Artificial intelligence’s meteoric rise has brought surprising environmental consequences, with the sector’s water consumption now exceeding the volume used by the global bottled water industry each year. According to recent research, the data centres and supporting infrastructure required to power large-scale AI models — from training high-performance systems to running everyday AI services — withdraw vast quantities of water for cooling and electricity generation, surpassing roughly 450 billion litres consumed worldwide as bottled water

This surge stems from the intense thermal demands of advanced chips and servers, which must be kept cool to function reliably. Many of today’s cooling systems draw from municipal supplies or stressed aquifers, highlighting a growing tension between tech expansion and water sustainability in regions already facing scarcity.

The study also points to a broader “transparency gap” within the tech industry, where water use isn’t systematically disclosed even as companies set ambitious goals to become “water positive” in the coming decade. Experts warn that without shifts toward closed-loop cooling, efficiency improvements and clearer reporting, AI’s water footprint — like its energy and carbon footprint — could continue to expand unchecked.

As regulators begin to consider environmental standards for digital infrastructure, understanding water use alongside energy impacts will be crucial for balancing innovation with planetary limits.