How Data Centres Sap Water from Our Lands

The idea of the ‘cloud’ storing our digital data feels weightless when we ask Gemini about our homework questions we don’t want to do, asking ChatGPT to help with speeches we need to get through, or the voice message we get when we pay for snacks through UPI. Compounding these apparently minor digital interactions across India’s behemoth of a youth population, and the scale of the infrastructure required to support them becomes difficult to ignore.

Of course, those are just a few of the common uses among the youth population. Professionals in the workplace are often encouraged to use artificial intelligence in their work for complex tasks, requiring even more energy for the system to complete the task. That’s where data centres come in- the physical, hulking infrastructure supporting it.

A data center—often appearing like a large, windowless warehouse—is a large facility that stores, processes, and distributes digital information. Modern data centers support many digital processes like cloud computing, streaming, artificial intelligence (AI), social media, banking and healthcare.

The focus of all the headlines however, is AI data centers. These warehouses full of computer servers run on electricity. This electrical energy produces heat. To prevent both the facility and the servers from overheating and combusting on us, they need consistent cooling.

This is where water comes in. Many data centers use a process called ‘evaporative cooling’ to cool their facilities, a process in which hot air is cooled by cold water, forming steam. Fans then push this hot air out of the data center — around 80% of the water used evaporates during this process.

Beyond direct water use through cooling, AI data centers often rely on electricity generated by power plants that use steam and cold water to produce energy. The structural components for data centres like servers and chips also require water for their own manufacturing process- another use of water indirectly.

Most of the water used is lost during cooling, but the remaining water is recycled a few times to help cool the servers. However, over time, this process leaves behind salt, bacteria, and mineral deposits as water evaporates, due to their minuscule amounts in water gradually becoming more concentrated and at risk of clogging the cooling systems and corroding equipment. To stop this, data center operators must continually add new freshwater to the system and discharge over-cycled water as wastewater.

Another obvious question is why data centers can’t use saltwater. The simple answer is the ‘salt’ in ‘saltwater’, which would be left behind when the water evaporates during cooling. Similarly to over-cycled freshwater, this can cause clogs and corrode equipment; thus, it is not used.

Since the main water waste occurs from cooling, AI data centres have been introducing new types of cooling in order to decrease the amount of water used in the process. Power Usage Effectiveness (PUE) is defined as the total data-centre facility energy divided by the IT equipment energy. A lower PUE corresponds to less energy waste on overhead systems and more energy going into IT equipment. Two promising cooling alternatives have emerged: Direct-to-Chip and Immersion Cooling. Direct to chip cooling circulates liquid coolant through cold plates mounted directly onto high-power processors. Immersion cooling is a technique where servers and computer hardware are completely submerged in a thermally conductive, electrically non-conductive dielectric fluid to dissipate heat efficiently. These fluids include mineral oils, synthetic hydrocarbons, esters, and fluorocarbons. Immersion cooling is slightly more efficient in pure energy usage and thermal performance, achieving a lower Power Usage Effectiveness: 1.02 to 1.10 for immersion vs. 1.05 to 1.30 for direct-to-chip (with the theoretical perfect being unity), but direct-to-chip (DTC) is more prevalent because it fits into standard rack layouts, and avoids the heavy upfront costs of specialized tanks.

India is becoming an increasingly important pillar of the global AI and digital economy. India now has more than a billion internet subscribers, while average mobile data consumption has reached around 25 GB per user each month. AI is also moving beyond being something used mainly by technology companies: more than 80% of organisations in India are reported to be using AI solutions, and the AI market is expected to grow substantially by 2027. All of this means more computing, more storage and, inevitably, more data centres. India already generates close to one-fifth of the world’s data while possessing only a small fraction of global data-centre capacity, creating pressure to build new facilities quickly. The problem is that this expansion is happening in a country where water and climate pressures are already serious. India’s data centre’s infrastructure’s environmental costs have to be considered alongside the benefits of becoming an AI and digital hub.

Looking back at an example from the introduction, we have UPI— India’s novel online payment method, used by millions. Yet to function on such a scale of digital payments, the National Payments Corporation of India (NPCI) operates its own highly resilient, dedicated Tier-IV data centers. Because UPI handles more than 640 million transactions every single day, its infrastructure requires intense computing power, inevitably leading to pulling more water.

It’s possible to reason that we’re using our own country’s water to support our economy, which is a logical conclusion to draw and we can apply similar arguments to our other technology. This argument does fall apart when we read news articles about tech giants setting up shop in India to use us as their extraction site to strip our lands of water and as a dumping ground for their carbon footprint, with an added bonus of using our population to harvest raw data under the banner of ‘accelerating innovation’. Brands like NVIDIA, Amazon Web Servics, Meta, Microsoft and Google have already set up or are planning to set up hyperscale AI data centres in water deficit areas. India’s data center water consumption is estimated at 150 billion liters and is projected to skyrocket to approximately 358 billion liters by 2030.

The geographical spread of data centres in India is also heavily concentrated, with 75% of all data centres located across just five states: Maharashtra, Tamil Nadu, Karnataka, Telangana and Uttar Pradesh, which together account for 278 facilities nationally. Many of these are situated in regions classified as having high or extremely high water stress, where demand significantly exceeds available renewable supplies.

Applications of more complex AI use (which require considerable amounts of energy) have applications on a larger, utalitarian scale and are possible, such as in optimisations in methane emissions reductions in oil and gas operations by detecting leaks quicker. Emissions can also be reduced in various sectors like industries, transport and buildings by optimisation of energy consumption and resource allocation.

However, does all of this justify building looming vampiric warehouses, sucking dry the life force from our very land?

WriteToWin Monsoon 2026