# Was Hank Green Right About the AI Water Problem? We Asked This Environmentalist. | EP 168 Source: https://www.youtube.com/watch?v=jCZIKA13KYI Recap page: https://rapidrecap.app/video/jCZIKA13KYI Generated: 2025-12-12T21:02:04.284+00:00 --- ## Quick Overview Blogger Andy Masley argues that the widespread fear regarding AI's water usage is largely based on misinterpretations and distractions, asserting that an individual chatbot prompt uses a minuscule amount of water (around 0.26 milliliters on-site), while the more significant environmental concern related to AI infrastructure is the massive increase in electricity demand, potentially leading to reliance on less environmentally friendly power sources. **Key Points:** - The common claim that one chatbot prompt uses an entire bottle of water is conclusively wrong, based on a misreading of an article, with best current data suggesting an on-site cost of about 0.26 milliliters per prompt. - The average American's daily water footprint is about 1,600 liters, meaning a median prompt (around two milliliters including off-site costs) contributes so little that a person would need to send about 8,000 median prompts to increase their daily footprint by just 1%. - Andy Masley, motivated by environmental concern, disputes the water narrative because focusing on it distracts from what he views as more serious issues, such as the massive power demands of AI data centers. - Data center water use in America in 2023 was estimated to be eight times the water usage of Masley's hometown (population 15,000), but this is comparable to other large industrial uses like agriculture, suggesting water systems can likely absorb it. - The power usage increase from AI is expected to be an order of magnitude greater than water use, potentially rising from 0.5% to 5% of the national electric grid by 2030, leading Masley to worry about reliance on fracking or older power generation. - The segment contrasts the AI water debate with Australia banning children under 16 from major social media platforms (Facebook, TikTok, etc.), which initiates a natural experiment on the effects of removing social media from teens, with initial success indicated by US school phone ban outcomes. - Masley dislikes his work being used by AI accelerationists but maintains his stance as an environmentalist, wanting to redirect concern from tiny personal impacts (like a single prompt) to systemic issues like the green energy transition. **Context:** The Hard Fork episode features an interview with blogger Andy Masley, known for his persistent debunking of narratives surrounding AI's environmental impact, particularly concerning water consumption. Masley, who previously worked as a high school physics teacher, advocates for shifting focus from individual AI use to systemic infrastructure concerns. The discussion is framed against the backdrop of widespread social media concern, exemplified by Australia's recent nationwide ban on social media access for users under 16, which the hosts use to pivot into discussing societal impacts of technology. ## Detailed Analysis Andy Masley vigorously argues that the narrative equating individual AI prompt usage with significant water waste is factually incorrect and distracting; he cites Google data suggesting a median prompt uses only about 0.26 milliliters of water on-site, which is negligible compared to the average American's 1,600-liter daily footprint. While Masley acknowledges the sheer scale of industrial water use by data centers—estimating 2023 usage in America was equivalent to eight new towns of 15,000 people—he contextualizes this against other massive water consumers like agriculture (potentially 1% of irrigated corn by 2030), concluding that this industrial use should be manageable by existing water systems. Masley pivots to express greater concern over AI's energy footprint, estimating its electricity demand will rise tenfold more than its water demand by 2030 (from 0.5% to 5% of the national grid), which risks keeping coal plants open. He admits that while he is an environmentalist, his work sometimes attracts allies who oppose environmentalism, a use he dislikes, but he insists on promoting accurate trade-offs. The hosts then transition to discussing Australia's new law banning social media for under-16s, which requires proactive age verification, noting that platforms like YouTube retain a logged-out experience loophole. The hosts view this as a crucial natural experiment to measure the mental health and social effects of removing social media access from a generation, drawing parallels to successful school phone bans, despite concerns about marginalized teens losing online support systems. ### AI Water Use Misconceptions - The idea that one chatbot prompt uses a whole bottle of water is based on a horrible misreading of an article - Best current data suggests on-site cost is 0.26 milliliters, or about two milliliters including off-site electricity generation costs - One pair of jeans purchase equals roughly 2 million median prompts. ### Shifting Environmental Focus - Masley prefers focusing on systemic issues, noting the electricity increase from AI will be an order of magnitude greater than water use by 2030 (0.5% to 5% of the grid) - He worries about data centers relying on fracking or less efficient power sources to meet energy demands - Masley is agnostic about the future of AI energy but wants full trade-offs considered. ### Contextualizing Industrial Water Use - 2023 data showed data center water use in America was equivalent to eight new towns of 15,000 people - By 2030, AI water use could equal about 1% of America's irrigated corn - Masley suggests data centers can provide positive externalities like tax revenue and utility upgrades to communities. ### Australia's Social Media Ban - Australia banned users under 16 from 10 major apps including Facebook, TikTok, and X, requiring proactive age verification, not simple drop-down menus - The impetus came from Jonathan Haidt's book, 'The Anxious Generation' - Platforms like Roblox and Discord were initially excluded, causing irritation among targeted companies. ### Natural Experiment & Teen Response - The ban creates a natural experiment to study mental health effects, with Australia funding research to track outcomes over several years - Teens are preparing by creating backup accounts or switching to banned apps like Coverstar and Yope, and even resorting to exchanging phone numbers for direct communication - Both hosts anticipate positive marginal effects over a 5-10 year horizon, similar to successful school phone bans.