A new UN report warns that the rapid expansion of AI infrastructure is driving massive water and electricity consumption. By 2030, global data centers could use 9.3 trillion liters of water each year for cooling, an amount that exceeds the domestic water needs of 1.3 billion people. The report from the United Nations University Institute for Water, Environment and Health (UNU-INWEH) estimates that a single ChatGPT-style text prompt consumes about 29 milliliters of water, while training a model like GPT-4 required roughly 600 million liters. The report argues that environmental discussions around AI must go beyond carbon emissions to include water and land use. As AI becomes more prevalent, data center cooling demands will strain local water supplies, especially in water-stressed regions. This adds a new dimension to the climate impact of technology, forcing policymakers and companies to consider resource allocation alongside grid decarbonization and renewable energy sourcing.
