A bipartisan pair of U.S. senators is demanding answers about a gaping hole in federal energy data — one that has persisted even as artificial intelligence infrastructure has become the single fastest-growing source of electricity demand in the country. Senators Elizabeth Warren and Josh Hawley, who agree on almost nothing else, sent a pointed letter to the U.S. Energy Information Administration on June 25, pressing the agency to explain why it still lacks comprehensive, granular data on electricity consumption by data centers.
The letter is blunt. It accuses the EIA of failing to keep pace with an industry that is reshaping the American power grid in real time.
As reported by The Verge, Warren and Hawley wrote that the EIA’s current data collection methods “ichever are inadequate to track the rapid expansion of data center energy use” and called the lack of transparency a threat to grid reliability and energy policy planning. The senators gave the agency until July 25 to respond to a series of detailed questions about its methodology, its plans for improvement, and the obstacles standing in the way of better reporting.
Their frustration isn’t academic. The United States is in the early stages of what grid operators and utility executives describe as the most significant surge in electricity demand in a generation. And data centers — particularly those built or expanded to support AI training and inference workloads — sit at the center of it. The numbers are staggering. According to the International Energy Agency, global data center electricity consumption could double by 2026 compared to 2022 levels, reaching roughly 1,000 terawatt-hours annually. In the U.S. alone, some estimates project data center power demand could triple by 2030.
Yet the federal government’s primary energy statistics agency doesn’t have a reliable way to measure it.
The EIA currently tracks data center electricity use through a combination of surveys and modeling, but the data is widely acknowledged to be incomplete. Data centers are not classified as a distinct category in most federal energy reporting frameworks. They’re lumped in with commercial buildings or, in some cases, industrial facilities. That means policymakers, grid planners, and the public are essentially flying blind on one of the most consequential energy trends of the decade.
Warren and Hawley aren’t the first to raise this concern. The Government Accountability Office flagged similar issues in a 2024 report, and multiple energy researchers have called for better tracking. But the bipartisan nature of this latest push — a progressive Democrat from Massachusetts joining a populist Republican from Missouri — signals that the political pressure is intensifying. Both senators have been vocal critics of Big Tech, though for different reasons. Warren focuses on market concentration and consumer protection. Hawley tends to emphasize cultural power and censorship. On data center energy, their interests converge.
The timing matters. Across the country, utilities are scrambling to accommodate massive new loads from hyperscale data center campuses. In Virginia’s Loudoun County — often called “Data Center Alley” — Dominion Energy has warned that demand growth is straining transmission infrastructure. Georgia Power has seen load forecasts jump dramatically. In Texas, ERCOT has revised its demand projections upward multiple times in the past year, driven largely by data center interconnection requests.
These aren’t hypothetical concerns. They’re showing up in integrated resource plans, in rate cases, and in capacity auction results right now.
The fundamental problem, as the senators’ letter highlights, is that without accurate federal data, it’s nearly impossible to coordinate a national response. How much generation capacity needs to be built? Where? What fuel mix? How should transmission planning account for loads that can appear almost overnight when a tech company decides to build a 500-megawatt campus? These questions require data the EIA doesn’t currently collect at sufficient resolution.
The tech industry itself has been reluctant to disclose detailed energy figures. Companies like Microsoft, Google, Amazon, and Meta publish sustainability reports with aggregate energy consumption and renewable energy procurement numbers, but facility-level data is generally treated as proprietary. There’s a competitive dimension: energy costs and availability have become strategic factors in site selection, and companies don’t want rivals — or regulators — to have a precise map of their power footprint.
That opacity is exactly what concerns Warren and Hawley. In their letter, they argue that the public has a right to understand how shared energy infrastructure is being allocated, particularly when data center growth could affect electricity prices for households and small businesses. According to The Verge, the senators specifically asked whether the EIA has the legal authority to compel data center operators to report energy usage, and if not, what legislative changes would be needed.
That question — about authority — is the crux of the matter. The EIA collects data primarily through voluntary surveys and mandatory reporting requirements tied to specific statutes. Expanding mandatory reporting to cover data centers would likely require congressional action, which means this letter could be a precursor to legislation. Given the bipartisan interest, a bill mandating data center energy reporting isn’t far-fetched. Whether it could survive lobbying from the tech industry is another question entirely.
Industry groups have pushed back against mandatory disclosure, arguing that existing voluntary frameworks and corporate sustainability commitments are sufficient. The Data Center Coalition, which represents major operators, has emphasized that its members are among the largest corporate buyers of renewable energy in the world. That’s true. But buying renewable energy credits or signing power purchase agreements for wind and solar doesn’t change the fact that total load is growing — and that incremental demand, at the margin, is often met by natural gas or even coal.
The environmental implications are significant. AI workloads are extraordinarily energy-intensive. Training a single large language model can consume as much electricity as hundreds of American homes use in a year. Inference — the process of running a trained model to generate outputs — is less intensive per query but adds up quickly at scale. As AI adoption accelerates across industries, the aggregate energy footprint is growing faster than efficiency gains can offset.
Some companies are pursuing nuclear power as a long-term solution. Microsoft signed a deal to restart a unit at Three Mile Island. Amazon has invested in small modular reactor development. Google has signed agreements with Kairos Power for advanced nuclear. These are real commitments, but the timelines are long. Most of these projects won’t deliver power before the end of the decade, if then. In the meantime, data centers are plugging into grids that are already under stress.
And that’s where the EIA’s data gap becomes a policy emergency rather than a bureaucratic inconvenience.
Grid operators like PJM Interconnection, which manages the largest wholesale electricity market in the country, have already acknowledged that data center demand is fundamentally altering their planning assumptions. PJM’s latest load forecast, released earlier this year, showed demand growth rates not seen since the mid-20th century — driven almost entirely by data centers and electrification. Without accurate, facility-level consumption data from the EIA, federal and state regulators can’t properly assess whether the grid buildout is keeping pace.
The senators’ letter also touches on a less discussed but equally important issue: water consumption. Large data centers use enormous quantities of water for cooling, and in drought-prone regions, this has become a source of local controversy. Better federal data on data center operations would presumably encompass water use as well, though the EIA’s mandate is primarily energy-focused. Addressing water would likely require coordination with the EPA or the Bureau of Reclamation.
Recent reporting has added urgency to the debate. Reuters has documented the rapid escalation of utility demand forecasts tied to data center growth across multiple states. And the broader policy conversation about AI’s physical footprint — not just its digital capabilities — has moved from trade publications into mainstream discourse. The Wall Street Journal, Bloomberg, and the Financial Times have all published extensive coverage of the strain AI infrastructure is placing on power grids, water supplies, and local communities.
So what happens next? The EIA has until late July to respond. The agency could announce new data collection initiatives, request additional funding from Congress, or simply explain the technical and legal barriers it faces. Any of those responses would advance the conversation. But if the agency’s reply is defensive or dismissive, expect Warren and Hawley to escalate — potentially with legislation that mandates reporting and imposes penalties for noncompliance.
The broader question is whether the federal government can adapt its statistical infrastructure fast enough to keep up with an industry that moves at the speed of venture capital. The EIA was designed for a world of coal plants, oil refineries, and natural gas pipelines — industries with long development timelines and well-established reporting norms. Data centers are different. A hyperscale facility can go from announcement to operation in 18 months. By the time the EIA publishes annual statistics, the numbers are already outdated.
This isn’t just a measurement problem. It’s a governance problem. And the Warren-Hawley letter, for all its bureaucratic formality, is really asking a more fundamental question: Does the United States government have the tools to understand what’s happening to its own power grid?
Right now, the honest answer is no.
WebProNews is an iEntry Publication