Logo
Nov 20, 2026 1:25 PM

A new MIT study links AI data centers to 650 premature deaths annually, revealing that some of the worst air pollution impacts occur hundreds of miles away, often in rural communities with few data centers.

The rapid expansion of artificial intelligence data centers is contributing to air pollution linked to hundreds of premature deaths across the United States each year, according to a new study that found some of the greatest health risks occur hundreds of miles from the facilities themselves.

Researchers at the Massachusetts Institute of Technology (MIT) estimate that electricity consumed by U.S. data centers in 2023 generated enough additional pollution to contribute to approximately 650 premature deaths annually.

Surprisingly, some of the highest mortality rates occurred in Midwestern states with relatively few data centers, rather than technology hubs like Virginia and Texas.

The findings, published in Environmental Research Letters, raise questions about who ultimately bears the environmental costs of America's rapidly expanding AI infrastructure.

Related stories

"When people think about the negative impacts of data centers, they usually focus on the neighbors,” lead author and MIT professor Dr. Noelle Selin said in a statement. “But our study indicates that the air quality impacts of data centers are very widespread across a large region.”

"Even if you don't think you are affected because there is not a data center near you, think again.”  

AI's Growing Electricity Demand Has an Unexpected Health Cost

Data centers already consume enormous amounts of electricity to operate servers, process information, and keep computing equipment cool.

In 2023, these facilities accounted for approximately 3.8% of U.S. electricity generation. However, researchers found that the additional power they require accounts for a disproportionately large share of electricity-related pollution.

According to the study, data center electricity demand accounted for approximately 6.7% of power-sector carbon dioxide emissions, 7.3% of nitrogen oxide emissions, and 5.6% of sulfur dioxide emissions.

To determine where those emissions originated, researchers developed a computer-based framework that simulated electricity production across the nation's interconnected power grid.

Rather than assuming data centers received electricity from the average mix of power plants in their region, the team modeled which individual generators increased production to satisfy additional demand.

They then combined those results with atmospheric simulations tracking how pollutants spread and react with other chemicals in the atmosphere.

The analysis attributed approximately 513 premature deaths to fine particulate pollution, known as PM2.5, and another 137 to ground-level ozone.

Together, those deaths represented an estimated 5.3% of all premature mortality associated with U.S. power-sector air pollution.

Why States With Few Data Centers Face Some of the Highest Risks

Perhaps the study's most surprising finding was the geographic disconnect between where data centers operate and where their pollution causes the greatest harm.

Virginia and Texas accounted for the largest data center electricity demands in 2023, consuming approximately 33.9 and 21.8 terawatt-hours, respectively.

Yet states including Nebraska, Iowa, Kansas, and South Dakota experienced some of the highest estimated mortality rates attributable to that demand.

Nebraska's estimated rate reached 6.5 premature deaths per million residents.

Researchers say this is due to two separate processes.

First, electricity grids routinely distribute power across state boundaries. A data center in Virginia, for example, can increase electricity generation at coal-fired plants elsewhere in the regional grid.

Second, pollutants released by those plants can travel hundreds of miles before affecting air quality in neighboring states.

In agricultural regions such as Iowa and Nebraska, ammonia already present in the atmosphere can react with emissions from power plants, creating additional fine particulate pollution.

Consequently, communities with relatively little data center infrastructure can experience substantial health consequences from electricity consumed elsewhere.

Annual mean air quality changes based on US data center electricity demand in 2023. (Image Source: Chen, et al, Environmental Research Letters)Annual mean air quality changes based on US data center electricity demand in 2023. (Image Source: Chen, et al, Environmental Research Letters)

Rural Americans Face Greater Health Risks

The researchers also identified an unexpected disparity between urban and rural populations.

Their analysis found that non-urban communities experienced an estimated 0.27 attributable premature deaths per 100,000 residents, compared with 0.21 deaths per 100,000 in urban areas.

That represents an estimated mortality rate approximately 29% higher for non-urban populations.

The findings question the assumption that communities closest to large data centers necessarily experience the greatest health impacts.

In fact, the researchers found little consistent relationship between a county's proximity to a major data center and its estimated mortality burden.

However, the study has limitations. Its estimates are based on modeled pollution exposure and established health-risk relationships, rather than individually documented deaths. It also uses 2023 electricity demand and excludes emissions from on-site backup generators, meaning it does not capture every potential source of pollution associated with data center operations.

Nevertheless, as AI infrastructure rapidly expands, the findings suggest that evaluating proposed facilities solely on local environmental conditions could overlook consequences for communities elsewhere.

"When people are thinking about data center siting and operations, it is important to consider more than just things like land and energy prices. They should also take air quality considerations into the equation," co-author and  MIT PhD candidate, Yuang Chen, notes. "This is a good baseline to think about where we should build data centers and how we should supply those data centers with electricity."

Tim McMillan is a retired law enforcement executive, investigative journalist, and co-founder of The Debrief. His reporting covers a wide range of breakthroughs in science and technology, with particular interests in defense, national security, the Intelligence Community, and psychology. You can follow Tim on X at @LtTimMcMillan or reach him by email at [email protected].

Comments
anonymous profile image
Powered by RoundtableAn AI/DeFi-powered Enterprise Media Operating System. Learn more at Roundtable.io.© Roundtable 2026. By using this site you agree to the Terms of Use and Privacy Policy