Technology

Hyperscale U.S. Data Centers Consume Power Equivalent to Entire Nations

The largest data centers in the United States, known as hyperscale facilities, are consuming electricity at a rate comparable to individual U.S. states or entire nations. According to research published this week in Nature Sustainability, these facilities—which can span the size of several football fields—accounted for 82 terawatt-hours of electricity usage between May 2024 and April 2025. This figure represents approximately 1.8% of the total electricity consumption in the United States during that period, an amount researchers likened to the power usage of Missouri or the country of Denmark.

The study, led by Harvard University scientist Gianluca Guidi, involved an analysis of 403 hyperscale data centers. The findings estimate these facilities produced approximately 26 metric tons of carbon dioxide. Guidi noted that these consumption figures likely understate the current reality, citing the rapid expansion of artificial intelligence infrastructure across the country. “The data stops in April 2025, which in AI time is basically ancient history,” Guidi said, adding that the number of large-scale facilities has since more than doubled, with another 2,000 currently planned or under development. “The industry is growing faster than anyone can track it.”

Explaining the high energy demand, Guidi described the facilities as “factories that never sleep.” These windowless buildings house tens of thousands of computers that operate continuously, generating significant heat and requiring constant cooling systems. “Every email you send, every photo you post, every show you stream, every question you ask of any generative AI software goes through a building like this,” he stated.

While Guidi characterized himself as an advocate for AI technology, he urged for a more deliberate approach to infrastructure development. He argued that the environmental impact should be considered before construction begins. “There are better ways to build this, maybe that cost more, maybe they take longer,” he said. “We’ve proven we can build AI really fast. Now, let’s prove that we can build it right.”

The study arrives as international focus shifts toward the climate implications of tech expansion. Earlier this week at the United Nations COP31 climate talks in Turkey, officials launched a voluntary pledge urging governments and companies to utilize cleaner energy sources for AI operations. United Nations climate chief Simon Stiell emphasized the need for immediate action, stating, “Claims about future emissions savings cannot excuse rising pollution now, making climate disasters even worse.”

Despite the push for commitments at the COP31 summit, U.S. technology companies driving the current AI buildout have not yet signed the voluntary pledge.

Amelia Foster

Amelia Foster covers technology, artificial intelligence, digital products, and cybersecurity. She explains emerging innovations and industry developments in practical language, helping readers understand how new technologies are changing work, business, and daily life.
Back to top button