Protests erupt in US arid regions over AI data center expansion

Public protests in Texas, Arizona, and New Mexico against the expansion of data centers have highlighted a friction point between artificial intelligence and environmental stability. While the global conversation focuses on the electricity demands of neural networks, the immediate trigger for local conflict is water scarcity. The crisis is not driven by the total volume of consumption—cooling servers accounted for less than 1% of total U.S. water use in 2023, roughly 66 billion liters—but by the geographic concentration of this demand in arid regions. The environmental impact of AI is split between two distinct mechanisms: direct consumption and indirect consumption. Direct water loss occurs through evaporative cooling, where water absorbs heat from processors—which can reach 80°C—and evaporates into the atmosphere. Indirect consumption happens upstream, where vast amounts of water are required to cool steam turbines at coal- or gas-fired power plants that generate the electricity powering the servers. This duality creates a systemic link between energy sources and water security. Shifting to wind or solar power inherently reduces the total water footprint, as these technologies require virtually no liquid to function. This suggests that the pressure on water tables is not an inevitable byproduct of AI, but a consequence of the current energy mix. As noted by Kirill Afonin, the load on water resources is tied to the specific technology used for power generation.