California is Looking to Mitigate Air Pollution Linked to Artificial Intelligence

Emissions from AI could exceed that of all 35 million vehicles in California by 2030, a new study found.

A visitor watches an artificial intelligence sign on an animated screen in Barcelona, Spain, on Feb. 28, 2023. Josep Lago/AFP via Getty Images

California Gov. Gavin Newsom said on Dec. 16 that the state faces unique challenges related to energy and natural resource consumption as the world leader in artificial intelligence (AI).

The governor highlighted the fact that 32 of the top 50 AI companies—including OpenAI, Anthropic, and Databricks—worldwide are headquartered in the Golden State.

“This is a world that’s radically changing,” Newsom told The Epoch Times. “California is the vanguard of this.”

Increased energy consumption to fuel demand for AI is contributing to air pollution, and risk models are failing to capture the costs, according to a newly published study.

“We want to maintain our competitiveness in this space, but we’re mindful of the energy side of this,” Newsom said.

The governor suggested that new technology known as small modular nuclear reactors—whereby carbon-free nuclear power is produced using expandable facilities with smaller footprints and lower investment costs than traditional nuclear plants—could play a role in helping the state achieve its goals.

“You’re seeing a race now around energy, just as it relates to [AI computing] and the component parts of that,” Newsom said. “That’s an evolving issue for the state of California as we continue to dominate and lead in terms of being the biggest beneficiary of this industry, and at the same time continue to advance our low carbon green growth efforts.”

Researchers from the University of California–Riverside, and the California Institute of Technology found that AI is affecting public health and is “significantly” contributing to air quality degradation by increasing the number of fine particulates injected into the air in nitrogen dioxide and sulfur dioxide emissions.

“These are really bad pollutants that can penetrate into people’s lungs, and we don’t really associate those things with AI computing because when we think about air pollution, we often think about cars on the road,” author Shaolei Ren, associate professor of electrical and computer engineering at UC Riverside, told The Epoch Times.

Training AI models can produce about the same amount of air pollution as driving a round-trip from Los Angeles to New York City more than 10,000 times, according to the study.

In 2023, health effects from AI equaled those of about one-third of the 35 million vehicles on California’s roads, and by 2030, AI-related emissions could surpass those from automobiles, the study found.

Researchers said the potential for a widespread effect exists because the industry projects electricity consumption for data centers serving AI projects to nearly triple over the next seven years, accounting for nearly 12 percent of all energy consumed nationwide in 2030.

Increased demand for data centers is putting pressure on power grids and affecting the environment by increasing water consumption and adding particulates to the atmosphere, the study and another conducted by Ren and others in 2023 concluded.

The authors suggested establishing criteria to report air pollution and other effects, and balancing workloads to take advantage of solar production and other energy sources that could mitigate health effects.

“We’re not trying to stop AI from growing, but ... at this point, it seems that we’re focusing on the benefits, and we’re sort of ignoring the immediate impacts on the health side,” Ren said. “We need to pay attention to the net costs, so we can make more informed decisions.”

The AI industry affects public health in what researchers describe as a “hidden toll.”

Estimates in the study suggested that AI-linked pollution could contribute to about 600,000 asthma cases and 1,300 early deaths in 2030, and total annual costs to public health could top $20 billion.

Human exposure to air pollution is linked to asthma, cardiovascular disease, heart attacks, and premature mortality, among other health problems.

About 4.2 million deaths globally were linked to air pollution in 2019, according to World Health Organization data, making it the second worst contributor of noncommunicable diseases.

The American Lung Association estimated in its State of the Air 2024 report that 40 percent of U.S. residents live in areas with unhealthy levels of particulates in the air.

Pollution affects communities across the country, as cross-state dispersal affects areas hundreds of miles from sources of emission, according to the AI study.

Because data centers require continual operation, power consumption is significant. Fossil fuel-fired power plants supply some of the energy, and diesel generators—which emit up to 600 times more particulates than natural gas-fueled power plants—are used for backup electricity in case of outages in many instances, with both sources contributing to air pollution, the researchers wrote.

The authors suggested a thorough cost-benefit analysis that calculates short- and long-term health effects.

Potential benefits of AI include improved efficiency, productivity, and activity, as well as reduced costs, among other things, according to university studies.

The fastest growing tech sector and potentially the most consequential innovation for a host of industries, AI will play a “significant” part in the lives of millions, according to Adam Wierman, Caltech’s Carl F. Braun professor of computing and mathematical sciences and director of information science and technology.

“At the same time, we need to make sure that we have our house in order, and that the negative impacts that come from it are recognized, quantified, minimized, and shared equitably,” he said in a statement.