Google’s decision to anchor half of Finland’s Loviisa nuclear plant output with a 20‑year contract is real progress, not noise, because it turns existing nuclear capacity into core infrastructure for hyperscale computing rather than a feel‑good add‑on to ESG reports. This is the first time a major tech company has effectively underwritten a reactor life‑extension at utility scale, and it will not be the last. Fortum gets a bankable revenue stream to justify keeping Loviisa online into the 2040s, Google gets firm zero‑carbon baseload for its AI buildout, and everyone else now has a concrete template for how nuclear and data centers actually transact.

What happened in Finland is simple but strategically important. Fortum plans to extend the operating life of the two‑unit Loviisa plant, and Google has signed a long‑term power purchase agreement that will cover up to 50 percent of the plant’s output for two decades starting in 2030. The contract steps in at lower volumes in 2028, then ramps to half the station’s capacity through 2049, directly supporting the economics of the life‑extension work and any associated uprates or digital modernization. In parallel, Google is lining up roughly 13 billion euros of data center investment in Finland over 2027‑2028, in a grid that is already highly decarbonized and friendly to long‑term industrial loads. This is not a speculative SMR MoU or a “joint study” announcement, it is a committed baseload procurement from a running plant, tied to concrete capital expenditure in servers, networking, and buildings.

The context matters. Over the last two years, Microsoft, Amazon, Google, and Meta have collectively signed or explored nuclear PPAs, SMR development agreements, and reactor restart deals to feed AI‑driven load growth that is now measured in tens of gigawatts rather than hundreds of megawatts. Some of those deals are vapor, designed to signal seriousness about climate while remaining nonbinding on actual nuclear delivery. But buried in the noise, a pattern is emerging: the most serious tech buyers are now willing to shoulder long‑duration, single‑asset risk on specific reactors or SMR projects, provided the regulatory environment is predictable and the local transmission operator plays ball. The Loviisa contract is a proof point that existing reactors can win those deals today, without waiting for SMR design certifications or first‑of‑a‑kind construction risk to be resolved. That is quietly transformative for nuclear owner balance sheets, and it raises the competitive bar for every generator still pitching generic “100 percent renewable” portfolios to data centers.

**THE ECONOMICS** Viewed through a finance lens, Google has effectively become a co‑sponsor of Loviisa’s continued operation, but without taking equity or operational risk. The plant currently produces on the order of 1 gigawatt of net electrical output, and Google is committing to buy up to 50 percent of that across two decades. That locks in a long‑term revenue floor for Fortum, the owner and operator, which can be capitalized into lower‑cost debt for life‑extension capex, potentially hundreds of millions of euros per unit for pressure vessel inspections, safety system upgrades, and turbine‑generator modernization. For Google, the trade‑off is a predictably priced block of firm power that can be matched to compute demand with relatively minor grid balancing, instead of relying on volatile spot markets and bundled “renewable plus certificate” packages that have weak physical correlation to actual data center load.

From a capacity factor perspective, nuclear already sits near the top of the stack. Well managed reactors routinely run above 90 percent annual capacity factor, and baseload PPAs from hyperscalers reduce the incentive to cycle or derate units for market reasons. That translates into lower effective levelized cost of electricity because fixed O&M and capital recovery are spread over more megawatt‑hours. In markets like Finland, where carbon prices and gas volatility both matter, a fully amortized but well maintained reactor selling a large fraction of its output under a long‑term corporate PPA can be cost‑competitive with new combined‑cycle gas at current input prices, especially once grid‑level system costs of balancing renewables are accounted for. For data center buyers, the alternative is building combinations of offshore wind, onshore wind, and solar with firming via batteries or imported hydro, which pushes all‑in costs higher and introduces multi‑asset performance risk. A single nuclear station with a known outage schedule and proven operating history is simply easier to underwrite.

The other economic lever is risk transfer. When a utility signs a long‑dated corporate PPA for a nuclear plant, it hedges wholesale price risk, but it also implicitly shares political and reputational risk with the buyer. If a government contemplates premature closure for nontechnical reasons, it now has to weigh the impact on a named strategic investor with public commitments. In Finland’s case, Google plans double‑digit billions in data center investment tied to the availability of reliable low‑carbon power. That creates a constituency for Loviisa’s continued operation that is very different from traditional “save our jobs” lobbying. The more reactors that lock in similar contracts with hyperscalers, the more nuclear becomes embedded in national digital infrastructure policy, which carries stronger economic arguments than legacy pro‑ or anti‑nuclear narratives.

**WHAT THIS ACCELERATES** The immediate acceleration is in reactor life‑extension programs and nuclear‑to‑data‑center contracting capacity, not SMR concrete. Loviisa’s deal will be studied by every nuclear operator with an asset approaching its initial design lifetime and every hyperscaler facing five‑ to ten‑year load growth plans. Expect copy‑paste conversations around reactors in Sweden, the broader Nordics, and stable regulatory environments in France, the UK, and selected US states with merchant nuclear fleets. Operators of plants like Duane Arnold and other restart candidates already have active dialogues with Microsoft and Google around full‑output PPAs that track this model: restart or extend, secure a single large corporate offtaker, then backfill remaining output with traditional utility customers.

For SMR developers, the signal is mixed but important. Companies like Kairos Power, TerraPower, Oklo, and their peers have spent years pitching nuclear as bespoke baseload for data centers. Some of those pitch decks promised 24/7 low‑carbon energy with load following attached, before a single NRC design certification was in hand. The Loviisa structure tells hyperscalers that they do not need to wait for first‑of‑a‑kind SMRs to get nuclear power at hyperscale. They can sign real contracts on operating reactors now, then layer SMRs in later where land availability, grid integration, and local industrial heat demand make sense. That bifurcates the developer landscape. Teams that have genuine prospects of grid‑connected, utility‑scale SMRs with credible schedules and supply chains benefit from nuclear‑to‑data‑center normalization. The pretenders, whose “reactor” is essentially a fund‑raising vehicle with no construction timetable, will find it harder to sell the need for their projects when customers can point to functioning reactors like Loviisa already meeting the load.

On the country level, Finland has quietly positioned itself as a test bed for nuclear‑powered digital infrastructure. Its grid mix, political stability, and permitting frameworks make it easier to combine large energy‑intensive campuses with firm low‑carbon generation. That advantage now compounds. Nordic neighbors pursuing new nuclear build or SMR programs can leverage the knowledge spillover in contract structures, risk allocation, and public communication. Countries that have allowed nuclear fleets to age without clear life‑extension plans, or that keep restarts in limbo for political reasons, are surrendering not just clean power but high‑margin digital investment. Japan’s recent governance scandals around plant restart applications are a good example of how quickly credibility can evaporate when safety culture and data integrity are questioned. Hyperscale buyers will not anchor 20‑year PPAs to reactors that sit inside regulatory soap operas.

**WHERE THIS IS HEADING** The deeper story in the Loviisa contract is that nuclear is being recast from legacy baseload to digital infrastructure, with hyperscale buyers acting as de facto anchor tenants for reactors. Once you accept that AI‑driven compute will add tens of gigawatts of relatively inflexible load to grids over the next decade, the number of physically coherent solutions is limited. You either overbuild variable renewables and storage, you burn more gas and accept higher emissions and price volatility, or you lock in a mix that includes high‑capacity‑factor nuclear. Big tech choosing the third path, with named reactors and real money, changes the risk calculus for utilities, regulators, and investors.

The next phase to watch is competitive response. Other hyperscalers will not let Google lock up the only tractable nuclear deals in hospitable jurisdictions. Expect a race to secure similar PPAs in Europe, North America, and select Asian markets. That will force more clarity on which SMR projects have actual siting, licensing, and supply chain progress versus those that only have press releases and renderings. It will also expose which utilities are prepared to treat nuclear as a strategic asset instead of a politically sensitive liability.

The single most important takeaway from this story is that nuclear’s revival will not be driven primarily by climate rhetoric, it will be driven by hard requirements of digital and industrial systems that cannot tolerate power intermittency or carbon risk at scale. Loviisa is an early example of that logic crystallizing into a contract. The industry players who internalize this, and build their project pipelines, financing structures, and regulatory strategies around nuclear as critical infrastructure for the AI economy, are the ones positioned to win. Those still arguing about nuclear in the language of twentieth‑century political battles are already behind the curve.