U.S. Senate seeks data from data centers on electricity bills and grid load

In the US, a new phase of pressure on AI infrastructure is beginning: the Senate wants to see exactly how much electricity data centers consume and who ultimately pays for the growing burden on the power grid. Senators Josh Hawley and Elizabeth Warren sent a request to the US Energy Information Administration, or EIA, demanding more detailed data on industry energy consumption.

What the Senators Are Demanding

On March 26, senators sent a letter to the EIA requesting mandatory annual reporting from data centers and other large energy consumers. Their main argument is straightforward: electricity demand is growing rapidly again after a long period of relative stagnation, and the government still has no standardized and reliable data to properly plan grid development. Without such a foundation, regulators and power grid operators are essentially assessing the impact of new facilities almost blindly, especially when it comes to the rapidly growing AI infrastructure segment.

Currently, the EIA already collects a large amount of information on US energy: prices, generation sources, energy efficiency programs, and general statistics by sector. But this data is broken down only into very broad categories like residential, commercial, industrial, and transportation consumption. Hawley and Warren want the agency to identify data centers as a separate class of load. Moreover, they are interested in the difference between AI computations and ordinary cloud services, because these consumption profiles may differ significantly.

Why the Topic Has Intensified

Political pressure on the industry is mounting from several directions at once. The day before Hawley and Warren's letter, Senator Bernie Sanders and Congresswoman Alexandria Ocasio-Cortez announced they were preparing legislation to suspend the construction of new data centers until Congress agrees on AI regulation rules. This is no longer a discussion about the distant future, but a dispute over how quickly computational infrastructure can expand and who should bear the costs of such growth.

The authorities have reason for concern. According to data presented in the material, Google's data centers doubled their energy consumption between 2020 and 2024. And looking more broadly, by 2035, planned new data centers are expected to nearly triple the sector's energy consumption. For the power grid, this means not only additional megawatt-hours, but also new peak loads, more expensive network upgrades, and pressure on tariffs in regions where major facilities are being built.

What Data Is Needed

The senators are asking for not abstract estimates, but a very specific set of metrics. They want to understand not only the total consumption volume, but also how data centers behave throughout the day, how much they pay, what upgrades their connections require, and whether they participate in mechanisms that help smooth peak loads.

• Hourly, annual, and peak load of major facilities
• Tariffs that companies pay for electricity
• What grid modernizations were required for new connections
• Who pays for these network upgrades
• Whether data center customers participate in demand response programs

A separate question is the speed at which the government is even capable of collecting such statistics. EIA Administrator Tristan Abbey called the agency an "important player" in collecting data on electricity demand from data centers back in December. But he also warned that launching a new full-scale survey is not a quick process: changes to EIA surveys go through approval procedures in the Office of Management and Budget and require a period of public comments.

"Starting a new survey from scratch usually takes about two years,"

EIA Administrator Tristan Abbey said in December.

At the same time, Abbey allowed that the agency has faster mechanisms: smaller-scale surveys can be conducted if a more rapid and immediate market signal is needed. Hawley and Warren asked the EIA to respond to their letter by April 9. So the political demand for transparency has already been formed, and the question now is how quickly the bureaucratic machine can turn it into a working reporting system.

What This Means

The era when data centers could be considered just another "large building with servers" is ending. For authorities, they are becoming a separate type of critical infrastructure with their own load profile, and for the AI market, this is a signal: it seems that further growth without more stringent reporting on energy use and impact on the grid will no longer be possible.