Data centres: there are 262 in Italia, with capacity having doubled since 2020

There are 262 data centres in Italia, of which around 45 per cent are in Lombardy, with an installed capacity more than double that of 2020: 609 MW by the end of 2025, of which 414 MW will be concentrated in the Milan area, particularly in the north-west.

These are the figures set out in the AI for Energy Report 2026, compiled by the Energy&Strategy department of the School of Management at the Politecnico di Milano. According to the study’s simulations, by 2030, Italy’s capacity could reach 0.9 GW under the most modest growth scenario, and rise to 1.2 and 1.8 GW under medium and high attractiveness scenarios, characterised respectively by a continuation of current trends or supported by ad hoc policies and increased investment.

The estimates take into account the requests for high- and extra-high-voltage connections, which Terna updates monthly: 101.4 GW at the end of August, with 1.9 GW currently at the final stage of the connection process.

As for energy consumption, in a broader context, it is estimated that data centres in Europe will account for 28 per cent (89 TWh, second only to industry, 177 TWh) to the rise in electricity demand in Europe between 2023 and 2030, with demand rising by 1.4 per cent over the same period (from 3,214 to 3,533 TWh).

“The growth of AI is making the sustainability of the digital infrastructure that enables its development increasingly crucial, and the challenge for Europe is precisely to develop technological supply chains and sufficient computing capacity to support the race towards artificial intelligence without compromising climate targets,” comments Federico Frattini, Deputy Director of E&S and scientific director of the report. “Improving the energy efficiency of data centres, optimising cooling systems, reducing water consumption and sourcing energy from renewable sources are key levers for limiting its environmental impact. Alongside these is the need to develop AI models capable of reducing computational and energy requirements without compromising performance. This is a complex balance, in which technological sovereignty and sustainability must advance hand in hand: in May, of the 50 most widely used AI models in the world, only two were not developed in the United States or China.”