America's pollution problem is already deadly. The AI boom threatens to make it worse.
A study published in November 2025 and led by Boston University School of Public Health researchers [1] showed groundbreaking results. The study estimated that 46.6 million people in the contiguous US live within about a mile of at least one piece of fossil fuel infrastructure. That represents 14.1 percent of the population. With 46.6 million people living this close to fossil fuel infrastructure, the US faces a dangerous reality. The researchers themselves say the health hazards of many of these facilities have not been fully studied, which means millions of people are living next to risks that nobody has measured yet [1].
What has been measured is bad enough. A study published in Science Advances in August 2025 and reported by The Guardian [2] found that air pollution from oil and gas causes more than 90,000 premature US deaths each year. It also linked the sector's nitrogen dioxide emissions to 216,000 childhood-onset asthma cases a year and its hazardous air pollutants to 1,610 lifetime cancer cases. The researchers used data from 2017, the most recent complete year available. Between 2017 and 2023, US oil and gas production increased by 40 percent and consumption increased by 8 percent, so the researchers say their numbers are most likely conservative [2].
While these effects are already outrageous, the continued development of artificial intelligence threatens to make them much worse. The World Meteorological Organization reports that 2024 was the hottest year on record at about 1.55°C above pre-industrial levels, and that 2025 was about 1.43°C above [3]. AI did not cause that warming, but it is arriving at the worst possible time. The danger is how fast its energy demand is growing and what is being built to power it.
The amount of energy and water needed to run AI data centers is simply outrageous. A March 2026 investigation by Consumer Reports [4] reported that hyperscale data centers demand "a nearly unimaginable amount of energy." The International Energy Agency says a typical hyperscale data center can use 100 megawatts (MW), as much electricity as 100,000 households [4]. However, this isn't all of it. We are also starting to transition into a gigawatt era, where a single data center campus uses at least 1 GW of power, which equates to the demand of about 1 million homes. Meta's Hyperion campus in Louisiana will need at least 5 GW to run [4]. Water is a problem too. A data center can use up to 5 million gallons of water a day for cooling, the same as more than 16,000 average US households, according to Environmental Protection Agency estimates cited by Consumer Reports [4].
Bloom Energy's 2026 Data Center Power Report [5], citing McKinsey projections, predicts that the total power demand of US data centers will nearly double between 2025 and 2028, jumping from about 80 to 150 gigawatts (GW). Consumer Reports compared that to adding a country with the energy needs of Spain in just three years [4]. Bloom sells on-site power to data centers, so it has a stake in these numbers. But if they are even close, the "gigawatt era" might actually come sooner than expected.
It's no wonder that electricity costs across the US are skyrocketing. Home electricity prices rose 7.1 percent in 2025, more than double the inflation rate, and by more than 20 percent in some states, according to federal data cited by Consumer Reports [4]. They are still climbing. The US Energy Information Administration's latest update shows the average residential price in July 2026 was 4.9 percent higher than a year earlier, and average electricity prices rose in 44 states and the District of Columbia [6]. With data center demand expected to jump, they are likely to keep rising. The AI data center rush isn't the only factor. Inflation and an aging power grid are pushing prices up too. But it is a significant one [4].
This is where the two problems meet. Many new data centers cannot get connected to the grid fast enough, so they are building their own power on site. One analysis found 46 planned data centers, with a combined capacity of 56 GW, that will skip the grid entirely, and for now that power will overwhelmingly come from fossil fuels [4]. That means more gas turbines and diesel generators, often close to homes and schools, in a country where 46.6 million people already live next to fossil fuel infrastructure [1].
The dangers that are being posed by Artificial Intelligence in the status quo are simply too large. When the US lets "Big Tech" run this country and develop such a potentially dangerous resource, there must be action. Whether the US establishes a moratorium on data centers, a stricter set of regulations, or even shuts them down completely, which may be overdoing it, the US must take action. If we are wasting the opportunity to create a stronger energy grid on Artificial Intelligence, the US is not protecting its people. As stated before, millions are already in danger, and with the growth of AI, more fossil fuel power will be built, electricity bills will climb, and the environment will become worse.