World AI Power

MARKET RESEARCH · 5 NOVEMBER 2026

A bigger energy story.
A sharper project brief.

Source-led context for the physical infrastructure behind AI. Separate global demand, local capacity and the decisions within your control.

THE DEMAND OUTLOOK

Read the scale.
Keep the scope.

The IEA’s April 2026 outlook puts worldwide data-center electricity use at 485 TWh in 2025 and projects 950 TWh in 2030 in its central case. These figures cover all data centers, including AI workloads.

Global data-center electricity consumption · TWh per year
2025
Reference estimate
485
2030
Central projection
950

A forecast is a scenario, not installed or reserved power.

View the underlying data table
IEA April 2026 demand outlook
YearElectricity useBasis
2025485 TWhReference estimate
2030950 TWhCentral projection

Source: IEA (2026), Key Questions on Energy and AI, executive summary ↗. Published 16 April 2026. CC BY 4.0. Chart adapted from the reported figures; checked 5 November 2026.

What this means for an individual site

Our planning interpretation: use global demand to understand the market, then build a separate evidence trail for your location. Electricity consumption over a year does not establish a site’s peak demand, a grid connection rating or an energization date.

TWh / year

Energy consumed over time. Useful for market scale and annual supply discussions.

MW

Power at a point in time or a stated capacity rating. Useful for a defined site scenario.

Procurement is one part of the power path

On 6 October 2026, Constellation announced a 20-year agreement with Google supporting planned uprates totaling 890 MW across its nuclear fleet. This is an announced future capacity initiative, not a listing of power available to a new project.

Read the Constellation announcement ↗ · Checked 5 November 2026.

For your own brief, separate the commercial supply arrangement from the physical connection, network studies, commissioning milestones and operating constraints. Record who must confirm each assumption.

RESEARCH METHOD

Clear sources.
Visible assumptions.

We use primary publications and identify the publisher, date, scope and basis of each quantitative statement. Our interpretation is separated from reported data.

This is a dated research snapshot. External links may change; review the original source and current local requirements before relying on a figure for a project.

Test a power scenario →