Cool by design: How India’s data centres are learning to run on far less water

The first thing you notice when you enter the main computing hall of a major modern-day data centre is not the technology, but rather the sound, a loud, constant rumble of moving air. The hall happens to be long, neat rows of metal cabinets called racks, each of which has servers, specialised computers that store and analyse data, stacked floor to ceiling. Tens of thousands of these machines can be housed in a single large facility. Like a strong electric heater left at full power all the time, each one of them produces heat continuously. A commercial kitchen’s amount of heat can be produced per square metre by a dense row of server racks. When you multiply that by thousands of racks operating continuously throughout the year, you have a massive and endless heat problem.

As a result, cooling is not a secondary consideration when designing data centres. It is one of the most vital support systems, after electricity and connectivity. Servers run the risk of overheating, shutting down, and suffering irreversible damage if cooling fails, even momentarily. If the cooling system isn’t working, a data centre is just an expensive, dangerously hot structure. The fact that the cooling plant, the apparatus that extracts and gets rid of all that heat, is frequently located physically distant from the computing hall, either on the roof, in a courtyard, or in a special plant room close by, is something that many people fail to recognise at first glance. Water enters the scene in a significant manner in this plant, which is almost invisible to observers.

Visit hilogame.news for more information.

From hot aisles to chilled water

One fundamental engineering guideline determines the organisation of the computing hall: hot and cold air must never be permitted to mix. Racks are arranged so that their back faces, where hot air escapes, face a common hot corridor, and their front faces, where cool air enters, face a common cold corridor. This configuration is known as the hot and cold aisle system, and it is used in almost every large facility that is operating today.

Computer Room Air Handlers, or CRAHs, are the devices that supply cooled air to the cold aisles. They are big devices with coils that continuously carry chilled water. Warm room air is drawn over these chilly coils by a CRAH, which then returns the cooled air to the server aisles. Computer Room Air Conditioning units, or CRACs, are used in older facilities. They are often less energy efficient and run on a refrigerant cycle instead of chilled water. In both cases, cooling air and distributing it to the racks operate on the same concept.

The two loops

Chillers are industrial refrigeration devices that operate on the same fundamental basis as a home refrigerator, but on a much bigger size. They are responsible for keeping that chilled water cold. Warm water that comes back from the CRAHs is sent out cold after a chiller absorbs its heat. The heat that has been removed needs to go somewhere, so it is put into a second, independent water loop that transports it outside to the cooling towers.

In reality, a typical large data centre has two separate water circuits. The first moves cold water between the chillers and the CRAHs within. The second moves heated water outside between the cooling towers and the chillers. The cooling towers in this outdoor loop is where actual water consumption, or irreversible and permanent water loss, occurs.

Feeding the system: Where the water comes from

A specific amount of the water going through the cooling circuit is permanently lost during each operational cycle, thus, it must be regularly refilled from an external source. This replacement water, also known as make-up water, usually comes from groundwater, a municipal supply, or increasingly treated recycled sources such as collected rainwater or processed municipal wastewater.

The water is treated before it enters the cooling system. To stop scale from forming on pipes and equipment, particles are filtered away, hardness-causing minerals like calcium and magnesium are decreased, and a small quantity of biocide is injected to stop bacteria from colonising the warm, humid environment of the cooling towers. The method is generally similar to that of treating a huge swimming pool: frequent chemical dosing, continued testing, and thorough monitoring, but it adheres to significantly tighter industrial and environmental requirements.

The tower that drinks heat

The physics of evaporation

Fundamentally, a cooling tower is a larger version of the desert cooler that millions of Indian homes rely on each summer. In a desert cooler, hot, dry outside air is drawn through pads soaked in water, when the water evaporates, it absorbs heat from the flowing air and cools it. According to Mytton, a data centre cooling tower operates on the same principle, but on an industrial level.

Hot water from the chillers is dispersed across corrugated plastic or wooden fill material inside the tower, which results in a large, continuously moist surface area. Outside air is drawn upward through this moist surface by strong fans. A portion of the water evaporates, and this evaporation safely removes heat from the remaining water. The physics explaining this is known as the latent heat of vaporisation. Evaporating one kilogram of water requires about 2,260 kilojoules of energy. That’s a lot more energy than just heating the same water a few degrees. This is also the reason why, on a hot afternoon in Chennai or Delhi, sweat evaporating from your skin effectively cools your body. In order to absorb additional heat and resume the cycle, the cooled water that does not evaporate is collected in a basin at the base of the tower and is carried back to the chillers.

Withdrawal vs Consumption

Properly interpreting data centre results requires an understanding of the distinction between water withdrawal and consumption. Withdrawal refers to the total volume collected from the supply source. The portion that is lost forever, mostly by evaporation into the atmosphere, is called consumption. About 70 to 80 percent of water loss in an open evaporative cooling system occurs through evaporation, with the remaining water getting out through a regulated process known as blow down. Only a little portion of the water flowing through the tower evaporates at each pass, but over time, that tiny portion grows into a significant amount due to the system’s constant operation and huge volumes.

What the water carries away

The blow down process

The dissolved minerals and salts that are left behind when water evaporates from the tower basin become increasingly concentrated in the water that remains. This is similar to the white scale that builds up within a vessel after boiling water repeatedly and allowing steam to escape. Minerals precipitate out of solution and create hard deposits on pipes, heat exchangers, and fill material, if this concentration is allowed to increase unchecked, harming the machinery and lowering productivity. Operators periodically remove a certain amount of the concentrated basin water and replace it with a clean, treated supply in order to avoid this. This method of controlled drainage is known as blow down.

Drift, discharge, and regulation

Blow down does not represent raw wastewater. Although it is a controlled, quantifiable industrial wastewater stream that is usually sent to municipal wastewater treatment systems or treated on site prior to any environmental discharge in full compliance with regulatory standards, it does contain elevated levels of dissolved minerals and residual treatment chemicals. Drift is a separate and much smaller loss that happens when tiny water droplets are transported out of the tower by moving air. Drift eliminators, which are baffled panels installed in modern towers, catch these droplets before they escape, lowering drift loss to less than 0.01 percent of the recirculation rate. It is important to note that the evaporated water itself is quite pure because it has left all of its dissolved minerals in the basin. It then disperses into the atmosphere as water vapour and finally returns as rains somewhere in the larger water cycle. 

Data centres are measured by water use because cooling is their biggest water demand, and that demand rises with the amount of power they consume. A rough benchmark is around 25 million litres per MW per year for water cooled facilities, though the exact figure changes with the cooling system and location. At India’s current scale, that adds up to about 150 billion litres a year, with projections rising sharply by 2030.  

Beyond the tower: The new science of keeping servers cool

So far, this article has told the story of the cooling tower, that tall, industrial version of your summer desert cooler that evaporates water to carry heat away. For decades, this was the dominant answer to the data centre’s heat problem. It worked. It was reliable. It was cost-effective. And for the modest server rooms of the early internet era, it was more than enough.

But something has changed dramatically in the last five years, and it has changed almost everything about how the industry thinks about cooling. The change has a name, artificial intelligence.

A modern AI training chip, the kind used to run the large language models reshaping how we work and communicate, can consume 700 watts of power on its own. A single rack of these chips, packed together in a dense cluster, can consume 100 kilowatts or more. To put that in perspective: a typical household in India runs on roughly 1 kilowatt. A single AI server rack, in other words, consumes about as much power as a hundred Indian homes, all concentrated into a space the size of a tall refrigerator. Air, even fast-moving, precisely directed air, simply cannot carry that much heat away fast enough. The physics do not allow it. This is why the data centre industry is undergoing its most significant engineering revolution since the invention of the cooling tower. And most of the new solutions use far less water, or none at all.

Direct liquid cooling: Bringing cold to the chip

Direct Liquid Cooling, or DLC, is taking over data centres right now, and for good reason. The name pretty much gives it away. Instead of cranking down the thermostat for an entire room and hoping enough cold air reaches the processors, DLC delivers cooling liquid straight to where things actually get hot.

It’s quite a simple setup. A thin metal pad filled with micro-channels, called a cold plate, gets mounted directly onto the CPU or GPU. Cold water flows through those small pathways, absorbs heat right at the surface, and carries it away inside a closed loop of pipes. Water pulls heat off metal way faster than air ever could. And because the water stays trapped inside that closed loop, the same fluid keeps circulating over and over, meaning virtually zero water evaporates. This is exactly how high-end liquid-cooled gaming PC’s work.

Hardware giants like Lenovo, HP, and Dell are already shipping DLC compatible systems as a standard option. Meanwhile, the companies running the world’s biggest data centres are retrofitting entire server halls just to keep up with heavy AI workloads. The result? Hardware stays cooler, energy waste plummets, and water usage drops to a tiny sliver of what traditional cooling setups burn through.

Immersion cooling: Drowning the servers on purpose

If Direct Liquid Cooling sounds radical, immersion cooling sounds confusing until you understand the physics. In an immersion-cooled data centre, servers are removed from their racks and submerged entirely in a tank of fluid, not water, but a specially engineered, electrically non-conductive liquid that covers every component of the server without causing a short circuit. Think of it like dropping your laptop into a special fluid that does not conduct electricity. The fluid absorbs heat directly from every chip, every circuit board, and every component simultaneously, without a single fan running anywhere

There are two types. Single phase immersion uses a fluid that stays liquid throughout, it absorbs heat and is then pumped through an external heat exchanger to cool down before returning to the tank. Two phase immersion is more inventive, the fluid is chosen so that it boils at a low temperature, around 50 degrees Celsius, right at the surface of the chips. As it boils, it turns to vapour, rises, condenses on cool coils placed at the top of the tank, and drips back down as liquid in a continuous, self-circulating loop that requires almost no energy to pump. The vapour you see rising from such a system is not water. It is a specialised engineered fluid that is expensive than water but never evaporates into the environment and is captured and reused entirely.

Companies such as LiquidStack and Submer are already operating commercial immersion-cooled data centres. Microsoft has run pilot immersion-cooled facilities. The water consumption of an immersion-cooled centre is, for practical purposes, close to zero, the fluid in the tank is a closed system.

Free cooling: When nature does the work free

There is a third approach that requires neither complicated fluids nor elaborate engineering, it simply requires choosing the right location and the right season. Free cooling, also called air-side economisation, works on a principle so simple it is almost embarrassing. When the outside air is already cooler than the servers need to be, you bring the outside air in and use it directly, bypassing the chiller plant entirely.

Microsoft has used free cooling extensively in its data centres in Ireland, Finland, and the Pacific Northwest of the United States, where the climate is cool enough for large portions of the year to make this viable. In those conditions, the cooling system uses zero water and a fraction of the electricity of a conventional evaporative system. In India’s hot climate, free cooling alone is insufficient for most of the year, but hybrid systems that switch between free cooling in winter months and conventional cooling in summer can still reduce annual water consumption meaningfully.

Trivia: Sinking a data centre in the sea to cooling servers in space

Back in 2018, Microsoft pulled off something straight out of a sci-fi movie. They crammed 864 servers into a massive steel tube about the size of a shipping container, pumped it full of dry nitrogen, and dropped it 117 feet down to the bottom of the Atlantic Ocean, just off the coast of Scotland. They called it Project Natick. It sat down there for two whole years, powered by local wind and tidal energy, while the freezing ocean water naturally kept everything cool.

When they finally hauled it back up in 2020, the results blew everyone away. The servers in that underwater tube broke down eight times less often than the exact same setups on land. Why? It turns out that a perfectly stable environment, with absolutely no humans bumping into things, no oxygen to cause rust, and zero humidity changes, keeps electronics incredibly happy. Microsoft hasn’t started selling underwater data centres just yet, but the experiment totally changed how the tech world thinks about cooling. The setup used zero fresh water and had zero evaporation. Just the cold, deep ocean doing its thing.

Now, if dropping servers into the ocean sounds a bit out there, wait until you hear what some aerospace startups are planning next. Putting data centres in orbit. Space startups such as SpaceX, along with researchers all over the US and Europe, are actively working on it. It sounds wildly ambitious, but the core logic makes perfect sense. Up in low Earth orbit, a computing satellite gets endless solar power on one side, and the freezing cold of deep space on the other.

To cool things down, you don’t need water, fans, or massive cooling towers. You just point thermal panels away from the sun and let the heat radiate out into the void. They’re already testing early versions of this right now, mostly to process satellite images directly in orbit instead of beaming all that heavy raw data back down to Earth. We might not see massive, floating space servers become the norm this decade, but the trend is obvious. The industry is moving fast toward cooling solutions that use the natural extremes of our environment instead of wasting precious fresh water here on the ground.

Why this matters especially for India

India’s push to build massive data centres isn’t just some future plan anymore, it’s happening right now, and it’s moving fast. A lot of this growth is taking place in cities that are already running low on water. Because of this, picking the right cooling system isn’t just a basic tech choice. It actually has a huge impact on the local water supply, how much the building costs to run, and whether the industry can stay green in the long run.

We’re already seeing the rules change to match this. Take Rajasthan, for example. Their 2025 data centre guidelines push companies to use recycled water and make sure no liquid goes to waste. At the same time, big tech companies are making Direct Liquid Cooling a strict rule for any new setups in India. Plus, as AI needs more and more power to run, liquid cooling isn’t just a nice option anymore, it’s becoming an absolute must. Now, regular cooling towers aren’t going to vanish overnight. They’re still cheaper and work just fine for normal, everyday computer tasks. But the next generation of data centres is going to look completely different from what is being built today.

For India, it’s a tough mix with extreme heat, a serious lack of water, and a booming AI industry. This creates a huge challenge, but it also gives everyone a really strong reason to switch to low-water cooling tech faster than ever.

The Water Usage Effectiveness metric

The data centre industry compares cooling water efficiency among facilities using a common metric known as Water Usage Effectiveness, or WUE. WUE, which is measured in litres of water per kilowatt hour, is simply the entire volume of water consumed divided by the energy required by the IT equipment inside. An operation that uses less water is indicated by a lower WUE. WUE values of 1.0 to 2.0 litres per kilowatt hour or even better can be achieved in well designed modern facilities, especially those that use hybrid cooling with intelligent controls. Facilities that are older or poorly maintained could score 3.0 or higher. 

The deeper engineering insight is that dry cooling uses more electricity than evaporative cooling. Compared to air-only or refrigerant-only solutions, it uses less electrical power to remove a specific quantity of heat. Therefore, selecting water-based cooling may actually lessen rather than raise a facility’s overall impact on the environment in areas where electricity is produced from coal or other carbon-intensive sources. This water-energy trade-off is what truly complicates cooling design and makes broad generalisations regarding data centre resource usage incorrect.

Warm, humid outdoor air lowers the driving force for evaporation and makes cooling towers work harder in India’s peculiar climate, which includes lengthy summers, frequently occurring temperatures above 40 degrees Celsius, and high humidity in coastal areas. For Indian operators looking to lessen their reliance on freshwater without sacrificing cooling dependability, careful tower design, higher cycles of concentration, and the utilisation of recycled or other water sources are particularly crucial.

 Conclusion

Water is used to cool data centres in large volumes. However, the method by which they employ it is a completely quantitative, well-understood engineering system with traceable, quantifiable losses rather than an enigmatic or uncontrollable depletion of a region’s water resources. The same physical process that has cooled the human body since the dawn of time causes the majority of water absorbed to evaporate into the atmosphere. A lesser amount is discharged as managed and regulated wastewater. Drift allows only a small fraction to escape. With the help of precise tools provided by modern engineering, such as WUE monitoring, hybrid cooling designs, and recycled water integration, operators can simultaneously lower energy and water consumption. Not all of the questions regarding data centres and water can be answered by knowing this physical reality. However, it gives such questions the solid, clear foundation they deserve. The next article in this series will analyse the main narrative and then debunk the false claims.  

Read full story at source