The Engine Room of the Digital World
At its core, a data centre is a physical building filled with thousands of powerful computers, or servers, that store, process, and distribute the data behind everything we do online. Think of it as the engine room of the internet. Just like any engine,
these servers generate an immense amount of heat as they work. To prevent them from overheating and failing, this heat must be constantly removed. The most common method involves cooling systems that use water. In many facilities, water is used in cooling towers where it evaporates to dissipate the heat from the servers, a process that requires a continuous and substantial supply of fresh water.
Why AI Changes the Cooling Game
Traditional computing, like streaming a video or sending an email, creates a predictable thermal load. The rise of artificial intelligence, however, has dramatically intensified the problem. Training large language models (LLMs) like those behind ChatGPT, or running complex AI queries, involves trillions of calculations. This keeps the specialised processors, known as GPUs, running at maximum capacity for extended periods, generating far more concentrated heat than conventional server racks. This increased thermal density means that AI-focused data centres require more aggressive and resource-intensive cooling, which in turn leads to significantly higher water consumption to keep the hardware from burning out.
A Thirst of Staggering Proportions
The numbers associated with AI's water footprint are startling. Research has shown that a single medium-sized conversation with an advanced AI can 'drink' the equivalent of a 500ml water bottle. On a larger scale, a single large data centre can consume up to 5 million gallons of water per day—comparable to the water usage of a town with 50,000 people. Tech giants have reported massive increases in their water consumption directly linked to AI expansion; Microsoft's water use, for example, rose 34% in one year due to its AI infrastructure growth. Projections estimate that by 2027, global AI demand could require between 4.2 and 6.6 trillion litres of water withdrawal, a volume that could satisfy the annual water needs of the entire population of Canada.
The Indian Context: A Looming Challenge
For India, this global issue has acute local implications. The country is rapidly expanding its data centre capacity to power its own AI ambitions and digital economy. However, this boom is happening in a nation that supports 18% of the world's population with only 4% of its freshwater resources. Many new data centres are being built in or near cities already facing extreme water stress, such as Bengaluru, Chennai, and Mumbai. India's data centres were estimated to consume around 150 billion litres of water in 2025, a figure projected to more than double by 2030. This creates a direct conflict between the goals of technological advancement and the urgent need for water security for agriculture and urban populations, a tension made worse by a lack of regulation requiring water use disclosure from these facilities.
The Search for Sustainable Solutions
The tech industry is aware of the problem and is exploring innovative solutions. One of the most promising is a shift from air-based cooling to more efficient liquid cooling methods. Techniques like 'direct-to-chip' cooling use pipes to deliver a coolant directly to the hottest components, while 'immersion cooling' involves submerging entire servers in a non-conductive dielectric fluid. These methods are far more effective at heat transfer and can dramatically reduce water consumption. Other strategies include locating data centres in colder climates to take advantage of natural cooling, using recycled or non-potable water, and improving transparency by reporting water usage efficiency (WUE). However, there are trade-offs; some water-saving methods can increase energy consumption, making the choice of solution highly dependent on local conditions.
