Europe’s Compute Gap and the Data Centre Backlash
cepInput

Artificial Intelligence

Europe’s Compute Gap and the Data Centre Backlash

Dr. Anselm Küsters, LL.M.
Dr. Anselm Küsters, LL.M.
  • Nordic countries such as Finland offer cheap, low-emission electricity, a cool climate and rapid grid connections for training large AI models. Meanwhile, Germany provides the industrial demand. Bundled, multi-year off-take agreements would make large-scale facilities financially viable.
  • Europe must also catch up in terms of pace. The cep is calling for binding deadlines for planning permission and grid connections.
  • Many objections to data centres do not stand up to empirical scrutiny. Above all, questions regarding ownership and the local distribution of tax revenue remain unresolved.

Germany and Finland could work together to play a key role in developing a sovereign European AI infrastructure. While Finland offers favourable conditions for large AI data centres, such as cheap energy and fast network connections, Germany has a strong industrial base and high demand for AI applications. A recent study by the Centre for European Policy (cep) recommends a European division of labour.

cepInput

Currently, Europe accounts for less than five per cent of global AI computing power, whereas the US accounts for around 75 per cent. ‘The gap between demand and available data centre capacity in Europe could rise from three gigawatts to 14 gigawatts by 2030,’ says Anselm Küsters, Scientific Director of cep Nordic Countries. Europe is also lagging behind in terms of construction speed: in Germany, for example, it takes an average of around 42 months to build a data centre, compared with around 24 months in the US. In some cases, Chinese operators are completing 100-megawatt facilities in less than twelve months.

Finland demonstrates how faster procedures and site planning can facilitate investment. However, even there, pressure on grid operators is mounting. ‘Building new facilities where electricity is generated rather than where it is consumed cuts connection times by years and reduces the need for new power lines,’ says Küsters. The cep recommends prioritising grid connections based on capacity and intended use. To ensure European technological sovereignty, the cep also suggests a targeted division of labour. Given their energy and climate conditions, the Nordic countries could become preferred locations for training large AI models. Meanwhile, Germany should expand its computing capacity for industrial AI applications mainly where latency is critical.

The study also examines the growing protests against data centres in both countries. Empirical research only partially supports concerns about rising household electricity prices or high water consumption. However, more relevant are issues of ownership and local distribution. Host communities must derive tangible benefits, for example through the business tax redistribution scheme planned by the German government. Agreements between operators and local authorities should also be publicly accessible. ‘Europe cannot afford the emotional intensity of the US debate,’ concludes Küsters. ‘A division of labour between Nordic computing capacity and German industrial demand, embedded within common EU standards, would be a pragmatic path towards greater technological sovereignty.’

Download PDF

Europe’s Compute Gap and the Data Centre Backlash (publ. 10.27.2026) PDF 777 KB Download
Europe’s Compute Gap and the Data Centre Backlash