Silent thirst: Water footprints of AI data centres in India and routes to resilience

India’s rapidly expanding data center industry, fueled by AI goals, is worsening water shortages in water stressed cities such as Mumbai, Hyderabad, Chennai and Bengaluru. Projections show water consumption more than doubling, from 150 billion liters in 2025 to 358 billion liters by 2030- enough to match the needs of millions of people. Training intricate AI models demand massive energy and the resulting heat buildup requires substantial water to cool the operating servers. For instance, a 1 MW data center uses around 26 million liters of water each year for cooling; Yotta's 30 MW facility in Greater Noida consumes about 780 million liters annually, matching the daily water requirements of 15,830 city dwellers. Expanding to 160 MW in a single park could reach 4.2 billion liters per year, equivalent to a small town usage for 85,000 residents. India possesses only 4% of the world’s freshwater despite supporting 18% global population, so data centers in these arid urban areas siphon off drinking water, forcing nearby villages to drill borewells from 30 meters to 250 meters deep. This paper is an attempt to study the growth of data Centre in India, it highlights related challenges of water sustainability and solutions being adopted to overcome the problem.