AI programs consume large volumes of scarce water
Danelski, David. 'AI programs consume large volumes of scarce water.' UC Riverside News, 28 April 2023.
Chapter 10. Mo Had Done the Math
10 sources
-
-
The Water Needed to Have the Dutch Drink Coffee
Roughly 140 litres per 125 ml cup.
Chapagain, A.K. and Hoekstra, A.Y. The Water Needed to Have the Dutch Drink Coffee. UNESCO-IHE / Water Footprint Network, 2003. (Roughly 140 litres per 125 ml cup.)
-
Energy and AI: Energy demand from AI
Data-centre electricity roughly doubling to about 945 TWh by 2030; geographic concentration.
International Energy Agency. Energy and AI: Energy demand from AI. IEA, 2026. (Data-centre electricity roughly doubling to about 945 TWh by 2030; geographic concentration.)
-
Black Canyon City Water Improvement District; Mountain Water & Sanitation District; Southwestern Water Conservation District. Board meeting minutes, accessed 23 June 2026
board minutes; source form is usable but may need exact document title
Black Canyon City Water Improvement District; Mountain Water & Sanitation District; Southwestern Water Conservation District. Board meeting minutes, accessed 23 June 2026.
-
Day Zero: Frequently Asked Questions
Municipal water restrictions, 50 litres per person per day, the ‘Day Zero’ naming and communications campaign.
City of Cape Town. 'Day Zero: Frequently Asked Questions.' 2018. (Municipal water restrictions, 50 litres per person per day, the 'Day Zero' naming and communications campaign.)
-
Day Zero: How Cape Town survived its water crisis
Sorensen, Paul. 'Day Zero: How Cape Town survived its water crisis.' The Guardian, 4 May 2018.
-
Written testimony in support of Minnesota SF 3852, Senate Committee on Environment, Climate, and Legacy, 26 March 2026. https://assets.senate.mn/committees/2025-2026/3130_Committee_on_Environment_Climate_and_Legacy/Mo-Feshami_SF-3852.pdf
Professional background; the 2.93 million gallons-per-day contract under City of Farmington Resolution 2024-111, signed 2 December 2024; the comparison to average daily use; the closing question.
Open this source Read the original
Feshami, Mo. Written testimony in support of Minnesota SF 3852, Senate Committee on Environment, Climate, and Legacy, 26 March 2026. https://assets.senate.mn/committees/2025-2026/3130_Committee_on_Environment_Climate_and_Legacy/Mo-Feshami_SF-3852.pdf (Professional background; the 2.93 million gallons-per-day contract under City of Farmington Resolution 2024-111, signed 2 December 2024; the comparison to average daily use; the closing question.)
-
Farmington residents push back against massive data center projected to double city’s water use
The 2.5 million square foot hyperscale project; the 708 MW figure; Farmington’s roughly 2.14 million gallons-per-day current use.
Webber, Melissa. 'Farmington residents push back against massive data center projected to double city's water use.' CBS News Minnesota, 11 February 2026. (The 2.5 million square foot hyperscale project; the 708 MW figure; Farmington's roughly 2.14 million gallons-per-day current use.)
-
Minnesotans Testify About the Need for a Large-Scale Water Permit Process
Minnesota Women's Press. 'Minnesotans Testify About the Need for a Large-Scale (Data Center) Water Permit Process.' 31 March 2026.
-
Making AI Less “Thirsty”: Uncovering and Addressing the Secret Water Footprint of AI Models
AI water-use estimates and the frequently misquoted ‘bottle of water’ claim: training GPT-3 in Microsoft’s US data centres estimated to have consumed millions of litres of water; roughly 10 to 50 medium-length responses could consume about a 500 ml bottle of water, depending on when and where the model is deployed; water efficiency varies spatially and temporally.
Open this source Read the original
Li, Pengfei; Yang, Jianyi; Islam, Mohammad A.; and Ren, Shaolei. 'Making AI Less "Thirsty": Uncovering and Addressing the Secret Water Footprint of AI Models.' arXiv, 2023. arxiv.org/pdf/2304.03271. (AI water-use estimates and the frequently misquoted 'bottle of water' claim: training GPT-3 in Microsoft's US data centres estimated to have consumed millions of litres of water; roughly 10 to 50 medium-length responses could consume about a 500 ml bottle of water, depending on when and where the model is deployed; water efficiency varies spatially and temporally.)