By Pradeep S Mehta and Sohom Banerjee
Imagine two neighbours drawing from the same aquifer. One is a farmer whose family has depended on it for generations. The other is a billion-dollar data centre processing the queries, models and algorithms powering India’s artificial-intelligence (AI) ambitions. Both need water. But only one represents an industry we still instinctively describe as being in “the cloud”. Therein lies an uncomfortable contradiction.
AI may appear weightless, but its infrastructure is profoundly physical. Servers generate heat; cooling systems require water; electricity generation carries its own water footprint; and semiconductor manufacturing adds another resource-intensive layer. The question, therefore, is not whether India should build AI infrastructure, it must, but who should bear the environmental cost of making intelligence computationally abundant when water remains scarce?
The scale deserves attention. Data centres consumed about 415 terawatt-hours of electricity globally in 2024, roughly 1.5 per cent of global electricity consumption, according to the International Energy Agency. In India, CEEW estimates that data centres used approximately 150 billion litres of water in 2025, a figure expected to more than double by 2030.
Research by Quantive Advisory estimates that global AI-related water withdrawals could reach 4.2-6.6 billion cubic metres annually by 2027, underscoring the rapidly growing water footprint of AI infrastructure.
These numbers should neither trigger technological pessimism nor be dismissed as the unavoidable price of progress.
India has compelling economic reasons to expand its AI and data-centre ecosystem. Digital sovereignty, productivity, cloud services, innovation, investment and employment. AI itself can improve energy management, agriculture, healthcare and climate adaptation.
Research suggests AI could facilitate progress towards 128 targets under the Sustainable Development Goals, even as it could inhibit 58. Sustainable development, after all, is not about stopping growth; it is about ensuring that today’s growth does not borrow irresponsibly from tomorrow.
The real policy failure would be to allow water to remain an unpriced environmental externality of India’s digital economy.
This is where the idea of water credits deserves serious consideration, but with an important caveat. Water credits cannot simply become carbon credits with a different label.
Carbon is substantially global, a tonne of carbon avoided in one geography contributes to a global climate objective. Water is intensely local. A million litres replenished in a water-abundant catchment cannot automatically compensate for a million litres extracted from a stressed aquifer elsewhere. Water has geography, seasonality, quality and, above all, competing human claims.
A credible Indian framework should therefore follow a hierarchy. Avoid, reduce, reuse, replenish and only then compensate.
Large AI and data-centre facilities should first disclose their water withdrawals, actual consumption, sources, recycled-water share, seasonal usage and Water Usage Effectiveness. Facilities in water-stressed areas should face stricter requirements to use treated wastewater, closed-loop systems and water-efficient cooling. Environmental approvals for major data centres should incorporate basin-level water assessments rather than examining electricity and land in isolation.
Only the residual, unavoidable footprint should become eligible for compensation through verified water credits.
Such credits could finance groundwater recharge, wetland and watershed restoration, municipal leakage reduction, rainwater harvesting, wastewater treatment and reuse, and improvements in irrigation efficiency. Importantly, methodologies need not begin from scratch. The World Resources Institute’s Volumetric Water Benefit Accounting already provides a science-based approach for quantifying water-stewardship benefits at catchment level.
India could go further by introducing a Water Stress Multiplier. A litre consumed in a severely stressed basin should create a greater replenishment obligation than a litre consumed in a water-secure one. Credits generated within the affected watershed should receive priority, while distant projects should not be allowed to conveniently erase local ecological damage.
But quantity alone is insufficient. A credible credit must measure additionality, water quality, ecological restoration and community benefit. Otherwise, the system risks becoming another sophisticated mechanism for greenwashing, where companies purchase certificates while communities continue watching groundwater tables fall.
This social dimension is indispensable. Our research on AI’s environmental repercussions similarly finds that environmental burdens are unevenly distributed and may fall disproportionately on vulnerable communities, making aggregate measures of energy or water use inadequate measures of sustainability. The same study identifies governance and transparency gaps as central obstacles to environmentally responsible AI.
The objective, therefore, should not be to make AI expensive for India. It should be to make resource efficiency part of India’s competitive advantage.
The countries that dominate the next technological era will not merely be those with the most GPUs or largest data centres. They may be those capable of building digital infrastructure without destabilising the physical systems—water, energy and ecosystems—on which their economies ultimately depend.
India should certainly pursue its AI dream. But if the cloud is going to drink from India’s rivers and aquifers, it must also learn to give something back. Water credits, designed carefully, could turn an invisible environmental liability into measurable corporate responsibility. The principle should be simple. Innovate boldly, consume efficiently, replenish locally, and never allow the right to compute to overwhelm the right to water.
Pradeep S. Mehta is the secretary-general of CUTS International, a leading global public policy research and advocacy group. Sohom Banerjee is associated with Jaipuria Institute of Management, Noida..
This article can also be viewed at:
https://economictimes.indiatimes.com/