India's Data Centre Boom Is Testing Its Water Security

India’s artificial intelligence ambitions are usually discussed in terms of computing power, semiconductor capacity, cloud infrastructure, investment, and digital sovereignty. But beneath almost every AI model, cloud service, digital payment, streaming platform, and government database sits something much more physical: a data centre that requires electricity, land, cooling infrastructure, and often large quantities of water.

That physical footprint is expanding rapidly.

India’s installed data centre capacity increased from about 375 MW in 2020 to around 1,500 MW by 2025. The government now estimates that electricity demand from data centres could reach 13.56 GW by FY 2031 to 32. Strictly speaking, the often quoted 36 times figure compares 2020 installed capacity with projected future electricity demand, so it should not be treated as a precise capacity growth forecast. But the broader message is unmistakable. India is preparing for a massive expansion of digital infrastructure.

The difficult question is where this infrastructure is being built.

One of India’s most important emerging data centre clusters is Gautam Buddha Nagar in Uttar Pradesh. The district already has at least 17 operational or upcoming facilities. At the same time, official groundwater assessment places its stage of groundwater extraction at 104.79 percent. In simple terms, annual extraction is already greater than the groundwater considered sustainably available.

India may therefore be building part of its AI future on top of aquifers that are already under serious pressure.

For UPSC aspirants, this makes the data centre boom much more than a Science and Technology story. It is also a question of environmental governance, industrial policy, agricultural water security, federalism, and the economics of scarce natural resources.

Why Data Centres Need So Much Water

A data centre is essentially a large facility containing thousands of servers operating continuously. These servers generate substantial heat, and that heat must be removed to prevent equipment failure.

Cooling is therefore central to data centre operations.

Some facilities rely heavily on air based cooling. Others use evaporative cooling, chilled water systems, cooling towers, direct liquid cooling, immersion cooling, or combinations of these technologies.

Water consumption can occur directly when water is used at the facility for cooling. It can also occur indirectly because electricity generation itself has a water footprint.

This distinction matters. A data centre claiming relatively low water consumption inside its campus may still have a much larger overall water footprint if the electricity supplying it comes from water intensive power generation.

The scale of the global challenge is becoming clearer. United Nations University researchers estimate that the water footprint associated with global data centre electricity consumption was approximately 4.5 trillion litres in 2025. By 2030, it could rise to around 9.3 trillion litres.

The same research warns against judging digital infrastructure through carbon emissions alone. An energy source that performs well on carbon may perform poorly on water or land use. Sustainable AI therefore requires simultaneous evaluation of electricity, emissions, water, and land.

This is particularly important for India because many major technology clusters are located in regions where water availability is already contested.

Gautam Buddha Nagar Shows the Policy Conflict

Gautam Buddha Nagar offers perhaps the clearest example of the problem.

The district includes Noida and Greater Noida, making it attractive for data centres because of proximity to Delhi, strong fibre connectivity, industrial infrastructure, real estate availability, and access to a large digital market.

But its groundwater position creates a serious contradiction.

The Dynamic Ground Water Resources of India assessment records the district’s stage of groundwater extraction at 104.79 percent. The same official table shows extraction occurring across irrigation, industrial, and domestic uses.

Once extraction exceeds annual extractable resources, the policy question changes.

The debate is no longer simply about whether a new industry has access to water. It becomes a question of who receives priority when a scarce common resource is already under stress.

Should additional water support digital infrastructure?

Should it remain available for agriculture?

Should household consumption receive absolute priority?

Should an industrial investor be permitted to extract groundwater if nearby farmers are already facing falling water tables?

These are fundamentally governance questions, not merely engineering problems.

Uttar Pradesh Wants to Become a Data Centre Hub

Uttar Pradesh has strong economic reasons to attract the industry.

The state formally introduced its Data Centre Policy 2026 in July 2026. In the discussions preceding the policy, the Chief Minister called for increasing the state’s data centre capacity to 2 GW by 2030 and emphasised greater support for AI enabled facilities.

The economic logic is understandable.

Data centres attract large capital investment, strengthen digital infrastructure, support cloud computing, improve domestic data storage capacity, enable AI development, and create demand for associated sectors such as telecommunications, power infrastructure, cybersecurity, construction, and technical services.

Uttar Pradesh had already attracted proposed data centre investment exceeding ₹21,000 crore and proposed capacity of more than 600 MW by June 2025.

The state therefore has an opportunity to become one of India’s major digital infrastructure hubs.

But the success of that strategy cannot be measured only in megawatts, investment proposals, and server racks.

It must also be measured in litres.

The National Water Policy Creates an Important Test

India already possesses a policy framework that can help answer the question of competing water claims.

The National Water Policy recognises water as a scarce natural resource fundamental to life, livelihood, food security, and sustainable development. It identifies safe water for drinking and sanitation as a preemptive need and gives high priority to other domestic requirements, food security, livelihood dependent agriculture, and minimum ecosystem requirements. Water remaining after these needs should be allocated in a manner that encourages conservation and efficient use.

That principle has major implications for data centre planning.

If an industrial facility is proposed in a region where groundwater is already over exploited, approving it cannot simply be treated as a routine industrial clearance.

The hydrological carrying capacity of the region should become part of the investment decision itself.

This is where industrial policy and water policy risk moving in opposite directions. One arm of government may offer incentives to accelerate data centre investment while another is attempting to reduce groundwater extraction.

Good governance requires these policies to converge.

Data Centres Versus Agricultural Water Security

The conflict becomes even sharper in an agricultural state such as Uttar Pradesh.

Groundwater is central to Indian food production. The national groundwater assessment notes that groundwater supplies roughly 62 percent of irrigation requirements, along with a major share of rural and urban water supply.

This means groundwater depletion is not only an environmental problem. It is directly linked to food security, farm incomes, rural employment, and vulnerability to drought.

Consider the economics from a farmer’s perspective.

A falling water table requires deeper borewells and greater pumping. That increases the cost of cultivation. Farmers with sufficient capital may adapt by investing in deeper infrastructure, while smaller farmers may lose reliable access altogether.

Industrial water demand can therefore create distributional consequences even when a company technically complies with extraction regulations.

The issue is not that agriculture should receive unlimited groundwater while industry receives none. Agriculture itself requires major improvements in water efficiency.

The deeper question is whether governments should encourage water intensive industrial clustering in areas where aquifers are already unable to replenish at the rate at which water is being withdrawn.

The Hidden Externality in India's AI Economy

Data centres illustrate a classic economic problem known as a negative externality.

A company may calculate the private cost of land, electricity, servers, construction, cooling equipment, and connectivity before deciding where to build.

But the social cost of groundwater depletion may not be fully reflected in those calculations.

If extraction reduces water availability for surrounding communities or raises irrigation costs for farmers, part of the real cost of the project is being borne by people outside the company.

That can distort investment decisions.

A location may appear economically attractive because land is available and state incentives are generous, while the environmental cost of its water consumption remains largely invisible.

In effect, society may subsidise digital infrastructure through depleted aquifers.

This is why water pricing and disclosure matter. When a scarce natural resource is effectively treated as cheap or abundant, companies have weaker financial incentives to minimise its use.

The Solution Is Not to Stop Building Data Centres

India cannot respond to the water challenge by simply rejecting data centre development.

Digital infrastructure has become strategically important.

India requires domestic computing capacity for artificial intelligence, financial services, government platforms, cybersecurity, research, cloud services, and data sovereignty. Dependence on foreign computing infrastructure carries its own economic and strategic risks.

The more useful question is how India can build data centres without transferring an excessive environmental burden to already stressed communities.

Technology offers part of the answer.

The government notes that the industry is increasingly adopting direct liquid cooling, adiabatic cooling, immersion cooling, and high density computing infrastructure to reduce power and water requirements.

But technological efficiency alone is insufficient.

A highly efficient data centre placed over an over exploited aquifer may still be located in the wrong place.

India Needs Water Sensitive Data Centre Siting

Location should become a central element of data centre regulation.

Before approving a major facility, authorities should examine the groundwater status of the district, long term recharge patterns, competing agricultural and domestic requirements, climate risks, and availability of treated wastewater.

Areas classified as over exploited should face much stricter conditions.

Data centres in such regions could be required to rely primarily on treated sewage water, recycled water, closed loop systems, or technologies designed to minimise freshwater consumption.

Water consumption should also be publicly disclosed.

Operators should report total water use, peak seasonal demand, source of water, proportion of groundwater used, proportion of reclaimed water used, and their water use efficiency.

Without such information, communities and regulators cannot properly evaluate cumulative environmental impacts.

The Energy and Water Tradeoff Must Be Managed Together

There is another complication.

Reducing water consumption can sometimes increase electricity consumption.

Dry cooling systems, for example, can significantly reduce water use but may require more energy under certain climatic conditions. Similarly, an energy strategy designed only to lower carbon emissions can increase water consumption depending on the technology used.

The United Nations University research describes precisely this problem. Carbon, water, and land footprints do not always move together. A solution that performs well on one environmental indicator can perform poorly on another.

India therefore needs integrated planning.

A data centre should not receive separate assessments for electricity, water, and land as though they were unrelated inputs. The entire infrastructure footprint should be evaluated together.

That is especially important as AI workloads become more intensive.

From Ease of Doing Business to Sustainable Doing Business

India’s competition to attract data centres resembles earlier competition for factories, semiconductor plants, logistics parks, and renewable energy projects.

States offer land, power infrastructure, fiscal incentives, and faster approvals to secure investment.

But data centres create a new governance test because their environmental footprint can remain largely invisible to ordinary citizens.

A smokestack makes pollution visible. A rapidly declining aquifer does not.

The next generation of industrial policy therefore needs to move beyond ease of doing business.

It should incorporate sustainable doing business.

States should compete not only on how quickly they can approve data centres, but also on how efficiently those facilities use water, how much reclaimed water they consume, how much renewable electricity they source, and whether local communities share in the economic benefits.

Conclusion

India needs data centres.

It also needs groundwater.

The challenge is that both demands are expanding at the same time.

India’s installed data centre capacity has already grown four times between 2020 and 2025, while future electricity demand from the sector could reach 13.56 GW by FY 2031 to 32. Uttar Pradesh is positioning itself as a major destination for this expansion. Yet Gautam Buddha Nagar, one of its most important data centre clusters, already records groundwater extraction at 104.79 percent.

That collision captures the central dilemma of India’s digital transition.

Artificial intelligence may appear weightless on a screen, but its infrastructure is deeply physical. It occupies land, consumes electricity, produces heat, and places demands on water systems.

The policy question is therefore not whether India should pursue AI infrastructure. It should.

The question is whether India can build that infrastructure without making farmers, households, and future generations pay an invisible water bill.

If data centre policy continues to count only investment and computing capacity, India risks solving its digital infrastructure deficit by deepening its ecological deficit.

The smarter path is to make water security part of digital infrastructure planning from the beginning.

India’s AI future should run on powerful servers, but it should not run its aquifers dry.