A psychological autopsy I. An issue of unfading importance A striking trend has emerged in public discourse around artificial intelligence: a persistent and often disproportionate focus on water usage. Despite serious concerns about copyright violations, exploitative labor practices, and inflated claims about AI’s capabilities, a significant segment of the conversation—particularly among young people, progressive activists, and AI ethics advocates—has fixated on the idea that AI companies are recklessly consuming vast quantities of water, effectively threatening the planet’s water supply. The argument goes like this: data centers that train and run AI models require massive cooling systems, and those systems consume enormous amounts of water. Critics claim that this water waste is so severe that it’s tantamount to cutting off access to essential water infrastructure. “It rains, but we don’t see a drop,” they say, suggesting that the water used in AI operations is being squandered at a rate that undermines both environmental responsibility and social justice. Yet, despite the emotional weight and frequency of this concern, the actual scale of water use by the AI industry has been consistently overstated. In fact, the data shows that the environmental impact of AI on water resources is far less significant than the narrative suggests. This misperception is not without its roots. The issue gained traction after a widely cited but ultimately flawed claim in The Empire of AI, a best-selling book by tech journalist Karen Hao—someone I deeply respect for her critical voice in holding AI companies accountable. The book included a figure suggesting that training a single large AI model could require the equivalent of a person’s lifetime water consumption. This claim, while attention-grabbing, was based on a misinterpretation of data and has since been widely discredited. Further analysis from independent researchers and industry reports reveals that while data centers do use water for cooling, the amount is relatively small when compared to other major water consumers like agriculture, manufacturing, and energy production. In fact, many modern data centers are shifting toward air-cooling systems and using recycled or non-potable water, significantly reducing their environmental footprint. Moreover, the water used in AI operations is not a direct or irreversible loss. It’s typically part of closed-loop systems, where water is reused or returned to the environment after treatment. The real water stress in many regions comes from long-term climate change, population growth, and outdated infrastructure—not from AI training. So, while the AI industry has undeniably made serious missteps in areas like labor practices, transparency, and environmental oversight, the narrative around water waste does not hold up under scrutiny. It’s not that the concerns are entirely baseless—there’s always room for improvement in sustainability—but the focus on water has become a symbolic stand-in for broader anxieties about technological excess and corporate irresponsibility. In truth, the water issue in AI is less about actual consumption and more about perception. It reflects a deeper cultural unease with the invisible, large-scale systems driving modern AI—systems that feel distant, opaque, and potentially destructive. But confusing symbolic outrage with measurable harm risks diverting attention from the real challenges that need urgent solutions.