While artificial intelligence, or AI, is transforming industries, its environmental impact is often overlooked. Shaolei Ren, an associate professor of electrical and computer engineering at UC Riverside, took to a TED Talk stage last fall in Vienna, Austria, to discuss a hidden consequence of AI’s rapid expansion.
In his talk, “AI Consumes a Lot of Water – But Why?,” Ren explained how the growing demand for AI-powered computing is driving significant water usage, particularly in cloud-based data centers. Such facilities can require millions of liters of water to cool the servers that power AI models, often straining local water supplies.
Ren’s talk was held at the TED AI Vienna conference in October, which brought together world experts on AI to discuss its promise and its societal and environmental consequences. The video of his talk was just recently released by TED Conferences, a non-profit that organizes several TED Talk conferences each year. TED stands for Technology, Entertainment, and Design.
Ren’s research has found that the water consumption of AI models is substantial. For instance, training GPT-3, one of OpenAI’s powerful models, required about 700,000 liters of water through direct evaporation—enough to manufacture approximately 370 BMWs or 320 Tesla electric vehicles.
Many AI data centers rely on water-cooled systems, where electricity consumption by AI servers generates heat that must be dissipated. This process often involves water evaporation, such as cooling towers that convert freshwater into steam, releasing it into the atmosphere for cooling.
“Every time you ask an AI chatbot a question, you are also consuming water—without realizing it,” Ren told the TED audience. “AI doesn’t just require computing power; it needs cooling, and that cooling comes with a cost.” AI’s water consumption is called blue water—water from rivers, lakes, and groundwater sources—that humans can directly utilize.
Addressing water consumption is integral for advancing AI infrastructure responsibly, as extended droughts are becoming a norm. Ren emphasized that AI companies must take responsibility for their water footprints. He proposed a simple but effecti
ve solution: timing AI training during cooler hours to reduce water evaporation.
“We don’t water our lawns at noon because it’s inefficient,” Ren explained. “Similarly, we shouldn’t train AI models when it’s hottest outside. Scheduling AI workloads for cooler parts of the day could significantly reduce water waste.”
Click here to watch Ren’s full TED Talk.
