AI Data Centres and Water

AI data centres create water risk because cooling systems can evaporate large volumes of water, and the electricity powering them also uses water. In India, the risk depends on location, cooling design and season, so siting in water-stressed areas becomes a business decision, not only an environmental one.

Until recently, the debate around data centres focused on electricity. Rapid growth in AI and cloud infrastructure is now drawing attention to water availability, cooling technology and competition with farms, homes and industry for the same local supply.

This article explains why AI workloads can raise water demand, how cooling choices change consumption, and which questions companies and communities should ask about data centre water use in India.

Servers turn almost all the electricity they draw into heat. If that heat is not removed, chips slow down or fail, so every data centre needs a cooling system. Water is widely used because it carries heat far better than air, the same principle behind a car radiator.

AI makes this harder. The chips used to train and run AI models, called GPUs, pack more computing power into each rack than conventional servers. Each rack therefore produces more heat that has to be removed.

Two terms matter throughout this article. Direct water use is the water a facility uses on site, mostly for cooling. Indirect water use is the water power plants consume to generate the electricity the facility buys. It also helps to separate withdrawal, which is water taken from a source, from consumption, which is water that does not return to that source, for example because it evaporates.

AI data centres link digital growth to local water supply through cooling and electricity. In India, location, cooling design and season decide how serious that link becomes.

As AI capacity grows, disclosure of WUE, water source and drought-month plans will be the practical test of whether operators have understood the risk.