How much of a problem is AI’s water use? (arstechnica.com)

Protests have erupted across the southwestern United States, expressing concerns over the water usage associated with the rapid expansion of data centers needed for artificial intelligence (AI) growth. These facilities, essential for processing and cooling the high-temperature hardware used in AI, rely on significant water to manage heat, drawing comparisons to human perspiration. While some claims suggest that AI operations are excessively taxing water resources, the reality is more nuanced. Experts indicate that the water consumption of data centers, estimated at 66 billion liters in 2023, constitutes less than 1% of the country's total water usage, though this figure could increase sharply as AI infrastructure expands, particularly in drought-prone regions. The significance of this issue lies in its potential impact on local water supplies and ecosystems, especially in areas already facing water scarcity. Despite concerns, engineers and experts emphasize that various solutions can mitigate the strain on water resources. Initiatives such as enhancing cooling systems and restoring natural water cycles could lead to sustainable practices, promoting a "net-zero water" approach for these data centers. The ongoing dialogue between technology advancement and environmental stewardship will be crucial as the AI/ML community navigates this complex landscape.