The Claim and the Paper Trail

Fact-checkers reviewing 12 primary sources, including the European Commission's Grids Package materials and the International Energy Agency's Energy and AI analysis, found a consistent theme: the buildings are not the long pole. Constructing the physical facility — halls, cooling, racks — is the faster half of the equation. The binding constraint is the interconnection queue: the wait to be physically linked to transmission and distribution networks capable of feeding power-hungry AI clusters built around dense arrays of GPU hardware.

Hub Figures, Not EU Averages

The two-year and seven-year figures are maximum values and hub-centric readings, not bloc-wide averages. Across EU member states, the documented range for these timelines spans roughly two to ten years depending on the jurisdiction, the voltage level, and the state of local grid investment. A headline quoting "2 years" and "7 years" therefore captures the acute congestion at Europe's most in-demand data center hubs while smoothing over regions where the process moves materially faster — or, at the other end of the range, even slower.

Why the Queue Is the Story

For anyone planning AI compute capacity in Europe — cloud operators, colocation providers, hardware suppliers — the sequencing matters more than the averages. A two-year construction clock is meaningless if the energization date slips to year seven, because idle GPU capacity earns nothing while it waits for electrons. The EU's own grid-focused policy push and the IEA's energy-and-AI work both point to the same structural diagnosis: permitting and construction regimes are adapting faster than the physical networks they depend on.

The Bottom Line

Measured against the full two-to-ten-year spread documented across the EU, the claim that data centers take two years while power connections take seven accurately names the bottleneck at its most strained points even as it oversimplifies the regional variation — a characterization best graded Mostly True.