By Kelly KIRSCH -Directeur Général ESG Europe
AI data centres are no longer “just” real estate projects. They are strategic infrastructure—as important to national competitiveness as ports, rail, and power plants. The ideal location is therefore a balancing act between reliable low-carbon electricity, grid capacity, climate and cooling needs, network connectivity, permitting speed, and social license to operate (water, land use, and community impact).
Two regions stand out in Europe’s current buildout cycle: France and the Nordics. Each offers a distinct advantage profile—and each comes with constraints investors and policymakers need to anticipate early.
🇫🇷 France: Low-Carbon Baseload + Connectivity + A Large Grid Pipeline
Why France is emerging as a European AI hub
France’s strongest structural advantage is the combination of stable, decarbonised electricity and a central geographic position in Europe’s digital and physical networks. Aurora Energy Research forecasts major infrastructure growth driven by digitalisation and AI workloads, with French electricity demand projected to surge 74% by 2050. France already hosts 250+ data centres, and momentum is accelerating.
A key signal is the scale of planned capacity: an estimated 11 GW of data centre projects are awaiting grid connection—a pipeline that reflects the surge in AI/cloud demand and investor appetite for French sites.
As Aurora’s Jonathan Hoare puts it, France offers “stable decarbonised power from the nuclear fleet” and a “central position interconnected to the global information network through undersea cables”—two factors that matter enormously for latency-sensitive, always-on AI workloads.
Demand characteristics: stable daily load, seasonal cooling spikes
Data centres are unusually “grid-friendly” in one sense: they have highly stable electricity demand, varying by only ~5% intraday. For utilities, this can improve predictability versus more volatile industrial loads.
But cooling changes the picture across seasons. Aurora notes seasonal variations up to 23%, driven by cooling requirements that can represent up to 40% of total energy consumption in these facilities. For planners, this matters because the system stress often appears not at average load, but at peak cooling periods.
A digitally strategic geography
France’s attractiveness is reinforced by its 20 submarine cable connections, strengthening links to global digital traffic routes—an often underappreciated factor in AI site selection. AI workloads depend not only on electricity, but also on network redundancy and international connectivity.
Efficiency improvements are real—but not sufficient alone
France’s data centre sector has improved energy efficiency over time. Aurora highlights that Power Usage Effectiveness (PUE) has fallen from >2.5 (2007) to around ~1.5 today, with modern builds capable of ~1.1. PUE measures total facility energy vs. IT energy—lower is better. Cooling remains the biggest lever for further gains.
Investors are already moving
Major players have been actively expanding or exploring French capacity—names cited include Digital Realty, Mistral, Data4, Microsoft, and “soon to be Google,” reflecting the broader European push for AI capacity close to demand, regulation, and skills.
Bottom line: France is positioned as a hub because it can offer scale, low-carbon baseload, and network connectivity—but the bottleneck will be grid connection timing and managing cooling-related peaks responsibly.
❄️ The Nordics: Clean Power + Cold Climate + Heat Reuse at Industrial Scale
Why the Nordics are becoming AI infrastructure magnets
Sweden, Norway, Denmark, and Finland are increasingly selected for AI infrastructure due to four core advantages:
The region’s “natural cooling” advantage is not just about cost; it reduces operational energy overhead and can enable better thermal management for dense GPU workloads.
Five notable Nordic AI infrastructure projects
1) Microsoft — $3.2B expansion in Sweden
Microsoft is investing ~$3.2 billion to expand cloud and AI infrastructure in Sweden, upgrading data centres with AI-optimised chips and adding ~1,000 MW of new clean energy capacity to support the buildout. The company also plans to train 250,000 people in Sweden on AI tools. Swedish operations are expected to run on 100% renewable energy, mainly wind and hydro.
2) Polar — AI-focused data centre in Tørdal, Norway (DRA01)
Polar is developing an AI-oriented facility in southern Norway, offering 12 MW in phase one, targeted to go live by end of 2025. It will run on 100% renewable energy, primarily hydropower, and use advanced liquid cooling for high-heat AI chips. It is planned to Tier III standards for high reliability.
3) atNorth — DEN01 campus near Copenhagen, Denmark
atNorth’s Copenhagen-area campus aims for up to 30 MW in its initial phase (with additional expansion planned). A standout feature is heat reuse: surplus heat will be routed to nearby buildings—turning data centre waste heat into local energy value.
4) Apple — Foulum data centre, Denmark
Apple operates a large facility in Foulum, supporting European services like iCloud, Siri, Maps, and the App Store. It is powered by clean energy including solar and wind, supported by Apple’s own renewable investments. Proximity to a major substation supports reliability—critical for AI-grade uptime expectations.
5) LUMI Supercomputer — Kajaani, Finland
LUMI is one of the world’s most powerful supercomputers, backed by the EU and ten European countries. It supports AI research and scientific computing and runs entirely on hydroelectric power, with waste heat reuse to warm nearby buildings—an example of system-level efficiency beyond PUE.
Bottom line: The Nordics win on clean power, climate efficiency, and heat reuse, making them ideal for long-duration AI compute—especially where operators want low operational emissions and scalable sites.
So what makes an “ideal” AI data-centre location, in practice?
Across France and the Nordics, the winning formula tends to include:
🔍 ESG.AI Insight
The next wave of AI infrastructure winners will be defined less by “who builds the most racks” and more by who secures the cleanest, most resilient operating envelope.
From an ESG lens, data centres concentrate three systemic risks:
In short: the “best” locations are those where AI can scale without externalising costs onto grids, water systems, and communities.
📌 What to Do Now
For data-centre developers
For policymakers and regulators
For investors