Growth is not market share
The IEA’s 17% figure describes a change from the previous year. It does not mean data centers consumed 17% of the world’s electricity. The agency also examines efficiency and constraints on expansion.
A national model is not a local meter
Berkeley Lab’s national estimate and EIA’s higher-demand scenario serve different purposes. Neither measures a proposed facility’s actual load or the change on your electricity bill. For that, inspect the local rate case and cost allocation.
Count the benefit on its own clock
Virginia JLARC found economic benefits concentrated in initial construction and potential substantial local tax revenues. Ongoing staffing, construction crews, incentives and public costs must be counted separately. Another state’s record is context, not your project’s guarantee.
How this could be read
A rapid-growth figure can suggest a giant footprint; a small national share can suggest no local problem. Neither reading is enough. The utility and neighbors experience the load where it is connected.
What would change the answer
Metered local demand, an approved tariff, documented tax collections and a credible build-versus-alternative comparison could change a project assessment. A developer’s investment total alone would not.
Earlier edition wording & complete retained context
What holds up
The IEA reports that data-centre electricity demand grew 17% in 2025, compared with 3% for electricity demand worldwide. Those are growth rates, not the share of all electricity consumed.
What else contributes
The IEA also examines hardware efficiency, grid constraints, and options for meeting demand. Faster demand growth is not the same question as where new power comes from or who pays for the infrastructure.
The U.S. scale check
Berkeley Lab’s 2025 Update, published in June 2026, estimates U.S. data centers used 192 TWh in 2024—about 4.7% of national electricity use. It is a model-based estimate of data centers generally, not AI alone. EIA’s separate higher-demand scenario illustrates possible local supply and wholesale-price pressures, not a measured effect on your household bill.
What it really means
There is a substantive planning issue. A global percentage alone cannot tell you what will happen to your electricity bill. That requires local utility, grid, and cost-allocation evidence.
How this is framed
A large growth percentage can make a sector sound larger than it is. Conversely, a small global share can obscure local pressure. Neither framing settles the local question.
What a rejection can cost
A rejected proposal can mean foregone local revenue, construction work and operating jobs—but those are project-specific opportunities, not guaranteed losses. Virginia’s 2024 JLARC review found substantial local tax benefits and that most economic benefits arose during construction. Temporary work and permanent employment must be counted separately.
The net community ledger
For a particular proposal, compare verified local tax receipts with incentives and public costs, permanent jobs with construction job-years, and infrastructure benefits with who pays for them. Also examine land, water, noise and alternative uses. A relocated project may be a local loss without being a national loss. A developer’s investment headline is not household income or net public benefit.
The paper trail
IEA: Data-centre electricity use in 2025 (opens a new tab)LBNL / DOE: U.S. Data Center Energy Usage Report, 2025 Update (opens a new tab)EIA: higher-demand scenario and regional implications (opens a new tab)Virginia JLARC, 2024: data-center benefits, construction and local trade-offs (opens a new tab)Three source organizations, different jobs: IEA for global growth, Berkeley Lab for a U.S. model-based estimate, EIA for a scenario. They are not three independent confirmations of the same number. Virginia JLARC provides regional economic context, not a forecast for a particular proposed site. Opportunity costs need an explicit build-versus-alternative comparison. This rewrite does not confer new editorial approval.