Opinion | Are Data Centres Draining India’s Water? | Yogesh Kochhar

A data center houses computer servers along with the power and cooling that keep them running. As life moves online, India needs more of them: according to the Indian Government, industry capacity, measured by the electricity servers can draw, rose from about 375 MW in 2020 to 1.57 GW in August 2026, and official estimates project nearly 8 GW by 2030.

Much of it must also be sovereign. The Reserve Bank of India has required payment data to be stored only in the country since 2018, and the Digital Personal Data Protection Act, 2023, is being phased in. Where data is stored and processed is now a question of who controls it, as much as of cost.

Running servers turns electricity into heat, and removing that heat often takes water. That matters more this year than most: the 2026 south-west monsoon ended 12.6 percent below normal, according to the India Meteorological Department, and Maharashtra, Karnataka and Andhra Pradesh have together declared drought in nearly 870 taluks, an official status based on rain shortfall and its effect on crops and water storage.

Data centres did not cause this shortfall, but they sit in the same cities and often draw on the same supply, so operators must plan cooling that does not add to the strain. NITI Aayog’s Data Centre Evaluation Framework, released in October 2026, makes water and cooling one of its five pillars.

A data centre's water use depends on its climate, design efficiency and source of electricity. This paper examines those choices and what Yotta has learned from seven years of running data centers in India.

Why We Wrote This Paper?

India’s data centre industry is still young: Data Center Map, an industry database, lists about 300 data centres in India against more than 3,000 in the United States. As it grows, so do questions about its water use. In the United States, water use is one of the common concerns raised by communities opposing new projects, according to Data Center Watch, which tracks local opposition. In India the same questions are arriving earlier and in cities where water is already under pressure. They deserve clear answers.

The government has pointed to where the answer lies: “The water requirement of data centres depends on the type of cooling technologies deployed,” the Ministry of Electronics and Information Technology told the Lok Sabha in March 2026.

This paper explains how cooling choices shape water use, how Yotta cools its largest campuses in Navi Mumbai and Delhi NCR, and what their measured water usage effectiveness (WUE) shows against published global benchmarks. We begin with the figures most often quoted and what they leave out.

How Data Centres Use Water?

The most quoted figure comes from the International Energy Agency (IEA): a 100 MW hyperscale facility in the United States consumes about 2 million litres a day, equivalent to about 6,500 households. Less often quoted is the IEA’s next finding: more than 60 percent of that water is consumed not at the data centre but at the power stations that supply it.

Country-level data on data centre water consumption remains limited. The International Energy Agency estimates that data centres worldwide consumed around 560 billion litres of water in 2023, with nearly a quarter used directly for cooling and the remainder linked to electricity generation and chip manufacturing. The US accounted for approximately 66 billion litres of direct consumption, while India has no comparable official estimate. According to CEEW and Systemiq, large operators such as Yotta use closed-loop, air-cooled chillers, whereas smaller facilities typically depend on municipal water for cooling. However, the sector’s overall water consumption remains unmeasured. In March 2026, the government provided no figures in response to a Rajya Sabha query, while MeitY acknowledged that it does not maintain official data on data centre capacity. Greater transparency and regular disclosure by operators are therefore essential to assess the sector’s water footprint and identify rising demand. This paper presents Yotta’s own data as a step towards that goal.

“Water use in data centers should be part of their annual compliance reporting, which needs to be backed up by existing environmental regulations,” says Dr Nitin Bassi, Fellow and Lead, Sustainable Water, at the Council on Energy, Environment and Water (CEEW). He argues for a common template through which larger data centres report withdrawals, consumption, used water, reuse and energy use "rather than submitting disconnected information to different agencies”.

Beyond the headline figures, four points matter.

The average hides the peak. In the United States, data centre withdrawals equal less than a tenth of annual municipal withdrawals, the IEA estimates. But water systems are sized for their busiest day, and data centres draw hardest in the hottest weeks. “Peak water demand is crucial for assessing the infrastructure needs,” says Shaolei Ren, Professor of Electrical and Computer Engineering at the University of California, Riverside. NITI Aayog’s framework agrees, asking states to judge water use at peak temperatures of about 40°C, not annual averages. In India, that pressure falls on a handful of cities.

Saving water costs electricity, most of all at peak hours. Cooling has two jobs: capturing heat at the servers and releasing it outdoors. The second decides water use. Cooling towers use the least electricity, but data centres relying on them consume 70 to 90 percent of the water they draw, against 5 to 15 percent for urban homes, where most water returns as wastewater. Dry systems evaporate nothing, but their compressors work hardest on 40°C afternoons, when the grid is most stretched, and with coal generating about 70 percent of India’s electricity in 2025, according to the Centre for Research on Energy and Clean Air, that extra power carries water of its own. Our forthcoming paper on power examines this.

Not every expert weighs the two the same way. Professor Vikram Vishal of IIT Bombay’s Department of Earth Sciences argues that on-site water deserves more attention because a data centre can change its power supply but not its need for cooling. “The power side can be substituted. If electricity comes entirely from renewables, the indirect water comes down,” he says. “But the water used for cooling inside the data centre is inherent and hard to abate.”

Anshuman, Director, Water Resources Division, The Energy and Resources Institute (TERI), sees the two as linked: “One is direct and has a direct impact on the local environment, while the other is upper in the value chain and impacts overall water use. Both are important to consider in resources use and efficiency planning.” He puts water use at India’s thermal power plants at about 2.5 to 5 litres per kWh generated, depending on plant design and efficiency.

The right answer changes with the season. Many dry-cooled facilities switch on adiabatic assist in summer, pre-cooling incoming air with mist or wetted pads. This trades a little water for a lower power peak. Ren argues for making the choice deliberate: “When power is more constrained, data centers could activate adiabatic assistance to reduce the demand. When water is more constrained, data centers can avoid withdrawing water from the community.”

Liquid cooling is not, by itself, water-free. Direct liquid cooling, now spreading with AI chips, changes the first job, not the second. Its warmer coolant, at about 30 to 40°C against about 20°C in air-cooled halls, allows heat to be released without compressors or evaporation on many more days, but the water saved still depends on what happens outdoors.

The industry’s standard yardstick, WUE, measures only the litres used on site per kWh of IT energy. NITI Aayog’s framework asks states to model WUE at local peak conditions and requires it to be reported by water source. No single annual figure shows when water is drawn, where it comes from or what the power station behind it consumes.

How Yotta Cools Its Data Centres?

Yotta’s largest campuses, in Navi Mumbai and Delhi NCR, are designed with one priority: to keep servers cool without leaning on water. Both use air-cooled chillers with adiabatic assist in summer. At the Navi Mumbai campus, the chillers feed chilled water to rear-door heat exchangers on the server racks. The water circulates in a closed loop and is not consumed; a small reserve is kept solely to offset losses in the unlikely event of a leak.

“When we designed our data centres, we recognised that water would become an increasingly critical resource as India’s digital infrastructure scaled,” says Sunil Gupta, co-founder and CEO, Yotta. “We chose a cooling approach that prioritises minimising on-site water use, even when that means a little higher electricity consumption. Even so, we have offset the incremental environmental impact of the higher power use by sourcing more than 60 percent of our power from renewable sources. Our focus has been on making these choices at the design stage, rather than trying to address water constraints after the infrastructure is built. We believe responsible data centre growth must account for both energy efficiency and the local availability of water.”

“The technology I witnessed at the Yotta data centre already recirculates a lot of its water,” says Professor Vishal, who has visited one of Yotta’s campuses.

In the hottest months, adiabatic cooling pre-cools incoming air to support the chillers: from about March to October in Delhi NCR and from about March to June (and in some years again in October and November) in Navi Mumbai, but not during the monsoon. In Delhi NCR’s winter, free cooling with cool night air does part of the work.

“As the monsoon arrives, we adapt our cooling strategies while maintaining a strong focus on humidity control and infrastructure reliability,” says Rajesh Kadu, Vice President, DC Operations, Yotta.

This design has a cost: air cooling uses more electricity than evaporative cooling. The halls are designed for a power usage effectiveness (PUE) of about 1.4, meaning that for every unit of power the servers use, the building uses about 0.4 more for cooling and other needs. Yotta accepts this trade-off for water because it has a strong counter-strategy on “Green” power.

Solar and wind use far less water than coal-fired power, so a higher renewable share lowers the water used to generate a campus’s electricity. At Yotta’s campuses that share varies month to month with load and renewable supply. In August 2026, the Delhi NCR campus reached 100 percent renewable power at its current load and the Navi Mumbai campus 80 percent; the average since the start of 2026 is 65 percent.

In FY2025-26, Yotta’s largest campuses recorded a combined WUE of 0.30 litres per kWh of IT energy: about 0.2 in Navi Mumbai and about 0.3 in Delhi NCR. The difference is climate, as Delhi NCR’s hotter, drier summers keep adiabatic assist running about twice as long. Both figures are measured on total site water intake, including non-cooling uses, so they are conservative. Each is below the 0.4 that Europe’s Climate Neutral Data Centre Pact targets for 2040. For comparison, AWS and Microsoft reported 0.12 and 0.27 respectively for 2025, and the US average was about 0.36 in 2023, according to the Lawrence Berkeley National Laboratory. These are annual figures: like most dry-cooled facilities, Yotta’s campuses draw more water in the hottest months, when adiabatic assist is on, and very little otherwise.

The next step is already being implemented. AI workloads run on GPU servers that produce far more heat than air can remove efficiently, so Yotta has begun deploying direct liquid cooling (DLC), in which coolant flows through plates on the chips. Water enters the coolant distribution units at about 30 to 40°C and can be cooled by dry coolers or high-temperature chillers. Because dry coolers alone cannot cool it enough on peak summer days in Navi Mumbai and Delhi NCR, Yotta has chosen high-temperature chillers, which draw less than 0.9 kW per ton of cooling, less than conventional

chillers. The 15 to 20 percent of heat still removed by air is handled by chilled water at about 21°C. In the DLC halls, design PUE falls to 1.2 to 1.3, against about 1.4 for the air-cooled halls.

DLC is also a closed loop, filled once and topped up about once a year. Leak detection in every rack connects to the building management system, so leaks are caught early and little water is lost.

The same thinking extends to backup power. For its GPU servers, Yotta uses an N+1 configuration of uninterruptible power supply (UPS) units, with one spare rather than a full duplicate set, which lowers costs and leaves less equipment to cool, maintain and dispose of. Outside the halls, rainwater harvesting already supports the Delhi NCR site, and recharge pits are included in the design of its next building.

What Other Markets Have Learned?

India is not the first country to face these questions, and those that faced them earliest offer a warning. In northern Virginia, Dublin, Amsterdam and Johor, capacity outgrew the grid, water and land around it, and NITI Aayog’s framework names all four as cautionary cases. The responses followed a pattern: first a pause, then rules.

Singapore paused new data center approvals in 2019 and, on reopening in 2022, required proposals to show a PUE below 1.3 and meet its top green building standard. Ireland now requires large new data centers to bring their own dispatchable generation or storage before connecting to the grid. The European Union requires data centers of 500 kW and above to report water use annually. In the United States, the Southern Nevada Water Authority, which serves the Las Vegas Valley, has stopped approving new evaporative cooling systems in any commercial building, and by March 2026 lawmakers in at least 11 states had introduced bills to pause new data centers, according to Stateline, a news service that covers state policy.

Two lessons stand out. First, disclosure tends to precede regulation, because rules cannot be written for what is not measured. Second, the strictest responses, pauses and moratoriums, came where growth ran ahead of planning.

What India Can Do?

India has an advantage those markets lacked: about four-fifths of the capacity expected by 2030 has yet to be built, so today’s choices will shape most of it. The rules, however, are still catching up. The environment ministry told the Rajya Sabha in August 2026 that data centers do not, by themselves, need environmental clearance, and oversight is split between groundwater permits and state pollution control boards. A February 2026 CEEW and Systemiq study found that only 5 of 15 state data center policies include sustainability provisions, and none sets standards such as WUE.

For Operators:

Design for the peak, not the average. Annual figures understate the pressure on city supply. Plan and report water use for the hottest weeks, as NITI Aayog’s framework asks states to assess.

Keep cooling water out of the drinking supply. Closed loops, dry heat rejection and direct liquid cooling cut a facility’s draw, and treated wastewater can cover much of the rest. India already asks this of power stations: under the 2016 Tariff Policy, thermal plants within 50 km of a sewage treatment plant must use treated sewage water.

Pair low-water cooling with clean power. Dry cooling saves water on site but draws more electricity, so the saving holds only if that electricity is increasingly renewable.

Report what matters. Publish annual and peak-season WUE by water source, measured consistently across sites. Claims of being water positive should rest on a transparent water balance and independently verifiable savings, preferably through a third-party water audit, with replenishment within the same basin, says TERI’s Anshuman.

For Government:

Set one way of measuring. NITI Aayog’s framework asks BEE, BIS and MeitY to standardise how WUE, PUE and carbon usage effectiveness are measured and reported; the government says BIS has already published a WUE standard. Benchmarks should allow for climate and compute density, so that a facility in Chennai is not judged by a standard written for Dublin.

Allow flexibility. The best cooling choice changes with the season, so rules should let operators switch between saving water and saving power as conditions change, rather than fix one limit for the year, as Ren suggests.

Policy should judge water use alongside power use, taking account of climate, cooling technology and water source, says Devroop Dhar, co-founder and managing director of Primus Partners. “Performance-based benchmarks should reward overall resource efficiency and encourage technologies that minimise freshwater consumption without creating disproportionate energy costs.”

Build the treated-water supply. Data center parks should come with dedicated treated wastewater, since the framework counts recycled water only when it is “contracted, connected and available from commissioning”. AMRUT 2.0 already asks states to meet 40 percent of industrial water demand through reuse, but Anshuman cautions that India lacks fit-for-purpose reuse standards and last-mile connections.

Protect existing users. The framework asks states to prepare a 20-to-30-year seasonal water balance for each catchment so that new allocations do not reduce supply for existing users. Anshuman says cities should treat large data centers as “major, continuous industrial water users” and let the state of local aquifers and water bodies decide where they are sited.

“A city-level water balance assessment that provides information on water-deficit or surplus status will be the starting point,” says CEEW’s Bassi. “Depending on the present and future freshwater status, need for resource allocation from alternative sources, such as desalinated water or treated used water, can be decided.”

CEEW’s water balance for the Chennai basin, prepared with the Tamil Nadu government in February 2026, puts unmet demand at about 546 million cubic metres in 2025, rising to 654 million by 2050. Reusing 40 percent of treated used water, alongside micro-irrigation on farms, could cut that gap by about 63 percent.

“If India gets one thing right, it should be recycling. Treated wastewater should be used for most purposes in data centers,” says Professor Vishal. “Alongside that, we need to identify and promote the cooling technologies with the smallest water footprint and avoid building in areas that are already water-stressed.”

“Sustainability must be at the heart of India’s AI infrastructure strategy. Data center growth must align with local water availability, promote recycled water and efficient cooling, and integrate water and energy planning,” says Rama Devi Lanka, former Senior Official, Frontier Technologies, NITI Aayog and the Government of Telangana. “Our goal must be to build globally competitive AI infrastructure while safeguarding environmental sustainability and long-term water security.”

The Road Ahead

Are data centers draining India’s water? The evidence suggests they need not. A data center’s water footprint is set not by its size but by its choices: how it releases heat, when it draws water, where that water comes from and what powers it. Each is being decided now, in buildings that will run for decades.

India’s monsoon will fail again in some years. What the country can still decide is whether its data centers add to that strain or are designed around it. Yotta’s experience suggests the second path is practical today. The power side of that equation is the subject of our next paper.

“We are publishing this paper to contribute to a more informed conversation about the environmental impact of India’s digital infrastructure,” says Gupta. “Water sustainability cannot be addressed through broad claims alone; it requires transparent data, sound engineering and decisions that reflect local conditions. At Yotta, we will continue to share what we learn from operating our facilities and explore ways to improve water efficiency as our infrastructure grows. We believe the industry must work together to ensure that India’s digital ambitions grow in step with responsible resource management.”

Experts Quoted

• Shaolei Ren, Professor of Electrical and Computer Engineering, University of California, Riverside Professor Vikram Vishal, Department of Earth Sciences, Indian Institute of Technology Bombay

• Dr Nitin Bassi, Fellow and Lead, Sustainable Water, Council on Energy, Environment and Water (CEEW) Anshuman, Director, Water Resources Division, The Energy and Resources Institute (TERI)

• Devroop Dhar, co-founder and managing director, Primus Partners

• Rama Devi Lanka, former Senior Official, Frontier Technologies, NITI Aayog and the Government of Telangana

Sources:

Introduction

• National Payments Corporation of India, UPI product statistics, September 2026.

• Press Information Bureau, Ministry of Electronics and Information Technology, explainer on India’s data center capacity, 14 September 2026.

• Reserve Bank of India, Storage of Payment System Data, April 2018.

• Digital Personal Data Protection Act, 2023, and Digital Personal Data Protection Rules, notified 13 November 2025. India Meteorological Department, end-of-season statement on the 2026 south-west monsoon.

• Drought declarations by the governments of Maharashtra, Karnataka and Andhra Pradesh, 2026. NITI Aayog, Data Centre Evaluation Framework, October 2026.

Why We Wrote This Paper?

• Data Center Map, India and United States listings, accessed October 2026.

• Ministry of Electronics and Information Technology, written reply in the Lok Sabha, 18 March 2026.

• Data Center Watch, research on local opposition to data center projects in the United States, 2025 (as reported by Data Center Knowledge, 15 May 2025).

How Data Centers Use Water

• International Energy Agency, Energy and AI, 2025.

• Rajya Sabha, written reply to Unstarred Question 4074, Water consumption by data centers in the country, 30 March 2026.

• Business Standard, “Data centres turn to air cooling to ease India’s water woes”, 7 September 2026.

• Standing Committee on External Affairs, 17th Report, India’s Role in Global Tech Diplomacy, Digital Governance, Outer Space Diplomacy, AI and Emerging Innovations, September 2026 (MeitY’s submission on data center capacity).

• Yuelin Han, Pengfei Li, Adam Wierman and Shaolei Ren, Small Bottle, Big Pipe: Quantifying and Addressing the Impact of Data Centers on Public Water Systems, March 2026.

• Centre for Research on Energy and Clean Air, India Power Sector Review 2025, January 2026. ASHRAE, thermal guidelines for data processing environments.

How Yotta cools its data centers

• Amazon, sustainability report, 2026; Microsoft, environmental sustainability report, 2026. Climate Neutral Data Centre Pact, water usage effectiveness target for 2040.

• Lawrence Berkeley National Laboratory, 2024 United States Data Center Energy Usage Report, December 2024.

What other markets have learned?

• Infocomm Media Development Authority and Economic Development Board, Singapore, data center call for application pilot, July 2022.

• Commission for Regulation of Utilities, Ireland, Large Energy Users Connection Policy, Decision Paper CRU/2025236, 12 December 2025.

• European Union, Energy Efficiency Directive (EU) 2023/1791 and Delegated Regulation (EU) 2024/1364. Southern Nevada Water Authority, restrictions on new evaporative cooling systems.

• Stateline, Temporarily banning data centers draws more interest from state, local officials, 6 March 2026.

What India can do?

• Ministry of Environment, Forest and Climate Change, written reply in the Rajya Sabha, 6 August 2026. CEEW and Systemiq, Scaling India’s Data Centre Ecosystem, February 2026.

• Ministry of Housing and Urban Affairs, AMRUT 2.0 Operational Guidelines, 2021. Ministry of Power, Tariff Policy, 2016.

• Tamil Nadu Green Climate Company and CEEW, Towards Climate-resilient River Systems in Chennai, February 2026.

About the Author: Yogi Kochhar is the President of the Yotta Foundation, as well as a globally acclaimed AI evangelist, maven, and author. A prominent thought leader, his work focuses on artificial intelligence strategy, technological innovation, and driving human-centric digital transformation.

Disclaimer: The views and opinions expressed in this article are those of the author(s) and do not necessarily reflect the official policy or position of the publisher. While every effort has been made to ensure the accuracy of the information, the publisher is not responsible for any errors or omissions, or for the results obtained from the use of this information.