“Computing power is the currency of AI,” says Friedrich Merz at the Digital Summit. A day later, Bitkom quantified just how much of it Germany possesses. The study “Data Centers in Germany 2026,” prepared by the Borderstep Institute, reports investments of 23.5 billion euros for this year—57 percent more than in 2025 and a record high. About 18 billion euros are allocated to servers and storage, and 5.5 billion to buildings and technology. The total IT power capacity of all data centers will be 3,300 megawatts in 2026; in 2015, it was 1,660. By 2035, it is expected to triple to around 9,700 megawatts.

AI Accounts for Half

The driving force is artificial intelligence. AI and high-performance computing currently account for 20 percent of capacity; by 2030, this figure is expected to reach 36 percent, and by 2035, 55 percent. In absolute terms, AI capacity will grow from 670 megawatts to more than 2,100 megawatts in 2030 and reach 5,300 megawatts in 2035. Hesse remains the largest location with about 1,200 megawatts, and 2,000 megawatts are under construction there; Brandenburg currently stands at 104 megawatts and has 1,301 megawatts in the pipeline.

The expansion targets are not ambitious enough for the majority

Nevertheless, the political target is in flux. The federal government aims to double capacity by 2030, to about 6,000 megawatts; the study’s projection tops out at 5,500. Only one-third of industry respondents consider the expansion targets ambitious enough, while one-fifth expect rapid implementation. Meanwhile, the federal government’s AI task force has set a target of a 300 percent increase in AI data center capacity by 2025 and aims to reduce the time between application and commissioning to a maximum of two and a half years.

A Different Order of Magnitude

In a European comparison, Germany leads with 3.3 gigawatts, ahead of the United Kingdom (2.9), France (2.0), and Ireland (1.7). Globally, that’s not much. The U.S. will have around 47 gigawatts by 2025 and is expected to reach 160 gigawatts by 2035; China will grow from 22 to 90 gigawatts; and the study projects 37 gigawatts for all of Europe. “The AI race is also a race for computing power,” says Bitkom President Ralf Wintergerst. Those who fall behind in infrastructure will quickly lose ground in AI as well. The experts surveyed expect Germany to focus primarily on training smaller models (68 percent); 46 percent see a special role for industrial AI—that is, the use of AI in machines, systems, and industrial processes.

A Quarter for the Hyperscalers

The quarterly reports from the U.S., which were released that same week, illustrate the scale of these figures. On October 28, Alphabet reported capital expenditures of $51.4 billion for the third quarter, maintained its guidance of $195 to $205 billion for 2026, and announced a “significant” increase for 2027; free cash flow stood at minus $9.1 billion. On the same day, Microsoft reported $45.2 billion for the September quarter, including finance leases. A day later, Amazon put its cash investments at $75.9 billion for the quarter and around $220 billion for the year, with AWS growth of 39 percent. Meta reported $34.9 billion for the quarter and plans to spend $135 to $145 billion in 2026. Combined, the four companies spend a good $200 billion in a single quarter; the breakdowns differ in detail. The entire German data center market invests 23.5 billion euros over the course of the year.

Electricity, Processes, Prices

According to Bitkom, the electricity demand of German data centers will rise from around 23 billion kilowatt-hours in 2026 to 35 to 38 billion in 2030 and 56 to 70 billion in 2035—a rise that is disproportionately lower than the increase in capacity because efficiency is improving: The PUE value has dropped from 1.63 to 1.42 since 2020. Greenpeace is calling on the federal government to make the water and energy consumption of the planned facilities transparent. Wildberger points out that prices for computing power are far higher than energy prices due to scarcity. Wintergerst is calling for affordable electricity, suitable sites, and faster approval processes.

Rented Sovereignty

The reality for decision-makers is sobering. Germany is expanding its capacity, but at a pace that is insufficient to close the gap with the U.S. and China. Those who need AI capacity will rent it for the foreseeable future, mostly from providers who invest several times more per quarter than is generated annually in this country. The task force is relying on strategic partnerships and joint European procurement for frontier models. As for domestic capacity, the question remains whether regulatory processes, electricity prices, and land availability will still make the 2030 target achievable.