AI vs almonds: which uses more water?
One California almond takes about 4 liters of irrigation water, roughly the same as 570 ChatGPT prompts at our Mid scenario. A handful of almonds out-drinks most people's year of chatbot use. That isn't an argument for or against either; it's a sense of scale.
One AI prompt (Mid)
7 mL
one almond
4 L
Prompt: one typical ChatGPT exchange at the Mid scenario (cooling plus the water behind the electricity). Item: its full water footprint, as published by the source.
One almond uses as much water as
≈ 571
AI prompts at Mid
1 dot = 10 prompts
The almond comparison became the internet's favorite AI-water rebuttal for a reason: California dedicates about 10% of its agricultural water to almonds, around 1.3 trillion gallons a year, and nobody types angry posts about marzipan. By raw volume, US data centers' direct cooling water (about 17 billion gallons in 2023, per LBNL) is a rounding error next to almonds alone.
The honest counterpoints: almond water mostly returns to the local water cycle differently than evaporative cooling losses do, almonds are food, and (the strongest one) agriculture's water problems don't excuse adding new demand in stressed watersheds. 'Almonds are worse' and 'this data center strains this aquifer' can both be true.
Where the comparison genuinely helps is personal scale: if your year of AI use computes to less than a bag of almonds, water is probably not the margin on which to optimize your footprint. The calculator below does that math for your actual usage.
Your number
What does your AI actually use?
Pick your model, set your usage, get the number, with sources.
Sources
Cited on this page, in order of appearance.
- 1Jegham, N., Abdelatti, M., Elmoubarki, L., & Hendawi, A. (2025). How Hungry is AI? Benchmarking Energy, Water, and Carbon Footprint of LLM Inference. arXiv preprint arXiv:2505.09598 (v6, Nov 2025). (accessed 2026-06) preprint
- 2Altman, S. (2025). The Gentle Singularity. blog.samaltman.com: 0.34 Wh and 0.000085 gal of water per average ChatGPT query. (accessed 2026-06) official disclosure
- 3Siddik, M. A. B., Shehabi, A., & Marston, L. (2021). The environmental footprint of data centers in the United States. Environmental Research Letters 16(6): watershed-scale direct + indirect water footprint. (accessed 2026-06) peer-reviewed
- 4Shehabi, A., Smith, S. J., Hubbard, A., et al. (2024). 2024 United States Data Center Energy Usage Report. Lawrence Berkeley National Laboratory, LBNL-2001637: 176 TWh electricity and 17B gal direct water in 2023, with 2028 projections. (accessed 2026-06) government
- 5Water Footprint Network (Mekonnen, M. M., & Hoekstra, A. Y.) (2011). Product water footprint database. Water Footprint Network: agricultural water footprints (coffee, beef, cotton, almonds). (accessed 2026-06) peer-reviewed