Enter your details below and we’ll send you the “Europe’s AI deficit: Building foundations to compete” report
In April 2025 the European Commission set out its ambition to become an “AI Continent”, but the gap between ambition and delivery remains substantial. Europe has less data centre capacity, more constrained access to power, shallower capital markets and fewer frontier AI companies than the US, while China is scaling rapidly across models, compute and infrastructure. This report examines the structural weaknesses holding European companies back across four domains: capital availability, existing data centre capacity, the ability to construct new capacity and the cost of operating it. It is written for data centre operators, investors and policymakers who want to understand where Europe stands and what it would take to compete credibly.
Methodology
The analysis benchmarks Europe against the US as the clearest and most quantifiable comparison, drawing on the Stanford AI Index Report 2026, Prosus, the European Commission, EUDCA, Cleanview and the IEA. Operating cost comparisons use STL Partners’ Data Centre Investment Calculator, which models net opex including power and technician labour costs across European and US markets. Power and grid constraints are assessed from published positions of transmission operators including Fingrid, Statnett and RTE.
Executive Summary
STL Partners’ analysis shows structural gaps against the US in all four domains. US private AI investment was 13.7 times the European level in 2025, rising to 51 times in generative AI, and European investors are funding fewer new companies. The US operates close to five times the data centre capacity of the EU27, and Europe’s share of global capacity fell from over 25% in 2015 to around 15% by 2024.
Building at scale is now the sharper constraint. The median US pipeline project is 144 MW against 50 MW in Europe, and the US has 44 projects of 1 GW or more against 10. Power compounds the problem: net opex in Frankfurt, London, Amsterdam, Paris and Dublin is around 50% higher than in US markets. Finland, Norway and France offer cheaper power, but each faces grid connection constraints, so the pool of locations that can host a multi-gigawatt AI campus is far smaller than the map of low electricity prices suggests.
EU policymakers cannot pursue AI leadership everywhere. The report argues for a strategic AI baseline: the ability for European companies to train and operate advanced models on European infrastructure, supported by a broad sovereign inference footprint. Data centre operators have particular agency over two of the four domains, the speed of expansion and the cost of operation, and the report sets out how they can use their investment weight and industry voice to help close the gap.
Data centre operators have particular agency over the speed of expansion and the cost of operation. To help close the gap, operators should:
- Push for accelerated planning and permitting routes that run approvals in parallel, as in Aragón’s PIGA model.
- Support tougher grid queue reform built on milestones, financial commitments and evidence of project maturity, drawing on the UK’s emerging approach.
- Campaign for differentiated treatment of strategic projects, based on defined and verifiable criteria.
- Engage communities early to maintain public legitimacy.
- Improve operating economics by lobbying for targeted tax and tariff reform, advocating a secure low-carbon energy mix, co-investing in grid infrastructure and investing in the skills pipeline.