Mint Explainer | Will water shortages derail India’s AI Data Centre boom?

India is pursuing an aggressive expansion of AI data centre capacity as part of its digital economy ambitions, attracting tens of billions of dollars in inve...

What Happened

India is pursuing an aggressive expansion of AI data centre capacity as part of its digital economy ambitions, attracting tens of billions of dollars in investment commitments from global technology companies.
However, this boom carries a largely underreported environmental cost: AI data centres are extraordinarily water-intensive, consuming far more water than conventional IT infrastructure because of the cooling demands of high-density GPU servers used for AI workloads.
Nearly 75% of India's existing data centres are concentrated in water-stressed regions — the same areas where El Niño-linked below-normal monsoons in 2026 are reducing freshwater availability precisely when demand for cooling water is highest.
Data centre water consumption in India could reach 358 billion litres annually by 2030 if current expansion trajectories continue, raising fundamental questions about the sustainability of India's AI ambitions in a water-scarce geography.

Static Topic Bridges

Evaporative Cooling and Water Consumption in AI Data Centres

AI data centres require continuous cooling to prevent GPU servers from overheating. The dominant cooling technology in India — and globally in hot climates — is evaporative cooling, in which water absorbs heat from server infrastructure and is then released as vapour. Approximately 80% of the water used in evaporative cooling is lost to evaporation rather than recycled.

GPU-based AI servers consume 2 kilowatts or more per chip — 4 to 6 times higher than conventional CPU-based servers (which consume 300–500 watts), proportionally increasing cooling water demand.
Each megawatt of data centre capacity requires approximately 25 million litres of water per year under evaporative cooling.
India's current operational data centre capacity is approximately 1.5 gigawatts; planned expansion would increase this six-fold.
Alternatives to evaporative cooling include closed-loop liquid cooling (which recirculates water rather than evaporating it) and immersion cooling (submerging servers in dielectric fluid), both of which dramatically reduce water consumption.

Connection to this news: India's rapid transition to AI workloads — which require GPU-heavy infrastructure — means per-megawatt water consumption is rising, not falling, even as individual data centres become more efficient in traditional IT metrics.

India's Water Stress: The Geographic Mismatch

India's data centre industry has clustered in major metro areas — Mumbai, Hyderabad, Chennai, Pune, and the Delhi-NCR — which offer connectivity infrastructure, power availability, and talent pools. However, the Central Water Commission and NITI Aayog assessments consistently classify most of these regions as water-stressed, meaning per capita water availability is below 1,700 cubic metres per year.

The National Water Mission (one of eight missions under India's National Action Plan on Climate Change) targets a 20% improvement in water use efficiency and maintenance of the ecological flows of rivers.
India receives approximately 4,000 billion cubic metres of precipitation annually, but uneven distribution, poor storage infrastructure, and groundwater depletion mean large urban regions face seasonal and structural water deficits.
The 2026 below-normal monsoon (linked to El Niño conditions) reduces reservoir levels and groundwater recharge in the same southern and western cities that host the majority of data centres.
Around 54% of India's groundwater blocks are already over-exploited or critical, according to the Central Ground Water Board.

Connection to this news: Locating water-intensive AI infrastructure in already water-stressed regions creates a compounding vulnerability — infrastructure that consumes more water when temperatures rise is placed in areas where rising temperatures simultaneously reduce water supply.

Environmental Impact Assessment (EIA) and Technology Infrastructure

Under the Environment (Protection) Act 1986 and the EIA Notification 2006, large industrial and infrastructure projects in India require mandatory environmental clearance through an EIA process. However, data centres have historically been classified as IT infrastructure rather than industrial facilities, exempting them from EIA requirements in most states.

The MoEFCC (Ministry of Environment, Forest and Climate Change) has not yet issued a national policy specifically classifying large data centres as requiring EIA clearance.
Some states have begun requiring Environmental Management Plans from large data centre projects as a condition of building permits.
EIA gaps for data centres mean water withdrawal, wastewater discharge, and energy source reporting requirements are inconsistently applied across jurisdictions.
Calls for mandatory Water Impact Assessments (WIAs) for data centres above a defined capacity threshold are growing in policy discourse.

Connection to this news: The absence of a mandatory EIA pathway for AI data centres means India lacks systematic data on the aggregate water withdrawal impact of its expanding data centre industry — a regulatory gap that must be addressed for credible sustainability planning.

India's National Action Plan on Climate Change (NAPCC) and Digital Infrastructure

India's NAPCC (2008) established eight National Missions, including the National Water Mission and the National Mission on Sustainable Habitat. These missions set the normative framework for sustainable urban infrastructure. However, they pre-date the AI data centre era and do not address the intersection of digital infrastructure and water resource management.

The National Water Mission targets 20% improvement in water use efficiency across sectors.
India's updated Nationally Determined Contribution (NDC) for 2031–2035, approved in 2026, commits to reducing the emissions intensity of GDP by 47% from 2005 levels and achieving 60% non-fossil fuel electricity capacity — directly relevant to the energy source choices of data centres.
The Union Budget 2024-25 allocated ₹551.75 crore to MeitY for AI infrastructure including GPU procurement, but sustainability requirements were not attached to this allocation.
Proposed solutions in policy discourse include: mandating use of treated wastewater (not potable water) for data centre cooling, requiring closed-loop cooling technologies for all new facilities above a capacity threshold, and incentivising coastal locations (where seawater cooling is feasible) through differential land pricing.

Connection to this news: Aligning AI infrastructure investment with India's climate commitments requires explicit environmental conditionalities on the IndiaAI Mission's compute investments — a policy integration that has not yet occurred.

Key Facts & Data

India's data centre water consumption could reach 358 billion litres annually by 2030, up from approximately 150 billion litres in 2025.
Each megawatt of AI data centre capacity requires approximately 25 million litres of water per year under evaporative cooling.
Approximately 75% of India's data centres are in water-stressed regions.
GPU servers consume 2 kW or more per chip — 4–6 times the power (and proportional cooling demand) of conventional CPU servers.
India's operational data centre capacity is approximately 1.5 gigawatts, with plans for a six-fold expansion.
Approximately 80% of water in evaporative cooling systems is lost to evaporation and not recycled.
India's updated NDC (2026) targets 47% reduction in GDP emissions intensity and 60% non-fossil fuel capacity by 2035.
54% of India's groundwater blocks are over-exploited or critical, per the Central Ground Water Board.