Data centres aren't India's water villain

India's water crisis has been created by decades of treating water as free. Prosecuting the newest useralone would be a policy failure

The moment data-centre water use starts appearing in cartoons in financial dailies and think pieces in business magazines, and President Donald Trump begins staking political capital on one of the most contentious issues in the United States, one knows that an engineering footnote has become a public policy dispute. 

Artificial intelligence, cloud computing and hyperscale server farms are now routinely painted as villains in a country’s water story.  A 100Mw data centre, critics warn, can draw roughly 2 million litres of water a day in conventional cooling conditions. Subsidising such facilities for a modest number of direct local jobs is increasingly described not as modern infrastructure-building, but as ecological asset-stripping. 

It is a neat, emotionally satisfying argument, yet it is also an incomplete one. Data centres do deserve scrutiny. They must disclose water withdrawals, avoid potable freshwater wherever alternatives exist, pay fairly for local resources, and be held to strict recycling and discharge norms. But singling them out as the main cause of India’s water problem is to mistake visibility for scale.  

The larger question is not whether data centres use water. They do. The question is whether India has a serious framework for comparing water use across sectors, locations and economic outcomes. Water, the country’s most valuable natural resource, is still treated as a cheap input for agriculture, industry and urban expansion rather than as a scarce economic asset that must be priced, measured, and governed. 

Consider the arithmetic of proportion. Agriculture accounts for around 90 per cent of India’s freshwater withdrawals, by most standard estimates. Within that, India continues to encourage crops and trade patterns that export enormous quantities of “virtual water”— the hidden water embedded in rice, sugar, cotton, livestock products and other commodities.  

Ashok Gulati and other agricultural economists have long warned of this distortion. India’s rice exports often exceed 20 million tonnes a year. Depending on assumptions about water intensity, that can translate into tens of billions of cubic metres of embedded water leaving the country annually, much of it drawn from the stressed groundwater belts of Punjab, Haryana and parts of Telangana. Sugarcane adds another layer of irony. The push for ethanol blending may reduce some fossil-fuel dependence, but when the feedstock is water-intensive cane or grain, the environmental burden is partly shifted from oil imports to groundwater depletion. 

This is the real paradox. As a nation we continue to subsidise the draining of our own water tables through free or under-priced power, open-ended procurement signals, and export incentives that do not account for embedded water. Rice, sugar, cotton, textiles, leather, mining, metals, steel, chemicals, pulp and paper are all part of the same water economy. Some withdraw water for irrigation. Others use it for washing, cooling, dilution, slurry transport, boilers, bleaching or effluent treatment. Yet public outrage is far more likely to gather around a data-centre campus than around the slower, dispersed and politically familiar overuse of water elsewhere. 

Of course, data centres’ local footprint can be real, especially if they draw on municipal freshwater, private tankers, or groundwater in already stressed urban regions. Their power demand can also increase indirect water use through the electricity system. But the solution is regulation, not reflexive demonisation. 

Indeed, data centres may be among the easier parts of the water problem to discipline. Modern facilities are moving towards closed-loop liquid cooling, air-cooled chillers, seawater cooling in coastal locations and reclaimed municipal wastewater. These systems are not costless; they need energy, infrastructure and thorough local safeguards. But they are measurable and auditable. A server farm can be required to report water usage effectiveness, source only non-potable water, use recycled supply, meet zero-liquid-discharge standards and publish watershed-level impact data.  

The right test, therefore, is economic productivity per litre. How much value, employment, strategic capability and ecological damage does each litre generate across sectors and places? Ronald Coase’s insight on social cost remains useful here. Externalities persist not because one actor is uniquely wicked, but because rights, prices and liabilities are poorly defined. Groundwater is extracted as if it were a private entitlement, though its depletion is a public cost. Electricity subsidies hide the cost of pumping. Industrial water is often priced below scarcity value. Urban wastewater remains under-treated even as industries queue up for fresh supply. In such a system, blaming one visible sector may produce satisfaction, but not conservation. 

India needs a water policy that is boring in design but radical in effect: Measure, price, disclose and recycle. First, all large users, including agriculture and certainly industrial clusters, mines, steel plants, chemical units, paper mills, textile processors and data centres, should face transparent water accounting. 

Second, high-stress basins should carry stricter permitting rules, with recycled water made mandatory for new industrial loads. Third, trade and subsidy policy must incorporate a virtual-water lens. It makes little sense to celebrate exports that earn thin margins while silently liquidating groundwater capital. 

Fourth, farm policy must move away from rewarding water intensity. Minimum support price signals, power subsidies and procurement systems should encourage crop choices aligned with local water endowments. Millets, pulses and oilseeds cannot remain rhetorical favourites while paddy and cane continue to enjoy the real incentives. Replacing unmetered free power for borewells with better-targeted income support would be politically difficult, but it is more honest than pretending that data-centre restrictions will rescue Punjab’s aquifers. 

The hard truth is that India’s water crisis has been created by decades of treating water as free and depletion as someone else’s problem. Data centres are part of the new demand curve and must be governed accordingly. But the real policy failure would be to prosecute the newest user while protecting the oldest distortions. 

 

The writer is group chief economist, L&T. The views are personal