The rapid expansion of artificial intelligence infrastructure could annually add 24 to 44 million metric tons of carbon dioxide to the atmosphere by 2030, according to new research from Cornell University. This is equivalent to the emissions of adding five to 10 million cars to US roads.
The study, published in Nature Sustainability, also found the industry’s growth at its current rate would drain 731 to 1,125 million cubic meters of water per year, equal to the annual household water usage of six to 10 million Americans.
Researchers found this cumulative effect would place the AI industry’s net-zero emissions targets out of reach.
“Artificial intelligence is changing every sector of society, but its rapid growth comes with a real footprint in energy, water and carbon,” said Fengqi You, the Roxanne E. and Michael J. Zak Professor in Energy Systems Engineering in Cornell Engineering, who led the project. “Our study is built to answer a simple question: Given the magnitude of the AI computing boom, what environmental trajectory will it take? And more importantly, what choices steer it toward sustainability?”
An actionable roadmap
The study, which used AI to analyse financial, manufacturing and environmental data, outlined an actionable roadmap to mitigate these impacts. Researchers determined that a combination of smart siting, faster grid decarbonisation and operational efficiency could cut carbon dioxide emissions by approximately 73 per cent and water usage by 86 per cent compared to worst-case scenarios.
Location was identified as one of the most important factors, as many data clusters are currently being established in water-scarce regions, such as Nevada and Arizona. The study found that locating facilities in areas with lower water stress, such as the Midwest or “windbelt” states like Texas and Montana, could slash water demands by about 52 per cent. New York state was also noted as a “low-carbon, climate-friendly option” due to its clean electricity mix.
Accelerating the clean-energy transition is also crucial. The research warned that if decarbonisation fails to keep up with computing demand, emissions could rise by roughly 20 per cent. Even in an ambitious high-renewables scenario, 11 million tons of residual emissions would remain by 2030.
The researchers also determined that deploying energy- and water-efficient technologies, such as advanced liquid cooling and improved server utilisation, could remove another seven per cent of carbon dioxide and lower water use by 29 per cent.
“This is the build-out moment,” You said. “The AI infrastructure choices we make this decade will decide whether AI accelerates climate progress or becomes a new environmental burden.”
