Google's '$15 Billion Nuclear Bet' Is Really a Power Contract

The headline reads like a purchase: Alphabet betting $15 billion on nuclear reactors to keep its artificial intelligence alive. That framing is the false narrative worth testing first, because the money and the reactors are two very different things, and the difference tells you what this announcement actually does for shareholders.

Start with the number, because it is the loudest and the least accurate part of the story. The $15 billion is not a nuclear budget. It is Google's total AI infrastructure commitment in Finland, its largest single investment in Europe, spread across 2027 and 2028. The euros go to building three new data centers in Kajaani, Muhos and Vaala plus an expansion at Hamina, with clean-energy and grid projects around the edges. Nuclear is a slice of a much larger construction bill.

What the "bet" really is

The nuclear piece is not Google buying a plant at all. Google signed a 22-year power purchase agreement with Fortum, the Finnish utility, for up to 50% of the output of the Loviisa nuclear station — a two-reactor site that supplies roughly a tenth of Finland's electricity. The deal starts at a smaller volume in 2028 and rises to half the plant's output for the 2030–2049 period.

That structure matters, because a power purchase agreement is a contract to buy electricity, not an ownership stake and not a reactor build. It also does not stand on its own. Loviisa is old — its two reactors date to the late 1970s and early 1980s — and without a roughly €1 billion lifetime-extension program it could not run past 2030. About €700 million of that work still lacks a final go-ahead. Google's contract gives Fortum the revenue certainty to spend it, in exchange for firm, around-the-clock power delivered near Google's nearby Hamina data campus.

So the honest way to read the deal is underwriter, not owner. Google is paying to keep an aging plant alive and to slot its output into its own grid math. What it locks in is predictability — a known, clean, 24-hour power source for the compute that never sleeps — not new capacity on a scale that changes Alphabet's AI economics. Half of a roughly one-gigawatt site is small change against the gigawatts of computing Alphabet is buying this year.

Power is the constraint; this is the hedge

None of this is a surprise to anyone watching energy and tech cross paths. The binding constraint on AI is no longer chips or talent in the near term; it is firm, uninterruptible electricity. Wind and solar need backups, batteries add cost and clock limits, and natural gas still carries carbon. Nuclear is the one source that runs flat all day, every day, which is why every hyperscaler is signing the same kind of deal. This is Google's first nuclear agreement outside the United States, where it has already contracted with Kairos Power and others. The repeat pattern — AI companies sponsoring someone else's plant to secure baseload power — is the structural story, and this Finland deal is another tile in it.

For a retail shareholder, the useful question is not whether Google wants the power, but what the deal costs in cash returned to owners. That is where the numbers get less comfortable. Alphabet has raised its global capital spending plan this year to as much as $205 billion. Its trailing free cash flow is roughly $53 billion, down about a fifth year over year, while capital expenditures — the biggest driver of that squeeze — run to about $132 billion. Cash on the balance sheet is still ample and net debt is negative, so there is no solvency problem. But a company pouring hundreds of billions into data centers and now underwriting nuclear lifetime extensions to feed them is at the opposite end of the spectrum from the mature, cash-returning machines this column usually prefers. The dividend remains a rounding error at a 0.25% yield.

Whose stock moved

The reaction is the cleanest signal of who this changes. Alphabet shares barely budged on the day, up about half a percent. Fortum jumped about 10%. That asymmetry tells you investors read the deal exactly as the numbers suggest: it is a measurable financing win for the utility, which now has a 22-year revenue lock to justify its extension capex, and a rounding error inside a $200 billion-plus spending plan for Alphabet. When a "big" energy headline moves the counterparty and not the company making it, the market is pricing the reality, not the novelty.

That is the useful correction to the headline. The $15 billion nuclear bet is really a $15 billion data-center build paired with a long-term power contract that props up an existing reactor — small, contingent, and strategically about certainty rather than capacity. The story worth watching is the other one, on Alphabet's own books: a capital plan so large that free cash flow is shrinking even as revenue grows at a 20% clip. If that trajectory flips negative or buybacks thin out to fund more concrete and uranium, the risk is not that the Finland reactor disappoints — it is that the machine funding it keeps spending faster than it generates. Nothing in this announcement changes that math; it just makes the direction of travel explicit.