If the development of artificial intelligence continues at its current trajectory, it will lead to significant increases in emissions and water use by the end of the decade, a new study by Cornell University has found.
According to Cornell’s study, which was published in the Nature Sustainability journal, the current trajectory of AI development is likely to lead to 24 million to 44 million tonnes of carbon dioxide emissions annually by 2030, while annual water usage could reach between 731 to 1,125 million cubic metres.
The emissions increase is equivalent to adding between five and 10 million cars to US roadways, while the water use is equivalent to that of between six and 10 million households.
‘Out of reach’
This cumulative effect would put the AI industry’s net-zero emissions targets ‘out of reach’, the report suggests, however an ‘actionable roadmap’ based on smart positioning of facilities, faster grid decarbonisation and operational efficiency could cut the emissions increase by close to three quarters (73%) and the growth in water use by 86%.
“Artificial intelligence is changing every sector of society, but its rapid growth comes with a real footprint in energy, water and carbon,” commented 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?”
Regional effect
As an example, the research team found that by placing data centres in regions with lower water stress and better renewable energy resources, such as the Midwest, Texas, Nebraska, Montana, and South Dakota, the energy and water use required could be significantly reduced.
However, even under the most ambitious ‘high-renewables’ scenario, carbon dioxide emissions are only likely to drop by 15% compared to the baseline, while approximately 11 million tonnes of residual emissions would remain. This would require approximately 28 gigawatts of wind or 43 gigawatts of solar capacity to reach net-zero.
“Even if each kilowatt-hour gets cleaner, total emissions can rise if AI demand grows faster than the grid decarbonises,” You said. “The solution is to accelerate the clean-energy transition in the same places where AI computing is expanding.”
The lead author of the study was doctoral student Tianqi Xiao in Cornell’s Process-Energy-Environmental Systems Engineering (PEESE) lab. Co-authors included researchers from the KTH Royal Institute of Technology in Stockholm, Sweden; Concordia University in Montreal, Canada; and RFF-CMCC European Institute on Economics and the Environment in Milan, Italy. Read more here.
