AI infrastructure is significantly warming surrounding areas, creating a “data heat island effect” with the potential to impact hundreds of millions of people living nearby, a new working paper found.

Using a dataset of land surface temperatures produced by NASA, a research team led by the Department of Computer Science and Technology at the University of Cambridge found from 2004 to 2024, the surrounding areas of more than 6,000 data centers worldwide saw an average increased land temperature of about 2 degrees Celsius, or 3.6 degrees Fahrenheit. In certain cases, nearby temperatures increased 9 degrees Celsius, or 16.4 degrees Fahrenheit. Researchers calculated these heat islands could be felt about 6.2 miles away from facilities, impacting up to 343 million people globally. 

“The data heat island effect could have a remarkable influence on communities and regional welfare in the future,” the study, which has not yet been peer reviewed, said.

Data centers, which store and process immense amounts of data to train AI, have become the foundation of AI-related spending, with capital expenditures for the facilities predicted to reach $760 billion in 2026, according to BloombergNEF estimates, up from $450 billion last year. Hyperscalers like Alphabet have doubled their spending on data centers this year, with Google’s parent company alone planning to invest $185 billion into AI infrastructure. The spending for these major tech firms exceeds the GDP of entire countries such as Sweden.

Data center climate impacts

The energy required to operate these data centers is immense. Modern AI runs on clusters of tens of thousands of graphics processing units (GPUs), which generate huge amounts of heat requiring ventilation and water to cool down. With some modern data centers spanning hundreds or even thousands of acres, the energy needed to power and cool the facilities can exceed a gigawatt, about enough to power between 750,000 to 1 million homes.

The magnitude of necessary power has raised concerns about the environmental impact of data centers, as well as the disruptions they can cause nearby residents. These centers can cause noise pollution, generating noise levels above 90 decibels. Prolonged volumes above 85 decibels are considered harmful to hearing. In arid climates, significant water usage to cool data centers has raised concerns for the potential of droughts.

More energy usage has also taxed the U.S.’s aging grid system—and combined with more extreme weather and increased natural gas costs, have hiked electric bills 7% as of December 2025, according to Goldman Sachs analysts. These increased electricity costs will be passed down to consumers, particularly lower-income Americans, because as businesses like restaurants grapple with increased energy costs, they may increase prices, including of food, to offset those higher costs.

“The income and spending drags will likely be larger for lower-income households because electricity accounts for a greater share of their spending, as well as for households in areas with higher concentrations of data centers where regional power markets will tighten more,” Goldman Sachs economists Manuel Abecasis and Hongcen Wei wrote in a note to clients in February.

To be sure, the heat island study has drawn criticism over how much data centers’ energy use is truly impacting the environment. Some independent researchers have noted much of the increased land heat from data center construction comes from the energy needed to construct any building where empty land and vegetation once sat, not the heat created by data center activity.

The risks of going all in on data centers

These ramifications are substantial, experts say, especially given the question marks about the sustainability of AI spending. According to a recent Moody’s analysis, of the total spending commitments made by Alphabet, Amazon, Meta, Microsoft, and Oracle, nearly two-thirds of it, $662 billion, is planned for data-center-related leases that have yet to begin. These hyperscalers issued $121 billion in new debt via bonds last year alone.

The risks of growing the data center footprint has been exacerbated by the ongoing war in Iran. Not only has the country threatened Nvidia, Apple, Microsoft, and Google with data center attacks, but slashed energy trade has strained data center supply chains.

As AI infrastructure expansion grows, so too do the financial and environmental risks associated with it.

“AI infrastructure is fundamentally an energy and cooling challenge wrapped inside a digital economy opportunity,” Lee Poh Seng, a professor specializing in thermal systems at the National University of Singapore (NUS), previously told Fortune.

Researchers, however, see a path forward to mitigate the heat island effect of AI infrastructure. They propose software-based solutions that increase the efficiency of computational methods and thus require less energy. Hardware-based solutions include improvements to integrated circuitry, or the structure of the chips themselves, to aid in energy recovery, as well as implementing hybrid cooling systems that combine “liquid cooling at the chip level with system-wide air cooling.”

“Although the impact of data heat islands can be intense (as has been previously discussed),” the recent Cambridge study said, “advances in technology in the semiconductor and energy material industries, as well as methodological developments in computer science and electrical engineering, can be used to mitigate their effects.”