The infrastructure behind the AI economy

Begin with three different questions

How much electricity might AI use? Where can new demand be supplied? Who will finance the assets that connect one to the other? These questions belong together, but they require different records. A global forecast can describe scale without establishing a connection date. An investment total can describe expenditure without identifying a finished system.

The IEA’s Energy and AI is a useful starting point because it examines both electricity for AI and AI applications within the energy sector. Read it alongside the electricity outlooks, which place data centres within a wider system of industrial demand, electrification, weather and generation.

Keep the forecast vintage visible

Electricity 2024 looked ahead to 2026. Electricity 2025 moved the forecast window to 2027. Their difference is a research opportunity. Before interpreting a changed number as a changed trend, ask whether the coverage, assumptions or information available to the forecaster also changed.

A retrospective source answers another question. Global Energy Review 2025 describes developments in 2024. Its publication year and observation year are different. Retaining both prevents a chart of publication dates from accidentally becoming a chart of economic activity.

Follow investment to the operating boundary

The World Energy Investment series adds the financing perspective: capital flows, sources of finance and differences across technologies and regions. It helps locate where spending is taking place. It does not, on its own, certify that an individual project is ready to supply a particular customer.

For a project-level inquiry, build a separate sequence: the announced demand, a credible power supply arrangement, the connection and commissioning milestones, and the operating record. Where those records are missing, retain the gap. Filling it with a global growth rate would make the answer appear more precise while changing the question.

What to take into the next inquiry

Use the reports below as a layered reading list: demand first, system context second, capital third. Our analytical conclusion is that the important evidence sits at their intersection. A stronger demand outlook makes delivery questions more consequential; it does not settle them. The next useful observation is a dated record that links a specific source of supply to a specific operating requirement.

Follow the power-delivery Signal →