Google’s long term power deal with Fortum in Finland is not a press release about “clean energy leadership”, it is a binding, 22 year bet that existing nuclear is the anchor fuel for AI growth in a deregulated European market. The most sophisticated electricity buyer on earth just told us what it believes about reliability, price stability and decarbonization, and it did not pick hydrogen, CCS, or a CFD on offshore wind. It picked half of a real reactor’s output for two decades, plus an option to help build more nuclear behind it.

Alphabet is buying up to 50 percent of the Loviisa plant’s output from 2030 to 2049, with a ramp that starts smaller in 2028 and locks in a nuclear baseload tranche right next to its Hamina data center cluster. Fortum gets a premium, creditworthy offtaker that de‑risks life extension and future new build, Google gets a hedge against the volatility that is now baked into European power markets. This is not a US style greenwashing deal where the energy is “carbon matched” on a spreadsheet. It is a physical tie‑in to a specific plant, in a country that has already done the hard work of licensing and building reliable reactors and integrating them with high shares of renewables.

The immediate significance for the nuclear revival is simple and important. First, big tech is now clearly willing to sign long duration contracts on nuclear, not just wind and solar bundled with RECs. Second, the geography matters. Google did this in Finland, not in the US, because Finland has functioning permitting, a stable nuclear fleet, and a political consensus around nuclear as a climate tool, and because the US grid and regulatory mess cannot yet guarantee the same combination of reliability, price and schedule. Third, the deal is structured around an existing plant but layered with a joint development agreement for new nuclear and renewables. That tells you where the capital stack is going next: long term AI demand backed by old reactors first, then expansion capital for SMRs or Gen III+ as the regulatory and technology risk profile improves.

For investors and utilities, the message is more uncomfortable. If you are still pitching new nuclear without an identified high value anchor load, you are behind the curve. Google just demonstrated what the new nuclear PPA template looks like, and it is not a generic “100 percent clean energy by 2035” pledge. It is a specific, asset linked, multi decade contract that shores up the economics of an existing plant and frames an option chain on future units. That is good news for operators with fleets in countries like Finland or Sweden, and for serious SMR developers who can actually offer a build schedule and a regulator with teeth. It is bad news for paper reactors whose business model depends on sovereign subsidies and vague future demand rather than hard AI loads that need power by a date certain.

**THE ECONOMICS** The economics of this deal start from a simple truth the AI crowd is finally internalizing, nuclear delivers extremely high capacity factor at a known cost over long horizons, and nothing else does that today at scale. Loviisa’s historical capacity factor has been north of 90 percent, which means Google is effectively reserving a tightly predictable block of MWhs that matches the always‑on profile of AI training clusters and latency sensitive inference workloads. Translate that into money, and you are looking at a hedge against both commodity price volatility and policy risk. Spot prices in the Nordic market can be wild, especially with hydro variability and interconnection constraints, but a long term nuclear PPA smooths that out into something close to a fixed fuel charge.

On the Fortum side, the contract is a revenue stabilizer that supports both continued investment in life extension and, crucially, the cost of capital for any new nuclear the company pursues. Banks and institutional investors care less about levelized cost in an abstract sense and more about contracted cash flows. A 22 year deal with a hyperscale buyer is better than any capacity market design in Europe as a de‑risking tool. That will show up in lower financing spreads compared to merchant nuclear or renewables exposed to wholesale prices. For Google, locking in nuclear reduces its exposure to the rising cost of balancing renewables with gas and batteries as carbon prices tighten. The company will still buy wind and solar, but the baseload slice lets it minimize the very expensive last 10 to 20 percent decarbonization burden.

This is also a direct signal to SMR vendors about where they must price to be taken seriously. If an existing large reactor can deliver power into a long term tech PPA at, say, 50 to 70 euros per MWh all‑in, including life extension capex, then any SMR that shows up north of 100 euros while still carrying significant schedule and licensing risk will not win this kind of customer. Real buyers will compare SMR proposals not just against gas and renewables, but against refurbishing and contracting existing nuclear. The floor for “competitive” SMR power in high value data center applications is being set in real time by deals like this, not by DOE slide decks.

**WHAT THIS ACCELERATES** The immediate acceleration is on the utility side. Fortum now has a case study and a template it can show regulators and shareholders when arguing for license extensions, uprates or even new nuclear in Finland and potentially in neighboring markets. The combination of high value industrial load and long duration contract is exactly what makes incremental nuclear politically and financially palatable. Expect other Nordic and central European utilities with solid operating fleets to start quietly pitching similar structures to cloud providers, chip fabs and hydrogen producers, with nuclear as the anchor and renewables playing around the edges.

For SMR and advanced reactor developers, this is a forcing function. The companies that are actually building hardware and have credible paths through licensing, like X‑energy, GE Hitachi with BWRX‑300 and the better funded European designs, now have leverage in conversations with tech companies. They can point to the Finland deal and say, here is the baseline for nuclear PPAs, here is what a real customer will sign for, now let’s talk about a project that comes online in the mid 2030s. Conversely, the PowerPoint reactors that cannot show a site, a regulator or a realistic construction sequence will find that the bar for “strategic MoUs” with tech firms has just moved. If a cloud provider can sign for real electrons from a real reactor, there is no reason to waste time on speculative MoUs that deliver only press coverage.

Countries also bifurcate. Those with functioning nuclear regulators, experienced operators and supportive public opinion, think Finland, Sweden, Canada, the UK if it gets its act together, move to the front of the line for data center and AI investment that demands firm, low carbon power. Jurisdictions that dragged nuclear through endless political fights or let their fleets decay without replacement learn the cost of that decision in lost industrial investment. The policy takeaway is blunt, if you want AI factories, you need either very cheap firm fossil with a free pass on CO2 or you need nuclear. The companies that own nuclear assets and can execute nuclear projects will increasingly dictate the terms.

**WHAT THIS REVEALS ABOUT NUCLEAR’S TRAJECTORY** What to watch next is who copies this pattern and where the first new build deal explicitly tied to data centers or AI training shows up. Google chose an existing plant because it needed certainty, but the joint development language around new nuclear and renewables tells you this is a forward contract on technology as much as on electrons. The next step is a cloud provider or chip maker signing a PPA that underwrites an SMR or Gen III+ unit from day one, probably in a country that has already proven it can build on budget and on schedule. If that happens, the financing model for new nuclear shifts from “government as buyer of last resort” to “industrial demand as anchor,” and the entire conversation about nuclear risk and cost changes.

The deeper lesson from Issue 18 is that nuclear’s comeback is not being driven by ideology or nostalgia, it is being driven by brutally practical buyers who cannot afford outages or price spikes. When AI and data center economics collide with grid reality, the technologies that survive are those that deliver electrons, not headlines. This Finland deal tells us nuclear has crossed that threshold for at least one of the world’s top energy buyers. The winners will be operators and vendors who can show up with steel, concrete and licenses instead of animations. The losers will be regions and companies that assumed nuclear could be indefinitely postponed while still attracting power hungry industry. Nuclear is returning to the center of the energy system, one hard contract at a time.