Synthetic intelligence (AI) is quickly rising as one of the transformative applied sciences of the twenty first century, reshaping industries, economies and on a regular basis life at an unprecedented tempo. From drafting easy emails and producing lifelike photos to powering superior chatbots, deep analysis, and decision-making programs, AI is more and more influencing how individuals work, be taught, talk and devour info.
However, it has a price, and this one is not merely in {dollars} or rupees. Each AI immediate prices electrical energy, land, minerals, and massive portions of water.
A recent report by the United Nations College Institute for Water, Atmosphere and Well being (UNU-INWEH) warns that the environmental footprint of AI is rising quickly because the know-how expands the world over.
Whereas a lot consideration has centered on vitality use and carbon emissions, one other useful resource is more and more coming below stress: water.
Water scarcity is already one of many largest rising considerations on this planet’s most populous nation, and but India is setting itself as much as change into one of many world’s busiest AI hubs. However it additionally wants to think about the water value that can come together with the fast enlargement of the AI infrastructure.
WHY DOES AI NEED WATER?
Each AI question begins inside a knowledge centre, which is actually an unlimited warehouse full of servers that retailer, course of and transmit info.
Each immediate you enter into an AI chatbot to ask one thing, write one thing, create a picture, or analysis, the system generates huge warmth to satisfy your demand and have to be cooled repeatedly to maintain functioning.
Many knowledge centres use evaporative cooling programs, the place water absorbs warmth from servers and is then launched into the ambiance. A considerable portion of that water is misplaced by way of evaporation through the cooling course of.
As AI fashions change into bigger and extra highly effective, the quantity of cooling required additionally rises sharply.
The size of that improve is quite alarming.
In accordance with the UNU-INWEH report, coaching ChatGPT-4 seemingly required round 592 million litres of water.
Whereas coaching giant AI fashions is resource-intensive, specialists notice that a lot of AI’s environmental footprint in the end comes from the billions of interactions that happen after a mannequin is deployed.
In accordance with a 2026 report by the Council on Power, Atmosphere and Water (CEEW), a typical 100-megawatt (MW) hyperscale knowledge centre, which has hundreds to a whole lot of hundreds of servers, can devour round 20 lakh litres of water per day for cooling, though precise consumption varies relying on cooling know-how and native circumstances.
“The present business customary for large-scale AI infrastructure is the 100-MW “AI manufacturing unit” — hyperscale services designed particularly to help intensive AI workloads,” mentioned Daswin De Silva, Professor of Analytics and AI at La Trobe College, Australia.
Because the business shifts in direction of even bigger MW-scale AI hubs, demand for each water and vitality is prone to rise considerably, De Silva added.
Estimates also show that India’s knowledge centres consumed roughly 150 billion litres of water in 2024-2025 and will require round 358 billion litres yearly by 2030 as digital infrastructure expands.
“Water demand varies broadly relying on design, cooling structure and native local weather circumstances,” famous Guillaume Dourdin, CEO of Veolia India, a multinational firm specialising in waste and water administration.
Citing Deloitte estimates, he mentioned {that a} 1-MW knowledge centre can devour roughly 68,500 litres of water per day. That could be a concern for India, a nation scaling up AI infrastructure whereas additionally going through water shortage in a number of areas.
INDIA’s AI BOOM MEETS WATER CRISIS
India’s AI ambition has paved the best way for a fast enlargement of digital infrastructure.
In accordance with government data, India’s data-centre capability elevated from about 375 MW in 2020 to 1500 MW in 2025.
Business forecasts counsel that determine might rise to between 8,000-10,000 MW by the top of the last decade.
Tech giants and home conglomerates are investing billions of {dollars} in new services. Google is creating a serious AI data-centre hub close to Visakhapatnam.
Microsoft is expanding its cloud and AI footprint throughout a number of Indian cities.
Amazon continues so as to add capability in India, whereas Reliance Industries has partnered with Mark Zuckerberg’s Meta to construct its first AI knowledge centre in India.
In accordance with CEEW, greater than 65% of India’s present data-centre capability is concentrated in Mumbai, Chennai, Hyderabad, Bengaluru and Noida.
Whereas water stays probably the most speedy concern, De Silva famous that enormous data-centre developments can even produce other environmental impacts, together with rising vitality demand, land use, heat-island results, noise air pollution and rising volumes of digital waste.
BUILT IN THE WRONG PLACE?
The problem is not only how a lot water knowledge centres devour, however the place they’re being constructed.
Data shows that a big majority of India’s data-centres are situated in areas already going through various levels of water stress.
That raises considerations as a result of a number of of India’s main know-how hubs have already skilled extreme water stress.
Bengaluru’s 2024 water crisis uncovered how rapidly a serious metropolis can run wanting water when reservoirs decline and groundwater extraction outpaces replenishment.
Hyderabad is projected to face vital future water shortages, whereas Chennai’s near-Day Zero crisis in 2019 stays one among India’s main examples of city water stress.
The UNU-INWEH report estimates that international knowledge centres had been related to roughly 4.5 trillion litres of water consumption by way of electrical energy era in 2025. If present tendencies proceed, that determine might exceed 9 trillion litres yearly by 2030.
Local weather change might additional complicate the image as the issue might change into extra acute as temperatures rise.
Rising temperatures improve cooling necessities inside knowledge centres, whereas more and more erratic rainfall patterns could make water provides much less predictable.
THE CHALLENGES AHEAD
The issue appears insurmountable however is fixable if addressed early.
In accordance with the CEEW, measures comparable to larger use of handled wastewater and cautious website choice can considerably cut back freshwater demand from knowledge centres.
De Silva argued that future cooling programs ought to more and more depend on reclaimed water, seawater, closed-loop liquid cooling and hybrid dry-cooling applied sciences quite than freshwater wherever attainable. Such approaches, he added, may help cut back stress on water-stressed areas as AI infrastructure inevitably expands.
For a rustic already grappling with recurring water shortages and rising demand, the query is not about whether or not AI will devour water. It’s about how a lot, the place, and who will bear the fee.
– Ends