Nebius's Two-Speed European Strategy: Why France Gets 18 Months and Finland Gets 2027

Nebius Group is simultaneously building two of Europe's largest AI data center campuses: €8 billion in Béthune, France (former Bridgestone factory, first power end 2026) and €10 billion in Lappeenranta, Finland (greenfield, first customer capacity targeted for 2027). Same developer, same capital conviction, same GPU-dense AI workload design.

Two radically different grid timelines. And the gap between them is entirely explained by one variable: whether existing HV infrastructure was on site when the project started.

THE BROWNFIELD ADVANTAGE — IN NUMBERS

Béthune is the clearest public proof of what the brownfield advantage looks like at scale. The former Bridgestone factory retained its HTB substation connection after closure. Nebius acquired the site, reactivated the grid connection, and is delivering 240 MW of AI compute capacity in 18 months — a timeline that no greenfield site in Western Europe, and certainly no site in the US, can match.

Lappeenranta is a different strategic bet. Finland offers genuine advantages: low-carbon electricity, cold climate enabling free air cooling, and a reliable Nordic grid. But it is a greenfield site requiring a new substation and targeted local grid reinforcements. The project's own risk disclosure acknowledges this explicitly. First customer capacity in 2027 means a minimum 12-month delta versus Béthune — and full operations stretch to 2028–2030.

Nebius is not choosing one strategy over the other. It is using both deliberately: brownfield France for speed, greenfield Finland for scale and long-term positioning. That is rational capital allocation — and it implicitly confirms that brownfield sites are a finite resource worth competing for now, before they are gone.

WHAT FINLAND IS NOT

Finland is a credible market for developers with a 2027–2030 deployment horizon, high ESG requirements, and tolerance for geopolitical complexity. The Lappeenranta site sits 30 km from the Russian border. Nebius's Yandex roots have already attracted scrutiny from the Finnish Security Intelligence Service (Supo) and the European Commission regarding ownership structure and data flows — a risk that a US hyperscaler or infrastructure fund would not carry in the same form, but that illustrates the regulatory overlay that comes with Finnish edge deployments.

For developers who need AI compute capacity in the 2026–2028 window, Finland is not the answer. France is.

THE WINDOW NEBIUS SAW IN 2025

When Nebius signed the Béthune acquisition in May 2025, the market was already aware of the European grid constraint. What Nebius understood — and acted on — was that brownfield sites with existing HTB connections represent a narrow, depleting stock of fast-path deployments. You cannot manufacture a former Bridgestone factory. You can only find one before someone else does.

RTE has identified 5 fast-track sites with approximately 4,800 MW of near-ready capacity across France. GridReadiness tracks 40+ additional brownfield sites with existing or recoverable HTB connections beyond the 63 government-listed locations. The Nebius Béthune timeline is reproducible — but only on sites where the grid infrastructure already exists.

The Lappeenranta announcement confirms the thesis: when brownfield is available, developers use it. When it isn't, they build greenfield and accept the 2027–2030 timeline. For developers evaluating Europe in 2026, the question is which category your target site falls into — and whether the brownfield stock has already been allocated by the time you engage.

→ Related: Europe brownfield sites · France site selection · Grid data center hub

Sources: Nebius newsroom March 2026 · Fingrid Grid Development Plan 2026–2035 · Choose France 2026 · GridReadiness field intelligence. Updated June 2026.