My Research Shows Why AI Might Not Be the Water Disaster We Feared (And What Actually Is)
News ClipOutside Magazine·AZ·4/6/2026
An environmental policy expert challenges the narrative that AI is a major driver of the Western US water crisis, presenting personal research showing significantly lower water usage for AI compared to other activities. The article highlights that all US data centers combined account for only 0.3% of national water withdrawals, with agriculture being the dominant water consumer in the Western US. It urges a shift in focus from AI to the larger systemic issues affecting the Colorado River.
waterelectricityenvironmentalgovernment
Gov: U.S. Geological Survey, Bureau of Reclamation, Argonne National Laboratory
The article, titled "My Research Shows Why AI Might Not Be the Water Disaster We Feared (And What Actually Is)" by an environmental policy expert and filmmaker, challenges the prevalent concern that artificial intelligence (AI) is a significant contributor to the Western US water crisis, particularly affecting the Colorado River. The author, who is Diné and works on Colorado River water issues, conducted an 11-week personal study to quantify his AI water use.
His research involved tracking 100 AI sessions, counting tokens processed, and applying publicly available data on energy and water intensity from Epoch AI, Microsoft, and Google. The study concluded that his extensive AI work, including developing an iOS app and writing policy briefs, had a total lifecycle water footprint of approximately five gallons. This contrasted sharply with a 383-mile road trip for work, which consumed an estimated 110 gallons of water for fuel—more than 20 times his AI usage.
The author emphasizes that all U.S. data centers combined, not just those supporting AI, account for roughly 0.3 percent of total national water withdrawals. He contrasts this with agriculture, which consumes about 80 percent of Colorado River water and 86 percent of Arizona's water. He also addresses concerns about AI's exponential growth, citing a 2025 Microsoft Research paper projecting significant reductions in energy per query due to efficiency gains. Most of AI's water footprint, he notes, originates from electricity generation at power plants (Scope 2 water), not direct cooling in data centers.
The article concludes by urging a re-evaluation of focus, advocating for attention on the long-standing over-allocation of the Colorado River, exacerbated by climate warming, and the massive water consumption by industries like agriculture. While acknowledging that individual data centers can locally strain small community water supplies, the author argues that framing AI as the primary driver of the Western water crisis distracts from the larger systemic issues draining the river, such as the 2026 Compact renegotiation.