The federal government doesn’t know how much energy AI uses. That’s the problem this survey is designed to start solving.
The U.S. Energy Information Administration launched a voluntary pilot survey targeting AI data centers, as reported by the Midland Reporter-Telegram. The survey identifies 196 companies operating 1,066 data centers across three regions, Texas, Washington state, and the Northern Virginia/Washington D.C. corridor, as candidates for voluntary participation. The EIA is the primary authority here; the Midland Reporter-Telegram is the confirmed reporting source. The survey was announced around March 27, 2026. For the most authoritative version of the survey’s scope and methodology, the EIA’s own publication is the primary reference: eia.gov.
Texas accounts for the largest share of targeted facilities among the three regions, according to reporting on the survey parameters. Specific breakdowns by company count and data center count per region appear in coverage but weren’t confirmed in the source material available for this cycle, take the aggregate figures (196 companies, 1,066 data centers) as confirmed, and the regional sub-figures as reported context that deserves verification from the EIA directly.
The “voluntary” framing deserves attention. Pilot surveys in federal regulatory contexts are typically the first stage of a sequence. The EIA runs this one, collects data, assesses completeness and industry participation rates, and uses that baseline to inform whether a mandatory reporting framework is warranted. No mandatory framework has been proposed. That isn’t a guarantee it won’t be. Companies in scope, particularly those operating in Texas, where the AI infrastructure buildout is most concentrated, should treat participation as a relationship-building exercise with a regulator that is clearly expanding its attention to this sector.
The timing isn’t coincidental. In the same 48-hour window, Meta announced a $10 billion-plus data center investment in El Paso, Texas. The EIA’s survey targets Texas as its largest region with 429 data centers identified. These aren’t connected events, but they’re connected conditions: the scale of private infrastructure investment in AI is outrunning the government’s ability to measure its energy implications, and the EIA is beginning to close that gap.
For data center operators in the three surveyed regions, the practical questions are immediate. What does participation involve? What data is being requested? How will responses be used? The EIA’s survey documentation will answer those questions more precisely than any secondary source can. Operators should retrieve that documentation directly rather than relying on press summaries.
For investors and infrastructure strategists, the survey signals something broader: federal attention to AI energy demand is no longer notional. The EIA has now defined a scope, three regions, 196 companies, 1,066 data centers, and is actively gathering the baseline data that future energy policy frameworks will be built on. That’s a different posture than monitoring from a distance.
The energy measurement gap won’t stay a gap. The question is how fast it closes and whether the industry shapes the measurement methodology before the methodology shapes the industry.