India should think about establishing a framework for disclosure of all the basic information about water sources, electricity usage and so on for large data centres
Anuradha P.S.
Published 16.09.26, 07:56 AM
Artificial Intelligence is presented in an abstract manner. Algorithms learn. Models think. Cloud data flow. However, every chatbot, image generator, and predictive model has a physical infrastructure of servers, data centres, transmission networks and cooling systems behind it.
These facilities use land, electricity and/or water, depending on their cooling system. The geography of AI infrastructure could be as critical as the technology itself. An increase in data centre capacity has been observed, from approximately 375 MW in 2020 to approximately 1,500 MW by 2025.
As part of the nation’s AI compute framework, over 38,000 GPUs have been on-boarded. The government projects the electricity consumption for data centres is likely to grow to 13.56 GW by 2031-32. These statistics showcase India’s digital aspirations. But what about this infrastructure’s environmental imprint?
Over 50% of India’s data centres are in water-stressed areas as per the World Resource Institute, India; 75% of the facilities are in Maharashtra, Tamil Nadu, Karnataka, Telangana and Uttar Pradesh.
A hint of the conflict is already apparent in Thane. There is concern of shortage of water supply for a hyper-scale data centre project at an estimated shortfall of around 30 million litres a day. According to the municipal commissioner, 12 MLD of recycled sewage water will be provided to the project. Amazon refuted allegations it would be using the water for cooling.
This conflict is more pronounced in Tusiana village of Greater Noida. A data centre park is emerging near the village whose residents have voiced concerns about the groundwater. According to Down To Earth, they were worried about threats to aquifers when the data centre complex expanded. The company, Yotta Data Services, did not use ground water or drinking water but, instead, used treated non-potable water and its cooling system was closed-loop and air-cooled.
Bengaluru has already learnt how easily water scarcity can become an everyday, urban crisis. Karnataka is also one of the five states which account for three-quarters of data centres in India.
An incident of contamination in a data centre’s fill-and-flush wastewater was linked to a rare bacteria in the reclaimed-water system in Cheyenne, Wyoming, by local wastewater officials. There was no impact on drinking water. An independent test ordered by the project’s general contractor reportedly showed no evidence of the bacterium. Cheyenne officials, however, prohibited such discharges into the city’s sewer system.
India should think about establishing a framework for disclosure of all the basic information about water sources, annual and peak usage, recycled water usage, electricity usage, cooling technology and so on for large data centres. The disclosure should be at the facility level and not limited to national sustainability reports.
Approvals may include a ‘computational carrying capacity assessment’. The same must be done for groups of digital infrastructure. Planners should consider the total water and electricity usage of existing and planned data centres prior to approving new, large data centres in areas under stress.
Incentives should be offered for both location and efficiency. Data centres are a fiercely competitive market among states. However, an advanced cooling facility in a water-constrained area could pose a higher risk over the long-term than a facility placed in a water resilient area.
A ‘resource sensitivity map for digital infrastructure’, with suggestions on how water-stressed, grid-capable and climate-resilient areas are best suited to host data centres, may also be considered.
The next digital divide is ecological. The next phase of AI policy in India should thus pose a query on top of the questions of the number of GPUs the country has, or how much data centre space it can create. It must reflect on the environmental impacts of data centres.
Anuradha P.S. is professor, department of commerce, CHRIST (deemed to be university)
