos, Vaala and Kajaani. Hamina carries special meaning for Google: it is where the company put down its first Finnish roots fifteen years ago, and it has since become a computing stronghold in the Nordics. The logic of choosing the Nordics is straightforward: data centers that train and run large models are energy hogs and water hogs, generating enormous heat, while Finland offers long harsh winters, a stable grid and a high share of renewables — cold air can be used directly for cooling, sharply improving PUE and cutting operating costs. The region's cold climate and stable grid are making this 'cold land' one of the hottest data center locations of the AI era — Reuters noted in the same story that OpenAI also announced its first European AI data center in Norway. Two frontier labs independently betting their compute on the edge of the Arctic Circle is no coincidence.

The most delicate piece of this deal is the nuclear cooperation with Fortum. Fortum's official press release reveals the key details: the companies signed a 'multi-decade power purchase agreement' under which Google will buy up to 50% of Loviisa's output for 22 years. The significance for Finland's energy landscape goes far beyond a single transaction — Loviisa employs about 580 people and supplies roughly 10% of Finland's electricity, with operating licenses that can run until 2050; but Fortum warned plainly that without the certainty of long-term revenue, the plant would lack the incentive to pursue the lifetime-extension investment program and could not operate beyond 2030. In other words, Google's nuclear PPA is effectively 'life support' for a critical carbon-free plant: it gives Fortum the financial confidence to invest in life-extension upgrades and keep stabilizing Finland's grid for decades. Fortum and Google also signed a memorandum of understanding to explore new nuclear, renewable and flexibility capacity, ensuring that as AI demand grows, additional low-carbon energy is deployed while safeguarding long-term supply for Finnish households and businesses. Google's EMEA Director of Energy Aris Karcanias said in the announcement that supporting Loviisa's lifetime extension 'keeps a critical energy source on the grid,' mitigating system impacts from meaningful capacity going offline in the south of the country.

Placed in a larger context, Google's calculus is clear. Alphabet has raised its global capital spending guidance this year to between $195 billion and $205 billion, going all-in on an explosion of AI compute demand — and at that scale of data center expansion, power supply has escalated from a 'logistics problem' into the 'number one bottleneck.' Over the past two years, tech giants including Google, Microsoft, Amazon and Meta have ignited an unprecedented wave of nuclear cooperation to lock down clean electricity for AI data centers — from signing deals with small modular reactor (SMR) developers to signing long-term PPAs with existing plants; 'tech companies buying nuclear power' is becoming routine rather than headline news. What makes the Finland deal distinctive is its textbook structure: climate siting (northern cold cuts cooling costs), reviving an existing asset (extending Loviisa rather than building from scratch), long-term locking (a 22-year PPA providing certainty), and an option on future growth (an MoU exploring new nuclear and renewables). For Google, it secures a critical piece of its European compute map while showing capital markets that its AI expansion has power; for Finland, it is a rare win-win — €13 billion of foreign investment flowing into the northern economy while preserving 10% of the nation's electricity supply and hundreds of skilled jobs.

For ordinary readers, the most telling part of this story is what it reveals about AI's shifting cost structure. People used to talk about AI's bottleneck as chips first — Nvidia GPUs in short supply, compute prices through the roof; but by 2026, 'electricity' is becoming a harder constraint than 'chips': clusters training next-generation models need hundreds of megawatts, and a large nuclear plant's single-unit capacity is roughly that magnitude. Google, OpenAI and their peers now compete not only for chips and talent but for power plants around the world — which explains why tech giants are building data centers in cold, green corners like Finland and Norway, and why nuclear, geothermal and SMR technologies that once felt distant have become regulars in AI headlines. The next metric worth watching: when tech companies stretch their power purchase agreements from 'a few years' to 'twenty-two years,' it means their judgment of AI compute demand now spans multiple technology cycles — not a position explainable by short-term hype, but a long-term bet on exponential growth in compute demand over the next decade. For everyone riding the AI wave, this is a footnote worth savoring: the AI arms race has moved from silicon wafers to nuclear reactors.

As I finish, I cannot help but marvel at a detail of our era: the iconic image of the last infrastructure wave was racks stuffed with GPUs; the latest footnote to this one is a press release about extending a Finnish nuclear plant to 2050. From silicon to reactors, from compute to power — AI's expansion logic is dragging humanity's oldest energy anxieties back to the table. For Google and its peers, €13 billion buys not just data centers but the certainty of 'no power shortage for the next decade'; for Finland, the deal preserves not just a plant but a critical ballast stone in the Nordic clean-energy map. And for the rest of us watching, the line most worth remembering may be this: when tech giants sign 22-year contracts for electricity, their confidence in the future of AI carries more weight than any keynote deck.