deniz.in

Markets

Weather

Loading weather

· via The Verge

Study: US hyperscale data centers burn more electricity than 148 nations, mostly on fossil fuels

A Nature Sustainability study estimates 403 US hyperscale data centers drew about 82 TWh of electricity in a year and emitted 25.7 million metric tons of CO2, with roughly 54% of their power coming from gas and coal.

Study: US hyperscale data centers burn more electricity than 148 nations, mostly on fossil fuels

US data centers out-consume most countries

The 403 largest data centers in the United States likely drew about 82 terawatt-hours of electricity between May 2024 and April 2025, around 1.8 percent of total US power consumption, according to a study published in Nature Sustainability and reported by The Verge. John Sterman, who directs the System Dynamics group at MIT Sloan and was not involved in the research, told The Verge the figure tops the annual electricity use of about 148 individual nations, a list that includes Austria, Israel and Venezuela but excludes large economies such as the US and China.

The emissions picture is similarly outsized. Generating that electricity produced an estimated 25.7 million metric tons of CO2, a footprint the paper compares to Denmark's, and Sterman says it exceeds the total annual emissions of roughly 113 countries. About 54 percent of the power consumed likely came from natural gas and coal.

Satellites instead of disclosure

The research team, led by Gianluca Guidi, a postdoctoral fellow at Harvard T.H. Chan School of Public Health, identified the hyperscale facilities using satellite imagery, then estimated each site's power draw from its electrical capacity. Using Environmental Protection Agency data, they modeled the carbon output each facility would have generated on its local grid.

That indirect approach was necessary because operators do not publish energy consumption figures, a transparency gap that both Guidi and Fengqi You, a professor of energy systems engineering at Cornell University who was not part of the study, criticized as an obstacle to accountability.

Where a data center sits matters as much as its size

A central finding is that geography can outweigh scale. Generating electricity in coal-heavy Wyoming produced roughly seven times more carbon per kilowatt-hour than in hydropower-rich Oregon, the study found. Guidi told The Verge that two facilities drawing exactly the same amount of power can end up with very different carbon footprints purely because of the grid they plug into.

Guidi sees that variability as encouraging: the environmental consequences of the same computing infrastructure can swing widely depending on the electricity system behind it, so the impacts are not fixed. Decisions about where to build and how to power these facilities are still open choices.

What the numbers leave out

The Verge points to several limits. The study does not capture local air pollution from new gas infrastructure built to satisfy data center demand, nor from gas turbines that some AI companies have installed on-site to power facilities without a grid connection. Hyperscale sites also cover more than just AI computing, and they account for only a fraction of the thousands of data centers in the US, with more arriving each year as developers race to build capacity for generative AI.

You warned of long-term "carbon lock-in": power plants and supporting infrastructure built to meet data center demand may keep operating for decades, meaning siting decisions made now could shape emissions well into the future.

Sterman, who disclosed that he used Claude to run his country-by-country comparisons, argued that AI can be extraordinarily useful but that its harms need limiting, since CO2 already emitted cannot be undone. He says policymakers should require operators to bring new clean energy online rather than simply connect to existing supply.

Pressure for transparency grows

Guidi and his colleagues are turning the research into a website meant to track the environmental impact of individual data centers. A bipartisan bill moving through Congress could also mandate broader national disclosure of data center electricity use. Gina McCarthy, former White House national climate advisor and EPA administrator, said in a statement shared by the research authors that reliable, independent information about the facilities' effects on air, health and climate is essential for communities and policymakers.

Why it matters

The study puts hard numbers on a trend that has so far been discussed mostly in estimates: the AI buildout is being powered substantially by fossil fuels, at a scale that rivals national economies. It also reframes the debate. AI's environmental footprint is not an inherent property of the technology but a function of grid choices, so siting, power procurement and disclosure rules made in the next few years will determine whether the fastest-growing source of US electricity demand accelerates or eases the energy transition.

  • #data-centers
  • #ai
  • #energy
  • #emissions
  • #sustainability

Related posts