Summary: An April 9 letter from the EIA administrator says work on the first mandatory nationwide data-center energy survey will begin after two pilot rounds. The pilots cover Texas, Washington state, and the Washington, DC-Northern Virginia area with 196 companies identified, and are due to finish by late September 2026; no effective date for the national survey is set.
A letter dated April 9 from Tristan Abbey, administrator of the US Energy Information Administration (EIA), put the agency’s plan on record: once two pilot rounds are complete, work will begin on the first mandatory nationwide survey of data center energy use. Wired covered the development on April 15, 2026. The reply went to an unusual bipartisan pairing — Democrat Elizabeth Warren and Republican Josh Hawley, the two senators who in March had asked the agency to start quantifying the industry’s power footprint. If the survey goes through, it will mark the first time the federal government compels data center operators to disclose how much energy they consume.
Warren made her view plain: “Americans deserve to know how much energy data centers are sucking up and what that’s doing to their utility bills,” and she labeled the survey “an important first step towards holding data centers accountable.” The signal to the AI industry is that the electricity carrying model training and inference now sits inside federal statistics — and inside federal accountability.
From a Letter to a Mandate: How the Rollout Unfolds
The true schedule moves in stages. In late March — essentially in parallel with the senators’ letter — the EIA unveiled a first pilot phase covering Texas, Washington state, and the Washington, DC–Northern Virginia metro area, with 196 companies identified so far. At least three more states will join a second pilot, none of them named yet. Both rounds are due to wrap by late September 2026, and only then does development of the mandatory nationwide questionnaire begin; no date exists for when the final version would take effect. In Abbey’s words, the pilots amount to “a necessary step in the methodical development of a nationwide mandatory survey.”
The better reading is track being laid, not a rule with an effective date: the pilots exist to validate the questionnaire design and the quality of the data it produces, and the national version follows from there. For the 196 companies already named, taking part amounts to an early rehearsal of the reporting format they will eventually owe.
Inside the Pilot Questionnaire
The ask is narrow — each company supplies data for a single facility — yet the field list runs wide: annual electricity use, behind-the-meter power generation, data center classifications, cooling systems, square footage, and IT efficiency metrics, with questions adjusted to each site’s location. What the fields are chasing is no mystery: not only “how much power” but “where it comes from and how efficiently it is used.” The behind-the-meter item maps straight onto developers’ growing practice of building their own gas-fired generation on site.
Two omissions stand out: the EIA would not explain how the 196 companies were chosen, and it declined to put a timeline on the national rollout. A statistical agency staying this quiet says something about the data’s sensitivity — for an operator, electricity consumption reads almost like a profile of capacity and customer mix.
The Timing: AI’s Power Appetite Collides With Household Bills
Three pressures are cresting together. Demand from AI training and inference has sent data center electricity use soaring, and developers racing to build are increasingly putting gas-fired generation on site. Residential rates have risen too, stirring community opposition that has stalled new projects in several states. The courts have now entered as well: the NAACP just sued xAI over gas turbines it allegedly ran without permits at a Mississippi facility. With no official numbers available, these policy fights run on dueling estimates — and turning “how much power data centers actually use” from a trade secret into a public statistic is exactly the survey’s first job.
Signals for Operators and Policy Circles
Operators should read this as an early compliance signal. The pilots are statistical exercises; the national version will not be — it is mandatory reporting — and governing energy data internally, from measurement through retention to submission pipelines, becomes a genuine operating cost whose numbers end up in public statistics. Policy circles get something else: a first federal-grade dataset that reshapes the evidence behind rate reviews, siting permits, and subsidy design, while handing state regulators a baseline they can cite. For developers and product teams, the practical note is simpler still: capacity cost models need to start carrying energy reporting and local grid constraints. Power is climbing from an infrastructure footnote to a decision variable at the product level.