10CountriesAbout 3 minutes

How do countries power their electricity demand?

Choose two places and compare annual demand, generation and the sources of electricity. These national totals frame AI infrastructure; they do not establish whether a particular data center can connect to a local grid.

A side by side electricity atlas

Choose two systems.

One source, one year, aligned units. Change either place to see a new pairing.

First place

Second place

Ember Yearly Electricity Data · CSV updated 22 September 2026 · annual TWh · 2025 data may be revised

Read the boundary

A national total is not a grid connection.

Generation and demand differ because of electricity trade and accounting. A generation mix describes a national annual average, not the electricity delivered to a particular building at a particular hour.

For the United States only, Berkeley Lab modeled data center use at 192 TWh, 4.7% of national electricity use in 2024. Its Reference Case reaches 649 TWh in 2030, with a 521–843 TWh uncertainty range. These are separate model estimates and projections; the 2024 share is not recomputed from Ember’s denominator. A comparable national data center estimate is not established here for the other places.

The IEA’s 2026 global outlook projects about 950 TWh of data center electricity in 2030. This tool does not allocate that global forecast among countries.

Sources and method

Ember Yearly Electricity Data, global CSV, retrieved 27 September 2026, source file last modified 22 September 2026. Licensed CC BY 4.0. Displayed demand, generation and fuel categories are taken from the same file for each year and area. The European Union entry is Ember’s EU aggregate, not IEA’s Europe region.
Ember methodology. Demand in this dataset is generation plus net imports. The clean/fossil grouping and source categories are Ember’s definitions. A category may be revised in future releases.
Berkeley Lab, United States Data Center Energy Usage Report: 2025 Update, June 2026. U.S. 2024 model estimate and 2030 Reference Case with compounded uncertainty range.
IEA, Key Questions on Energy and AI, April 2026. Separate global 2030 projection.

Method. Three-year change equals Ember 2025 demand minus Ember 2022 demand for the same area. Mix shares use Ember clean or fossil generation divided by Ember total generation in the selected year. This is a comparison of national systems, not a ranking of AI readiness, grid reliability or available data center capacity. Selections stay in the URL fragment for sharing; no personal data are sent.