Fortum to support a global AI infrastructure company Nscale’s plans to establish a data centre in Harjavalta, Finland

Nscale has expressed interest in building a data centre in Harjavalta, Finland (Sievari industrial area), with Fortum acting as site development partner. The project aligns with Fortum’s strategic priority to drive electrification in the Nordics; Fortum previously signed a planning reservation with the Town of Harjavalta in June 2024 to establish the site. This is a strategic infrastructure/AI-infrastructure development that is positive for Fortum’s service pipeline and local electrification demand but is unlikely to have material near-term market impact.

Analysis

Electrification demand from AI-scale compute creates concentrated, lumpy grid load that forces visible second-order winners: owners of regional transmission upgrades, transformer and high-voltage cable suppliers, and utilities that can monetize connection and capacity services rather than commodity power alone. Expect single-site projects to pull 30–200 MW of incremental load, creating multi-year lead times for reinforcement (18–36 months) and near-term optionality in development fees and capacity contracts that rarely show up in headline power price moves.

Local municipal balance sheets and industrial landowners become de facto infrastructure financiers — zoning/permitting speed and the ability to offer interruptible tariffs will determine which regions capture long-term data center rents versus those that simply export construction activity. This dynamic favors vertically integrated utilities and equipment OEMs with turnkey grid upgrade capabilities; it also raises marginal nodal power prices and capacity contract value ahead of commodity energy price changes.

Tail risks are concentrated: a stalled PPA, a change in EU state‑aid/tax treatment for discounted industrial power, or a macro slowdown in AI capex would reverse the re-rating quickly; those are 3–12 month catalysts. Watch connection agreements, announced PPAs and local capacity auction results as binary triggers — absence of these makes project equity optionality worth materially less despite headline announcements.

The consensus underestimates the speed at which engineering procurement timelines convert into visible electrical demand and tariffs. Markets are still treating AI data centers like generic real estate; in reality they behave like industrial gas plants for grid planners — high utilization, predictable draw, and therefore premium-priced capacity economics that can be captured by a narrow set of suppliers and utilities over a 1–3 year window.

Key Decisions for Investors

• Long HE:FORTUM — 6–12 month, size 2–4% NAV. Thesis: capture development/capacity upside and transmission service optionality; target +30% vs downside -15% if permitting/PPA fail. Close on negative PPA announcements or municipal opposition.
• Long NYSE:ABB — 9–18 month, size 1–3% NAV. Thesis: direct exposure to transformer/substation and HV cable demand from concentrated compute builds; look for order flow/earnings upgrades as triggers. Risk: capex pullback in AI slows orders — stop at -12%.
• Long MIL:PRY (Prysmian) — 12 month, size 1–2% NAV. Thesis: benefits from incremental land and submarine/land cable demand for new grid links; asymmetric payoff if EU electrification capex accelerates. Take profits on >25% move or if tender wins miss expectations.
• Pair trade: Long EQIX (Equinix) Jun-2027 1.5x notional calls vs Short DLR (Digital Realty) Jun-2027 0.75x notional calls — 9–12 month horizon. Rationale: favor premium interconnection platforms with stronger European footprint and higher tariff capture; reduce beta by shorting a more commoditized REIT. Close on macro credit stress or broad REIT sell-off.