Artificial Intelligence Energy Demands Shift European Data Center Landscape
November 21, 2026 · 3 min read
Also published in Português (Brasil)
The growth of artificial intelligence is becoming increasingly dependent on access to energy and power grids, turning data centers into strategic infrastructure for the digital economy.
According to a report by EY titled Artificial Intelligence and Energy: Mutual Dependence, electricity consumption by data centers could nearly double by 2030, approaching 1,000 TWh annually. This trend is pushing investors to analyze new markets, including Poland, as traditional European locations face limitations. Potential development requires grid expansion, an increase in low emission energy sources, and the mitigation of environmental and community impacts.
Global installed capacity for data centers rose from 60 GW in 2020 to 114 GW in 2025. An IEA base scenario suggests this could reach 226 GW by the end of the decade. Global investments in data centers may reach 7 trillion USD between 2025 and 2030. Consequently, the pace of AI development will depend on grid connection timelines, energy availability, and the ability of power systems to handle large, continuous loads, alongside the availability of chips and models.
By the end of 2025, nearly 12,000 data centers operated worldwide, with 45 percent in the United States and 27 percent in Europe. A typical AI data center consumes as much energy as approximately 100,000 households, while the largest centers currently under construction may use 20 times more. Power density is also rising: traditional server racks typically draw 5 to 15 kW, whereas AI racks require 40 to 60 kW or more, with some modern facilities exceeding 100 kW per rack.
Data centers currently account for about 1.5 percent of global electricity consumption. An EY ERTA model suggests this could rise to 2.5 percent by 2030 in a current trajectory scenario, while the IEA base scenario indicates approximately 3 percent. The final scale remains uncertain, depending on factors such as cooling technology, geopolitical tensions affecting supply chains, and hardware efficiency.
Grid connection times are a major constraint. While a hyperscale campus can be designed and built in about four years, large transmission infrastructure can take a decade. Network bottlenecks may delay about 20 percent of planned global data center capacity by 2030, and AI driven growth may require 720 billion USD in grid investments by the end of the decade.
Pressure is high in Frankfurt, London, Amsterdam, Paris, and Dublin, which accounted for over 60 percent of European data center capacity in early 2025. Expansion there is hindered by grid congestion, land shortages, and wait times for connections reaching seven years in Germany and ten years in the United Kingdom and the Netherlands.
In the Central European region, which includes Nordic countries, Central and Eastern Europe, and Turkey, there were about 25 percent of all European data centers by number at the end of 2025. In this region, about 65 percent of electricity came from low emission sources. Poland offers available land and development potential, though it has lower low emission energy shares and relatively higher prices. However, Warsaw's average temperature of 10 degrees Celsius may reduce cooling needs compared to Frankfurt at 11 degrees Celsius and Stockholm at 9 degrees Celsius.
Polish data center capacity could grow to about 500 MW in 2030 and 1.2 GW by 2034. The national high voltage grid investment plan predicts the ability to support this capacity by 2034, provided there is timely grid expansion and the adoption of solutions like heat recovery and water reuse.