Google Signs 22-Year Nuclear Deal, €13B AI Investment in Finland

The First Non-American Reactor to Power Google

Google and Fortum signed a 22-year power purchase agreement (PPA) on Wednesday, September 9, granting Google rights to up to half of the output from the Loviisa nuclear power plant, located on Finland's southern coast. The contract begins in 2028 with reduced capacity and covers 50% of the output between 2030 and 2049. It is Google's first nuclear contract outside the United States.

Ruth Porat, Google’s President and Chief Investment Officer, summarized the logic in one statement during the press conference: 'We call this BYOP, bring your own power.' The company has previously struck deals with Kairos Power for small modular reactors and with NextEra to reactivate Duane Arnold in Iowa. Loviisa represents a geographical shift in this strategy.

€13 Billion, Four Data Centers, and a 94 MW Battery

The nuclear PPA comes as part of a larger package. Google announced a €13 billion investment in Finland over the next two years, its largest European contribution to date and one of the biggest industrial investments in the country's history. The plan includes expanding the existing data center in Hamina and building three new plants in Kajaani, Vaala, and Muhos, all in northern Finland, where winter temperatures drop below -10°C, reducing cooling costs.

Complementing the package is a 94-megawatt battery near Kajaani, set to go live at the end of 2027, providing balancing capacity for the Finnish electrical grid. Google has also signed separate agreements for onshore wind energy and grid optimization services. For Fortum, the agreement enables approximately €1 billion in necessary investments to keep Loviisa operational until 2050, the deadline for a license extension that would expire without an anchor customer.

A Reading Across Three Geographies

The agreement is not an isolated positive development for Finland. It reorganizes at least three fronts. In the United States, where Google already focuses much of its nuclear portfolio, the European PPA serves as a replication pilot: if it works in a regulatory jurisdiction with strong unions and a carbon tax, the model can be exported. In the next 24 months, Amazon, Microsoft, and Meta will have to decide whether to purchase part of existing European reactors or invest in SMRs under construction.

In Germany, the contrast is stark. Berlin closed its last three reactors in April 2023, and the country now imports French nuclear energy during peak demand. A hyperscaler committing to a multi-year agreement with a Finnish reactor while German operators seek to repurchase gas safeguards highlights which side of the European energy debate carries more weight for data center builders today. For banks and consulting firms with Global Capability Centers in Germany, energy cost per MWh continues to rise.

In Brazil, Petrobras's decision to allocate a portion of pre-salt revenues to fund sovereign digital infrastructure faces a challenging comparison. While Loviisa secures a firm 22-year contract with a single buyer, Brazilian data center projects for generative AI still depend on assurances regarding hydropower dispatch and the sluggish reserve auction process. A multinational cloud customer choosing between Kajaani and Camaçari in 2028 will consider energy tariffs, regulatory predictability, and fiber proximity simultaneously.

What the Board Needs to Recognize

Skepticism about the AI capex bubble has its merits. Training costs continue to rise, and several analysts view part of this expenditure as transitory. However, there is a distinction that generic discourse overlooks: a profitable hyperscaler signing a two-decade nuclear PPA is not a lab burning through venture capital. It’s a marginal cost calculation per token served, with energy prices locked in. Companies needing to decide where to deploy future AI load in 2028 and 2029 will have to compare contracted kilowatt-hours for 20 years against renewable electricity contracts refreshed every two. It is this arithmetic—rather than excitement over the next model—that will define who will have margins in the next cycle.