The US Energy Information Administration (EIA) is moving toward its first mandatory national survey of data center energy use, reflecting a growing view in Washington that modeled estimates are no longer enough.
Utilities and regulators are being asked to plan for rapidly rising AI-driven electricity demand, even though national reporting on data center power use remains incomplete. WIRED reported that the EIA told senators Elizabeth Warren and Josh Hawley it plans to develop a mandatory nationwide assessment after its pilot work is complete.
What the EIA pilot is actually asking
In its March 25 announcement, the EIA said it launched three voluntary pilot studies in Texas, Washington state, and the Northern Virginia-Washington, DC corridor. The agency identified 196 companies operating data centers in those regions and asked each to report on at least one facility.
The questionnaire covers energy sources, electricity consumption, site characteristics, server metrics, and cooling systems. Those categories highlight what the government still cannot see well enough today.
Energy source data can show how much of a facility’s demand is met directly by the grid and how much is supported by on-site generation or storage. Site characteristics can indicate where large concentrations of load are forming and which regions may face the greatest strain.
Server metrics matter because AI workloads are changing not just how much electricity data centers use, but how quickly that demand is climbing. Cooling data matters too, because the infrastructure burden from a data center extends beyond the servers themselves.
WIRED reported that the next round of pilot work is expected to extend to at least three more states, although the EIA has not yet published a national timeline or enforcement mechanism.
The pilot is arriving at a moment when utilities, regulators, and lawmakers are already struggling to plan around data center demand with limited facility-level visibility. That uncertainty is already shaping the broader conversation around AI’s power crisis and who pays the costs of AI data center growth.
Until planners have better facility-level data, they are still making major infrastructure calls with an incomplete picture of where future demand will land.
Why this matters beyond one survey
Even a mandatory electricity survey would leave important gaps. In its analysis of AI infrastructure reporting, the Federation of American Scientists argued that policymakers should also track water use, greenhouse gas emissions, and other lifecycle impacts, because electricity alone does not capture the full environmental footprint of AI infrastructure.
That wider pressure is showing up alongside industry concerns over powering AI’s next wave and growing public resistance to new facilities. The EIA pilot does not settle those fights, but it does make one thing clearer: data center energy use is becoming harder for both industry and government to treat as a black box.
Better reporting will not resolve disputes over grid upgrades, water use, or who should bear the cost of expansion. It will, however, make those disputes harder to ignore.
Also read: Power, pollution, and protests are fueling resistance to AI data centers.
