This article was originally published at The Conversation. The publication contributed the article to Space.com’s Expert Voices: Op-Ed & Insights.
Even before an Elon Musk-owned artificial intelligence company opened a data center in southwest Memphis, Tennessee, air pollution was so bad that residents of a nearby neighborhood were far more likely to get cancer from industrial air pollution than average Americans. Our analysis found that air pollution got only slightly worse as a result of the data center.
The xAI Supercluster began operations on Sept. 1, 2024, powered by natural gas turbines that began operating before the company applied for the required air pollution permits. As environmental health researchers at the University of Memphis, we were immediately concerned about the potential for the turbines to pollute the air even more and decided to investigate.
Combustion from natural gas turbines releases a complex mixture of pollutants, including fine and coarse particulate matter, nitrogen dioxide, sulfur dioxide and hazardous chemicals such as benzene, toluene, ethylbenzene and xylenes. Each of these compounds has been linked to serious health consequences, such as respiratory and cardiovascular illnesses, neurological effects, cancers and elevated mortality rates.
Southwest Memphis is home to predominantly Black people with low incomes. Local residents were concerned that the new data center would worsen long-standing problems of industrial pollution in their community, which includes levels of fine particulate matter, sometimes known as PM2.5, that have long been at or near the level the U.S. government says is the maximum allowable concentration.
There were, and still are, no permanent air-quality monitors operating in the neighborhoods of southwest Memphis that are closest to the xAI data center. So we developed an approach that combined several types of measurements and calculations to determine what air pollution was like in the area before the data center opened, and what, if anything, changed after it was up and running.
Examining multiple pictures
We focused on two neighborhoods: the Boxtown Subdivision, located 2½ miles (4 km) east of xAI, which is the community closest to the facility, and the Riverview Subdivision, 6.8 miles (11 km) northeast, a known air pollution hot spot near multiple industrial and traffic emissions sources.
To create a picture of local air quality, we looked at three elements. Using company-provided technical details about the turbines and information about how many were running at any one time, we examined how emissions move through the air in the local area. We looked at satellite data showing fine-particle pollution both before and after the turbines began operating. And we looked at data on current air pollution levels in Boxtown using a third-party company’s monitors on the ground.
The company reported to county health officials that the turbines would emit different amounts of 11 different pollutants, including 30 tons of sulfur dioxide and 94 tons per year of carbon monoxide. Using a computer model recommended by the U.S. Environmental Protection Agency, we calculated how that pollution would spread across the neighborhoods.
Our calculations found that the xAI turbines would contribute minimally to ambient air pollution in both neighborhoods. We also calculated that concentrations of sulfur dioxide, carbon monoxide and nitrogen dioxide would remain well below national standards.
Fine particulate matter
The modeling estimated that fine particulate matter would increase about 1% – though that increase would come on top of a level of fine particulate matter pollution that was already higher than the national limit.
We released our initial findings, based on the computer modeling of company-reported emissions, in March 2025. Since then, our findings have been confirmed by additional research involving direct measurements of air quality in the area.
To examine whether the xAI turbines had, in fact, increased fine particulate matter concentrations, we compared satellite measurements from before and after Sept. 1, 2024. The comparison showed no significant changes.
In addition, an independent accredited lab conducted a two-day monitoring campaign in June 2025. Its findings confirmed that our model’s predictions for fine particulate matter, carbon monoxide, nitrogen dioxide and formaldehyde aligned closely with observed concentrations.
Limitations remained, however: The lab’s monitoring techniques were not sensitive enough to detect trace levels of benzene, toluene, ethylbenzene, xylenes or sulfur dioxide. That makes it impossible to directly compare our model and real-world data for those pollutants.
A long-standing concern
Our analysis offers evidence that at least as of when we did our work, xAI’s natural gas turbines had not measurably degraded air quality in the surrounding neighborhoods. That said, any changes to the equipment used to generate power would likely change the data center’s emissions. And all our analyses assumed regular, normal turbine operations: Malfunctions or accidents can lead to emissions of excessive quantities of air pollutants until they are fixed or resolved.
Our findings confirmed that fine particulate matter has long been, and remains, a concern in the area. If we had access to sustained, community-based monitoring data, we could more clearly examine pollution levels and their public health effects in the community. We believe this type of monitoring by regulatory agencies and public health groups would be beneficial to the people of southwest Memphis, whether or not there is an xAI data center operating there.
Through our work, we aim to not only clarify the air pollution effects of a specific facility, but also highlight the importance of sustained scientific engagement in communities disproportionately affected by industrial emissions. By understanding and documenting the environmental health challenges faced by the residents of southwest Memphis, we hope to contribute to their ultimate mitigation.
This article is republished from The Conversation under a Creative Commons license. Read the original article.
