The 22-Year Landmark: Google and Fortum Anchor Nuclear Capacity for AI

On September 9, 2026, Finnish state-owned utility Fortum Corporation and Google announced a 22-year corporate Power Purchase Agreement (PPA) tied directly to the Loviisa nuclear power plant. Under the terms of the agreement, commercial power deliveries will begin at a reduced capacity in 2028 before scaling up to 50% of the total generation capacity of the Loviisa facility from 2030 through 2049.

The transaction marks Google’s first nuclear PPA executed outside the United States. Crucially, it stands as the first corporate partnership globally to directly fund the operational lifetime extension and uprates of existing commercial nuclear assets, explicitly securing continuous, non-intermittent clean power for hyperscale artificial intelligence compute infrastructure.

€13 Billion Commitment and Nordic Infrastructure Buildout

Parallel to the energy contract, Google announced a committed capital investment of €13 billion (approximately $15.1 billion) spanning 2027 through 2028 into Finnish digital infrastructure. This capital program will finance the expansion of Google’s existing Hamina hub alongside the construction of three greenfield data center campuses located in Kajaani, Muhos, and Vaala.

The operational scope also incorporates an agreement to optimize a 94-megawatt (MW) battery energy storage system (BESS) co-located adjacent to the newly planned Kajaani facility, providing essential grid balancing and frequency response services.

For Fortum, the guaranteed revenue baseline provides the definitive financial underwriting required to finalize its €1 billion lifetime extension program for Loviisa’s two pressurized water reactors through 2050. Approximately €700 million of that life-extension capital program had previously remained uncommitted pending final investment decisions.

Scope, Boundaries, and the SMR Memorandum of Understanding

While the power purchase contract for the Loviisa plant is fully binding, Google and Fortum also entered into an accompanying Memorandum of Understanding (MoU) to jointly explore New Nuclear initiatives, specifically Small Modular Reactors (SMRs) and potential new builds across the Nordic region.

However, verifiable evidence confirms that this MoU entails no binding financial commitment or capital deployment to construct SMRs at this stage. The active, enforceable agreement remains strictly confined to off-taking 50% of the existing output from the Loviisa pressurized water reactors.

Practitioner Perspectives: Energy Realism vs. Grid Cannibalization

The transaction sparked sharp discussion among infrastructure engineers and energy economists. One camp framed the arrangement as strategic realism, noting that corporate AI capital is effectively rescuing critical European nuclear infrastructure that utilities struggled to finance under standard municipal or regulatory balance sheets.

Conversely, grid observers raised pointed concerns over hyperscaler energy cannibalization. Critics argued that locking up 50% of a baseload clean energy asset for proprietary AI compute could strain regional capacity and indirectly raise electricity tariffs for residential and light-industrial consumers. However, claims that this procurement will force surrounding communities onto fossil fuel alternatives remain unverified by formal grid impact assessments. Concurrently, technical practitioners debated hardware durability, questioning whether radical algorithmic and model efficiency gains might outpace the 22-year physical lifespan of these specialized data center builds.

Strategic Implications for Thailand’s Enterprise and Energy Landscape

For enterprise leaders and policy architects in Thailand, Google’s €13 billion deployment underscores that the fundamental bottleneck for generative AI scaling is no longer solely GPU availability, but access to firm, non-intermittent clean baseload electricity.

As Thailand actively positions itself to attract global hyperscalers for regional data center and cloud infrastructure hubs, reliance on intermittent renewable energy such as solar and wind will struggle to meet stringent 24/7 carbon-free enterprise mandates. Thai energy conglomerates and regulatory authorities must examine long-term PPA structures, co-located high-capacity battery energy storage systems (BESS), and firm clean baseload strategies to maintain long-term competitiveness as a viable destination for institutional AI capital.

Why it matters

Marking Google’s first nuclear PPA outside the United States, this transaction proves that hyperscale AI expansion demands dedicated baseload power and is directly reviving aging nuclear utility assets.

Primary material