AI’s water problems run deep

AI's water problems are complex and far-reaching. Data centers consume vast amounts of electricity, contributing significantly to overall U.S. electricity use. By 2030, demand for AI is projected to double, straining water resources further. Cooling data centers with water is an efficient method, but electricity generation and chip manufacturing also heavily rely on water.

These factors, combined with the water intensity of AI facilities, can triple or quadruple their total water footprint. The water crisis is already severe, and AI exacerbates the issue. Cooling towers and on-site water systems often draw attention, but water embedded in the power generation and semiconductor manufacturing processes can be even more significant.

For instance, thermoelectric power plants withdraw and consume large volumes of water for cooling. Moreover, producing a single advanced processor can require thousands of gallons of ultrapure water. When considering these upstream demands, an AI facility's total water footprint can be two to three times higher than what's visible on-site.

Even as major chipmakers aim to reduce data center water consumption, the water used outside the data center in electricity generation and chip manufacturing can be nearly equal to the amount consumed within. This hidden water cost is often overlooked, leading to an understated water impact. The scale of water demand by 2030 could be nearly equal to New York City's daily water consumption.

This raises concerns about whether communities bear the environmental and resource costs of AI growth while companies reap most of the benefits. Water availability may soon shape where data centers can be built, especially in drought-prone regions like Utah, where proposed data center development has already sparked public pushback.

Many U.S. water infrastructure systems were not designed for hyperscale data centers, and communities are grappling with limited water access. To address this, data centers should start thinking about water as a strategic resource, similar to energy. This approach could involve closed-loop systems and decentralized treatment models to optimize water usage.

Proactively managing water resources could give communities and operators an advantage in permitting, cost predictability, and scalability. Ultimately, water strategy is becoming infrastructure strategy as AI continues to transform industries and economies.