AI Destroying Our Water? Inside India’s Hidden Water Crisis

The world is racing to build AI. India is racing to become its next global hub. But behind every chatbot response, every image generated, every model trained, lies a resource most people never think about: fresh water. And in India, the collision of artificial intelligence and water scarcity is creating a silent emergency.

The Invisible Thirst of Artificial Intelligence

When you ask an AI a question, you probably think about electricity. You might even think about carbon emissions. Almost nobody thinks about water.

But data centers — the industrial nerve centers that power every AI model — are among the thirstiest buildings on earth. They run on servers that generate enormous heat. To prevent those servers from melting down, operators pump millions of liters of water through cooling towers, where it evaporates into the air, carrying the heat away.

That water does not come back.

The numbers are staggering. Microsoft reported that its global water consumption jumped approximately 34% between 2021 and 2022, a spike the company directly linked to the explosive growth of AI workloads. Google reported a similar 22% increase over the same period. Training a single large language model — the type that powers modern AI assistants — is estimated to consume around 700,000 liters of fresh water. That is enough drinking water for one person for roughly 1,900 years.

And that is just training. Every time a user runs a query — every inference, every generated image, every real-time translation — water is consumed again, at scale, millions of times per day.

Why India Is Ground Zero for This Crisis

India is not just participating in the global AI boom. It is at the center of it.

In 2024, the Indian government launched the IndiaAI Mission, a sweeping national initiative to position the country as a leading AI power. The mission is accelerating investment in data centers, funding AI infrastructure, and attracting global technology giants to build on Indian soil. By most projections, India’s data center capacity will more than double before 2030.

This is happening against a backdrop that makes it uniquely dangerous.

India is already ranked among the most water-stressed nations in the world. According to the NITI Aayog — the government’s own policy think tank — India hosts approximately 18% of the world’s population but holds only about 4% of the world’s freshwater resources. Groundwater aquifers, particularly in Maharashtra, Tamil Nadu, and Rajasthan, are being depleted at rates that far outstrip natural recharge. Several of India’s most important agricultural belts are already drawing on reserves that took thousands of years to accumulate.

The collision of these two forces — one pushing water demand up, the other pushing water supply down — is creating conditions for a crisis that has received almost no serious public attention.

The Geography Problem: Where Data Centers Meet Drought

India’s data center industry is heavily concentrated in a handful of cities. Mumbai, Chennai, Hyderabad, and Pune together host the vast majority of the country’s existing and planned data center capacity. These cities were chosen for practical reasons: they have reliable power grids, strong fiber connectivity, skilled technical workforces, and good logistics.

What they also have is water stress.

Mumbai, India’s primary data center hub, draws its water from a system of lakes and reservoirs that regularly run low during the pre-monsoon months. The city’s own residents frequently face water rationing. Yet the data center belt expanding along the Mumbai-Pune corridor is placing additional demands on the same stretched supply.

Chennai came within weeks of a complete water system collapse in 2019, during what officials called “Day Zero.” Tanker trucks queued outside government offices. Residents waited hours at public taps. The crisis eventually eased when the monsoon arrived — but the structural causes remained. Since then, Chennai’s data center expansion has continued at pace, with multiple large facilities announced and under construction.

Hyderabad, promoted as “Cyberabad” and a major technology hub, sits on the Deccan Plateau, where groundwater tables are among the most depleted in southern India.

The pattern is consistent and troubling: India is building its AI infrastructure precisely in the places least able to absorb additional water demand.

The Seasonal Timing Makes It Worse

India’s water crisis is not a flat, year-round problem. It pulses with the seasons. And the seasonal pattern for water stress almost perfectly inverts the seasonal pattern for data center cooling demand.

The April through June window — the months before the monsoon arrives — is when Indian temperatures peak. In many parts of the country, temperatures regularly exceed 40°C (104°F) during this period. Data centers require more cooling when ambient temperatures are high, which means water consumption climbs precisely during those months.

Those same months are when India’s water tables are at their lowest. Reservoirs drawn down through the long dry season are at or near minimum levels. Groundwater, pumped throughout the year, has had no chance to recharge. Urban water systems are under maximum stress.

The result is a direct seasonal competition between data centers and communities for the same diminishing water supply — at exactly the moment when that supply is most fragile.

The Transparency Gap: Nobody Is Counting

In the European Union, regulations are tightening around data center environmental disclosure. Companies operating at scale are increasingly required to report water usage, energy efficiency, and environmental impact. The data, imperfect as it is, at least exists.

In India, no such requirement exists.

There is currently no mandatory water reporting framework for data centers operating in India. No government agency publishes comprehensive figures on how much water the sector consumes. No environmental impact assessment routinely captures water use alongside land use or power consumption.

This means the actual scale of AI-related water consumption in India is largely unknown. Researchers and journalists must work backwards from power draw figures, facility types, and cooling technology specifications — inferring water use rather than measuring it. The estimates exist, but they are estimates.

This transparency gap is not a minor technical detail. It means India cannot accurately assess the impact of its AI infrastructure on local water systems. It means communities near data centers have no way to know how much water those facilities are drawing. It means policymakers are making decisions about where to approve new data centers without reliable data on water trade-offs.

For a country already facing a structural water emergency, this is a serious governance failure.

How Data Centers Actually Use Water: Three Channels

Understanding the water footprint of AI requires looking beyond the obvious cooling towers. The water consumption flows through three distinct channels.

Direct evaporative cooling is the most visible. In standard cooling tower systems, water is pumped across hot surfaces and allowed to evaporate. This phase change carries heat away efficiently — but the evaporated water is gone from the local water cycle. A large data center running evaporative cooling can consume tens of millions of liters per year.

Upstream power generation adds another layer. India’s electricity grid is still heavily dependent on thermal power plants — coal, natural gas, and some nuclear. These plants also use water for cooling. Every unit of electricity consumed by a data center carries an embedded water cost from the power plant that generated it. This is sometimes called the “water-energy nexus,” and it means data centers’ true water footprint is significantly larger than their direct cooling consumption alone.

Hardware manufacturing represents a third, less-discussed channel. Semiconductor fabrication — the process of making the chips that power AI — is extraordinarily water-intensive. A modern chip fabrication plant can consume millions of liters of ultrapure water per day. While most chip manufacturing happens outside India, the country’s ambition to build domestic semiconductor capacity brings this issue closer to home.

What the Tech Industry Promises — and What It Delivers

To be fair, the major technology companies are not ignoring this problem.

Microsoft, Google, and Meta have all made high-profile pledges to become “water positive” by 2030 — meaning they commit to replenishing more water than they consume globally. The pledges involve investing in watershed restoration projects, funding water access infrastructure in water-stressed communities, and developing more efficient cooling technologies.

Some progress is real. Newer data center designs increasingly use air cooling or direct liquid cooling to chips, which can dramatically reduce water consumption compared to traditional evaporative towers. In cooler climates, data centers can use outside air for cooling entirely, using almost no water at all. Facilities built in more temperate regions — parts of Europe, northern North America, Scandinavia — are significantly more water-efficient than those in hot climates like India’s.

But several problems limit the impact of these pledges in the Indian context.

First, “water positivity” is calculated globally. A company can consume water in Chennai while claiming credit for restoring a watershed in Oregon. Local communities in India facing data center water competition receive no direct benefit from distant conservation projects.

Second, efficiency is racing against growth — and growth is winning. Even as water per computation drops, the total number of computations is increasing faster. The net result, as seen in Microsoft and Google’s own reported consumption figures, is rising absolute water use year over year.

Third, India has no mechanism to hold international companies to these pledges on Indian soil. Without mandatory disclosure or enforceable standards, voluntary commitments remain voluntary.

India is not without options. Several practical steps could substantially reduce the risk:

Mandatory water disclosure is the starting point. Any data center above a certain capacity threshold should be required to report annual water consumption to a designated regulatory authority. This data should be public. Without measurement, management is impossible.

Water-stressed zone restrictions could prevent the worst outcomes. Regulators should map groundwater stress levels against planned data center locations and apply stricter approval criteria — or outright moratoria — in the most depleted zones. Incentives to locate new facilities in less water-stressed areas, such as parts of northern or eastern India, could help shift the geography of infrastructure development.

Cooling technology standards could set a floor. Requiring new data centers above a certain scale to demonstrate water efficiency ratios — and to prefer air cooling, closed-loop systems, or recycled water sources — would push the industry toward better practice.

Community water impact assessments should be mandatory before large facility approvals. Local municipal water systems should be modeled, and data center water demand should be stress-tested against projected seasonal water availability.

AI’s Resource Reckoning

The water crisis is one dimension of a broader reckoning that AI’s rapid growth is forcing.

The narrative around artificial intelligence has been dominated by capability — what AI can do, how fast it is improving, what industries it will transform. The resource costs have received far less attention, partly because they are harder to see and partly because the industry has had limited incentive to highlight them.

But those costs are real and growing. Electricity demand from data centers is already straining power grids in multiple countries. Water consumption is rising. Land use for large-scale facilities competes with agriculture and housing. The carbon footprint of training and running large models remains substantial despite efficiency improvements.

India’s specific situation — high water stress, rapid AI infrastructure expansion, governance gaps, and seasonal weather extremes — makes it a particularly high-stakes case. But the underlying dynamic is global.

The question facing India, and every country investing in AI infrastructure, is whether the governance frameworks can keep pace with the technology. History suggests that they rarely do — at least not without public pressure, civil society engagement, and eventually political will.

The water under India’s data centers does not have to disappear before the conversation starts. But the window to get ahead of this problem is narrowing with every new facility that breaks ground.

Key Facts at a Glance

• Training a single large AI model consumes approximately 700,000 liters of fresh water
• Microsoft’s global water consumption rose ~34% in 2022, driven by AI growth
• Google reported a ~22% increase over the same period
• India holds 4% of global freshwater but supports 18% of the world’s population
• India’s main data center hubs — Mumbai, Chennai, Hyderabad — are all in high water-stress zones
• The April–June cooling peak directly coincides with India’s lowest seasonal water tables
• India currently has no mandatory water reporting requirements for data centers
• The EU is moving toward mandatory environmental disclosure; India has no equivalent legislation

sanjoy gorh

Sanjoy Gorh – Founder & Editor, FinBuzz India
Sanjoy Gorh is the founder and editor of FinBuzz India (finbuzzindia.com), an independent digital news platform delivering accurate, clear, and timely news to readers across Assam, Northeast India, and beyond.
Driven by a deep passion for digital journalism, Sanjoy launched FinBuzz India with a clear mission: to give grassroots stories the attention they deserve and bring local voices to a national stage. Hailing from Assam, he brings hands-on, on-ground experience in news reporting, content creation, and digital media management.
His editorial focus spans Assam local news, Northeast India developments, government schemes and exam updates, finance, technology and AI, business and startups, sports, and national affairs — always with an emphasis on making important topics simple, relevant, and accessible to everyday readers.
At the heart of his work lies an unwavering commitment to factual, unbiased reporting. Sanjoy believes journalism’s greatest responsibility is building reader trust, and every story published on FinBuzz India reflects that belief.
With a vision to grow FinBuzz India into the most trusted digital news voice of Northeast India, Sanjoy continues to raise the bar, one story at a time.
Connect with Sanjoy: [Twitter/Xhttps://x.com/amolgorh84648?s=11 ] | [https://www.linkedin.com/in/finbuzz-india-6b0a00307?utm_source=share_via&utm_content=profile&utm_medium=member_ios]