AI drives global datacenter capacity to quadruple by 2035 as Europe’s share declines despite new policy support

Study Download

• Roland Berger study forecasts global installed datacenter capacity to rise to as much as 340 gigawatts by 2035.
• Europe currently needs up to seven years to connect a new datacenter to the power grid.
• Europe’s share of global datacenter capacity is expected to shrink from 13% to 10% by 2030.

Munich, July 2026:The global race for artificial intelligence is increasingly being decided by datacenter capacity, grid connections, and permitting processes. According to Roland Berger’s new study, The AI Datacenter Study – Part 1: The Global Build-out Race, global installed datacenter capacity is projected to nearly quadruple from 88 gigawatts in 2025 to as much as 340 gigawatts by 2035. Over the same period, AI’s share of total datacenter load is expected to increase from 20% today to almost 60% by 2035. The study is based on the Roland Berger Datacenter Model, which analyzes developments in datacenter capacity, energy demand, and investment activity. The analysis is complemented by a survey of 70 decision-makers from across the European datacenter industry.

Despite ongoing policy discussions aimed at reducing barriers for the sector, Europe risks falling further behind in the global AI race. In 2025, the United States alone already hosted more datacenters than Western Europe and Asia combined. According to Roland Berger’s analysis, Europe’s share of global installed datacenter capacity is expected to decline further, from approximately 13% today to around 10% by 2030.

Europe’s digital sovereignty requires domestic datacenter capacity

The authors view current political initiatives aimed at expanding digital infrastructure as an important step toward greater European independence, but stress that additional action is required.

“Digital sovereignty starts with power and compute. To deliver the infrastructure build-out Europe needs, we urgently require stronger power grids, faster planning and permitting procedures, and competitive energy prices,” says Dr. Edeltraud Leibrock, Global Managing Director at Roland Berger. Local datacenters enable the processing of sensitive data and the deployment of AI applications under European jurisdiction, while providing the foundation for a competitive European AI ecosystem.

The study identifies three key factors holding back Europe’s datacenter expansion:

• Grid connections: Connecting a new datacenter to the electricity grid can take up to seven years in Europe. In the United States, grid connections are generally delivered more quickly, or operators rely on various forms of on-site power generation.
• Energy costs and access: 69% of surveyed industry experts expect energy availability to become even more challenging by 2035. Electricity accounts for roughly 40% of datacenter operating costs, while European energy prices remain significantly higher than those of competing regions.
• Regulatory complexity: According to respondents, permitting procedures and regulatory requirements are among the key barriers to expanding datacenter capacity in Europe.

The scale of the challenge is illustrated by the immense infrastructure requirements of modern AI facilities. A state-of-the-art AI datacenter with a capacity of one gigawatt consumes as much electricity as a city with around one million households.

Yet domestic datacenter capacity is not only essential for Europe’s digital sovereignty but also for future economic growth.

“AI and datacenters already make a significant contribution to US economic growth. Europe participates in parts of the value chain through selected supplier industries, but is becoming increasingly dependent on external providers,”says Dr. Nikolaus Lehmann, Partner at Roland Berger. “Without sufficient compute capacity, Europe risks missing out on the growth opportunities created by this critical technology.”

However, financial incentives alone will not be enough to close Europe’s gap. What is needed instead is anticipatory grid planning and technology-neutral regulatory frameworks that enable investment rather than delay it.

About the study

The AI Datacenter Study – Part 1: The Global Build-out Race is based on the Roland Berger Datacenter Model, which analyzes global developments in datacenter capacity, energy demand, and investment activity. The study is complemented by a survey of 70 decision-makers from the European datacenter industry, including datacenter operators as well as companies from the technology, energy, infrastructure, construction, and engineering sectors. Approximately 80% of respondents are directly involved in strategic or investment-related decisions concerning datacenter infrastructure.