AI and Air Pollution

One of AI’s many effects on the environment is its contribution to air pollution. Air pollution from AI involves tiny toxic particles that can enter your lungs and heart (like nitrogen oxides and sulphur oxides). They are proven to cause heart attacks, lung cancer, and asthma, even in small quantities 1.

These tiny toxic particles can travel vast distances through the air, even harming people far away from AI data centres and computer chip factories 2.

In 2023, at the start of the generative AI boom:

  • An estimated 360 deaths were caused by data centres polluting the air in the United States (U.S.) 3, that’s about the same number of people as in 12 classrooms.
  • Data centres in the U.S. caused a total public health burden of US$5.6 billion 3, that’s approximately $17 to every American 4.

The Health Impact is Growing

As AI data centres rapidly expand, so does the pollution and the public health impact.

By 2030, it’s estimated that air pollution from AI data centres in the United States will cause:

  • 1,300 deaths per year (or 25 deaths per week) 3. That’s almost four times the amount in 2023.
  • 60,000 cases of asthma per year 3, that’s roughly the population of a small town.
  • A public health burden of US$20 billion per year 3, That’s almost four times the amount in 2023, and more than the US spends on the Department of the Interior 5.

Poorer communities will experience the worst impacts, potentially up to two hundred times worse than richer communities 3. In fact, this pattern applies to other forms of pollution as well, giving rise to the environmental justice movement which advocates for the environmental rights and protections of marginalized communities 6.

How AI contributes to Air Pollution

AI companies contribute towards air pollution either directly, or indirectly, in three main ways.

  1. Diesel Backup Generators: AI companies use diesel backup generators to keep data centers running during power outages, and these emit harmful nitrogen oxides. These alone are estimated to cause 14,000 asthma cases and 13-19 deaths each year in the United States. 3
  2. Electricity from Fossil Fuels: Rising energy demands of AI are postponing the closure of coal-fired power plants and boosting fossil fuel use. Virginia Electric and Power Company, which supplies the world’s largest concentration of AI data centers, will now rely on coal until at least 2039 and is expanding its gas-fired plants to meet AI data center energy needs. 7
  3. Computer Chips: AI systems need large quantities of computer chips and circuit boards. Production of computer chips generates toxic materials and harmful air emissions. In 2023-2024, a single computer chip factory is estimated to have a public health impact of US$26-39 million. 3

References

  1. U.S. EPA (2024, December 12). Research on health effects from air pollution. https://www.epa.gov/air-research/research-health-effects-air-pollution ↩

  2. U.S. EPA (2024, September 30). What is cross-state air pollution? https://www.epa.gov/Cross-State-Air-Pollution/what-cross-state-air-pollution ↩

  3. Han, Y., Wu, Z., Li, P., Wierman, A., & Ren, S. (2024). The Unpaid Toll: Quantifying the Public Health Impact of AI (No. arXiv:2412.06288). arXiv. https://doi.org/10.48550/arXiv.2412.06288 ↩ ↩2 ↩3 ↩4 ↩5 ↩6 ↩7 ↩8

  4. Epstein, B., & Lofquist, D. (2021, April 26). First 2020 Census Data Release Shows U.S. Resident Population of 331,449,281. United States Census Bureau. https://www.census.gov/library/stories/2021/04/2020-census-data-release.html ↩

  5. U.S. Department of the Interior. (2022). Fiscal Year 2023 The Interior Budget in Brief. https://www.doi.gov/sites/doi.gov/files/fy2023-bib-entire-book-508.pdf ↩

  6. Skelton, R., & Miller, V. (2023, August 22). The Environmental Justice Movement. The Natural Resources Defence Council. https://www.nrdc.org/stories/environmental-justice-movement ↩

  7. Virginia Electric and Power Company. (2024). Integrated resource plan. https://www.dominionenergy.com/-/media/pdfs/global/company/IRP/2024-IRP-w_o-Appendices.pdf ↩