The India Meteorological Department (IMD)’s first long-range forecast for the 2026 southwest monsoon puts rainfall at 92 per cent of the long-period average, the lowest opening forecast in 25 years. The United States NOAA puts the probability of El Niño forming between June and August at 62 per cent. El Niño years bring higher temperatures across India, compress the monsoon, and intensify heatwaves. Higher temperatures mean higher cooling demand. Higher cooling demand means data centres consume more water. And India is currently building data centres at a scale with no historical precedent.
This is not a coincidence. It is a resource conflict with a timeline.
India’s data centre capacity reached 1.5 gigawatts by the end of 2025. Deloitte projects 8 to 10 gigawatts by 2030, a six-fold expansion in four years. The Union Budget 2026 extended a tax holiday until 2047 for foreign companies building data centres in India. At the February 2026 AI Impact Summit, Prime Minister Modi stated that India would invite the world’s data to reside here. The government’s framing is one of digital ambition. The resource arithmetic underneath it is something else.
Most data centres in India use evaporative cooling, the dominant technology in high-temperature environments. Water removes heat from servers and roughly 80 per cent of that water evaporates and is lost. Karnataka’s IT Minister Priyank Kharge told the state assembly in March 2026 that each megawatt of data centre capacity requires 25 million litres of water per year. Applied to India’s current 1.5 gigawatt operational capacity that implies 37.5 billion litres annually from data centres alone, before the planned six-fold expansion. A September 2025 Morgan Stanley report projects global AI data centre water consumption reaching 1,068 billion litres annually by 2028, an eleven-fold increase from 2024 levels, with the range running from 637 billion to 1,485 billion litres depending on cooling technology and regional energy mix.
El Niño sharpens this in a specific and underreported way.
Evaporative cooling systems consume more water as ambient temperatures rise. In Noida, where summer temperatures reach 48 degrees Celsius, a data centre draws significantly more water during a heatwave than during moderate conditions. El Niño years are heatwave years. A March 2026 preprint study led by Andrea Marinoni at the University of Cambridge’s Department of Computer Science and Technology, mapping two decades of satellite-derived land surface temperature data across more than 8,000 AI hyperscaler sites globally, found that land surface temperatures rise by an average of 2.07 degrees Celsius after a data centre begins operations, with the heat effect extending up to 10 kilometres from the facility. The authors term this the ‘data heat island effect’. More than 340 million people globally live within that radius. The study awaits peer review but the methodology draws on verified remote sensing records across multiple continents.
Nearly 75 per cent of India’s data centres are already located in water-stressed regions. El Niño raises temperatures in those same regions, increases cooling demand inside the facilities, and reduces the water available in the aquifers and reservoirs those facilities draw from. The stress compounds in both directions simultaneously.
The geography of this problem is not hidden.
Bengaluru’s data centres are concentrated in its most water-stressed eastern and southeastern corridors. Hyderabad faces a projected water deficit of 870 million litres per day by 2027. Chennai hit Day Zero in 2019 when its four main reservoirs fell below one per cent capacity. Mumbai and Chennai together account for 70 per cent of India’s total data centre absorption in 2025, according to Savills India. These are not cities with water to spare.
The governance gap is visible in the data that does not exist.
Karnataka’s IT Minister Kharge warned in the state assembly in March 2026 that data centres were heavy water and energy guzzlers and that the state needed a fundamentally greener approach. Yet Karnataka’s Data Centre Policy 2022 commits the government to facilitating uninterrupted water supply to data centre entities across the state without specifying what happens when that commitment competes with residential water security during a drought year. No state data centre policy among the 15 currently in force contains mandatory water stress mapping as an environmental clearance condition. The Union Ministry of Environment, Forests and Climate Change does not require thermal impact assessment for data centre approvals. Because data centres are classified as IT or ITeS entities rather than large infrastructure projects, they do not go through the standard environmental impact assessment process.
This matters more in an El Niño year than in a normal one.
When reservoir levels are declining and municipal supply is stretched, the question of how much an industrial facility draws from the same aquifer as surrounding residential communities is not a transparency issue. It is an allocation question. In March 2024, during Bengaluru’s drinking water crisis, BWSSB mandated a 20 per cent supply cut to bulk industrial users including tech parks. That cut came after the shortage was visible. El Niño gives India a 12-month warning that 2026 conditions will be worse. No pre-emptive allocation framework exists for data centres.
The paradox at the centre of this is structural.
The AI systems trained and deployed in these facilities include the climate forecasting models generating the warnings India is now acting on. NOAA’s operational AI systems and Google DeepMind’s GraphCast are among the tools extending El Niño prediction horizons from six months to 12 to 18 months. India’s IMD is integrating AI-based forecasting into its monsoon advisory systems. The computational infrastructure that generates early warning of water scarcity consumes water to run, and that consumption rises precisely during the climate conditions the models are predicting.
The communities bearing the water cost are not the communities benefiting from the forecasts.
Data centres are expanding fastest in peri-urban corridors where agricultural communities still depend on shared groundwater. Karnataka’s own Central Ground Water Board has found that Bengaluru extracts almost 10 times the water it recharges each year. El Niño reduces recharge further. The residents of these corridors have no mechanism to know how much an adjacent data centre draws from the same aquifer, because no agency currently requires that disclosure.
The policy correction needed already has a legal precedent within Karnataka itself.
A 2015 state provision requires thermal power plants within 50 kilometres of Bengaluru to use only treated water for cooling. The Yelahanka gas power plant draws 15 million litres per day of treated water from the Jakkur sewage treatment plant under this rule. A pilot model near Devanahalli supplies secondary treated water to a KIADB tech park with tertiary treatment on site. The infrastructure to supply treated wastewater to industrial users exists and is operational.
Extending that mandate to data centres, and making it a national environmental clearance condition rather than a state discretion, would not constrain India’s AI expansion. It would require operators to use water that would otherwise go unused from sewage treatment plants already running, instead of drawing from aquifers that El Niño is already stressing. The Union government’s own Jal Shakti Ministry has pushed treated wastewater reuse as a priority since 2019. The data centre sector has been explicitly excluded from that push by omission, not by design.
The 2026 monsoon will be below normal. El Niño will raise temperatures in the same cities where India is building its AI infrastructure. Those facilities will consume more water to stay cool precisely when less water is available. The AI systems inside them will simultaneously be generating forecasts warning of that shortage.
India is building the infrastructure of prediction while ignoring the resource cost of running it. That is the contradiction this El Niño season will make visible. Whether it produces a policy correction depends on whether the government treats data centre water consumption as an allocation problem requiring regulation before the deficit arrives, not after.
Sagari Gupta is a public policy researcher with over eight years of experience in social development, governance reforms, and data-driven policy analysis in India.
Views expressed are the author’s own and don’t necessarily reflect those of Down To Earth