This resource has been factchecked by policy experts, using the latest scientific research. Find all our sources linked below.
Artificial intelligence (AI), love it or hate it, is here to stay. Among other applications, it has taken over online searches, which now default to AI-generated results. Although there are steps to take to prioritize web links in searches, there’s no way to fully opt out of AI.
While this sounds more like a cultural than an environmental concern, AI presents potentially huge climate and air pollution problems. AI systems are powered by a tremendous amount of electricity, which generates massive amounts of climate pollution—potentially even more now that pollution controls at coal and oil and gas plants are at risk under the second Trump administration. AI operations also require large quantities of water to cool the equipment, and they cover a lot of land.
With demand skyrocketing globally, AI is expected to grow. In the United States, becoming the “AI capital of the world” is one of Environmental Protection Agency Administrator Lee Zeldin’s five key priorities. AI data centers, the gigantic spaces where supercomputers and infrastructure needed to power, train, and deploy AI are housed, already exist or are being planned across the country, with Virginia and Texas considered notable hubs.
Here’s a look at what this means for air pollution, public health, and the communities living closest to AI data centers.
February 2024: Construction of a large Amazon data center next to homes in Loudoun County, Virginia.
Powering AI and air pollution
The climate impacts of AI depend largely on the source of electricity generation, as some (solar) are much cleaner and produce far less air pollution than others (coal). While the current mix of energy in the U.S. grid includes renewables, nuclear, coal, and natural gas, two-thirds of the additional power needed for AI data centers is expected to come from fossil fuels. Additional air pollutants associated with AI come from gas turbines at data centers that power the projects as well as backup diesel generators on site that supply increased electricity demands.
The enormity of the electricity needed for AI to operate is hard to translate into everyday numbers. One simplified estimate is that every AI query needs 10 times as much electricity as a regular web search. Calculations of AI’s impact can vary widely, but according to some experts, AI data centers currently represent around 4% of total U.S. electricity demand; by 2030, this number is expected to jump to 9%. This means that, if the power grid can handle the increased demand, carbon dioxide emissions of AI data centers may also more than double by 2030.
AI and water
A significant amount of water is needed to cool AI data centers. AI’s projected water usage could hit 6.6 billion cubic meters by 2027. It’s hard to imagine what that large of a number even means. Breaking it down to equivalents, some say composing a 100-word email using an AI-powered writing tool consumes as much as a bottle of water. However it’s calculated or described, the water footprint of AI is a huge problem, especially now, when many places in the world are experiencing water scarcity. In some communities, treated wastewater is being used for cooling data centers to avoid straining local aquifers.
A large Apple data center east of Reno, Nevada.
AI and human health
Air pollution associated with AI is expected to result in as many as 1,300 premature deaths per year by 2030. The weakly regulated gas combustion turbines being used to boost power and avoid outages at AI data centers alone emit hazardous chemicals, including the known carcinogens benzene and formaldehyde, and worsen ground-level ozone. Particulate matter from turbines can increase the risk of heart attacks, respiratory infections, asthma attacks, or death. The electronics at data centers can also pose health risks—from the mining of the raw materials needed for the computers, including microchips, to the resulting e-waste, which can contain hazardous substances like mercury and lead.
Increased fossil fuel energy demand for AI also increases planet-warming greenhouse gases, which in turn exacerbates extreme weather, including heat, flooding, and wildfires. Extreme weather has wide-ranging established health impacts, from asthma to cardiovascular disease to death.
January 2025: Data center cooling equipment at an Amazon data center in Virginia.
AI and community impacts
Data centers are frequently located in neighborhoods already burdened by pollution. South Memphis’ Boxtown community, for example, where Elon Musk opened an AI supercomputer facility to power his startup xAI, was already home to an oil refinery, a steel mill, and chemical plants. The neighborhood’s majority-Black residents have high rates of asthma, lower life expectancy, and four times the national rate of cancer—health impacts that have all been linked to industrial pollution.
Data centers are immense—some, including the Memphis facility, are many football fields in size—and, depending on location, can deprive nearby residents of greenspace, which can reduce air pollution. Many new data centers are planned for areas close to federal and state park land.
Another community impact is expense; these centers could increase electric rates for residential customers.
May 2025: Gas turbines outside the xAI data center in Memphis, Tennessee.
Call to action
AI enthusiasts say its potential benefits include finding climate solutions by quickly analyzing vast datasets. Still, the environmental tradeoffs for any possible progress are steep and must be mitigated. Concerned individuals can get involved by demanding companies uphold a strong commitment to climate when building out AI—and voting with our dollars to support those that do.
You can also join Moms Clean Air Force to urge elected officials to keep—not roll back—federal climate protections for AI, including those for gas turbines, as well as demand strong regulations at the local and state level where data centers exist or are currently being built.
Full list of sources.
Released: May 2025