Thirsty Machines

“Thousands have lived without

love, not one without water...”

— W.H. Auden

We are the new, modern world. We pride ourselves on miracles that seemingly occur without effort. Our messages travel across continents in milliseconds. Artificial intelligence drafts treatises, diagnoses disease and composes music on demand. Cloud servers remember everything we deign to forget. The digital age, we have been informed, is immaterial; it is just an elegant realm of code and computation floating somewhere in the ether. In reality, though, the cloud is anything but weightless. It is built from steel, silicon, copper and water. A hell of a lot of water.

Every day, vast warehouses filled with humming servers generate immense heat while processing the world’s digital traffic. Cooling these machines requires astonishing volumes of water. Research bodies have estimated that 800 billion litres of freshwater are used worldwide every single day to cool our computing infrastructure, industrial systems and digital networks – those machines that sustain the globe’s modern economies. Just one large data centre can consume around 18.9 million litres of water each day, roughly the amount used by a town of 50,000 people.

And the numbers are climbing rapidly as artificial intelligence flexes muscle. Analysts warn that AI-led data centre operations alone will consume hundreds of billions of litres annually within the next few years, some even suggesting that global demand will cross trillions of litres by Year 2030. If you still haven’t figured out what these words are saying, here’s an unabridged version to scare the pants off your revered bottoms: the same tech that promise to transforms civilisation is quietly sucking the planet dry. Consider the irony of asking an algorithm about global sustainability; and then imagine it evaporating water somewhere to keep itself cool while it figures out what answer to throw at us.

Heat of Intelligence

For all of its mathematical elegance, artificial intelligence has a rather unpoetic offshoot. It runs hot. Very hot. Training large AI models requires enormous clusters of specialised processors, all packed tightly into server racks that generate massive amounts of intense heat. Without constant cooling, the hardware would cook and fail. Water is the most efficient way to counter this heat. Most data centres use evaporative cooling systems – essentially industry-scale versions of the human body’s perspiration mechanism. Water absorbs heat and evaporates into the air, carrying thermal energy away from delicate electronics. The problem is that just like sweat, the water used up does not magically return to the body (the industrial tank in this case). It vanishes into the atmosphere and needs repeated replenishing. Again, and again, and again. In perpetuity.

It that’s not enough of a bother and you still have your pants on, here’s more. Beyond direct cooling, AI’s water footprint extends deeper into the supply chain. Electricity generation itself consumes large volumes of water, particularly in thermal power plants that use steam turbines and cooling towers. Experts estimate that indirect water use through electricity generation can account for 80 per cent or more of the total water footprint of data centre operations.

Even manufacturing the technology adds to this unpalatable tally. Producing microchips, servers and semiconductors involves water-thirsty processes. A single chip needs several gallons of purified water during fabrication. This cacophonic symphony creates a peculiar paradox of the digital age. The more virtual our lives become, the more physical resources are consumed in facilities few people ever see.

India’s Digital Dilemma

For water-rich regions, the water problem is somewhat manageable. For India, it is alarming, for the country is among the most water-stressed in the world. Groundwater tables are falling across large parts of the nation, with many major cities periodically facing acute shortages. If this backdrop weren’t scary enough, India has also embarked on an ambitious mission to become a global hub for artificial intelligence, cloud computing and digital infrastructure.

The result is a collision of aspirations. India hosts around 250 data centres, with clusters concentrated in Mumbai, Chennai, Hyderabad and Bengaluru. These cities are already struggling with water scarcity. Thus, 60-80 per cent of all data centres in India will face critical water stress in just a few years, given current water usage trends. The scale is staggering. Data centre water usage in India is projected to rise from 150 billion litres in 2025 to more than 358 billion litres by 2030.

That is not a mere statistic. It represents water that might otherwise supply households, agriculture or ecosystems. Policymakers are finally beginning to understand the quandary, and some states are reassessing data centre policies, weighing the economic benefits of digital infrastructure against the hidden environmental costs. While industry promises investment and technological prestige, only a few hoarse whispers reveal that data centres generate relatively modest employment, especially when compared to their enormous resource footprint.

In a country where summer headlines feature reservoirs running dry and taps turn into discussion points in Parliament, the optics are awkward – thousands of litres of drinking-grade water being used to cool machines so that someone can generate AI-written political or marketing spiel.

Hidden Economy

To give the devil his due, the global technology industry is not completely oblivious to the issue. Major tech firms now issue water-usage data and sustainability reports, and some have even pledged to turn ‘water positive’ by replenishing more than they consume. Some are experimenting with alternative cooling systems such as immersion cooling, advanced liquid cooling loops and the use of recycled or treated wastewater instead of potable supplies.

While promising, these innovations are far from universal. In many regions, the simplest and cheapest solution remains traditional evaporative cooling, which means freshwater usage. This brutal reality means that companies build facilities where land, electricity and water are easily available; not where ecosystems can afford the extra demand. That has led to warnings that water scarcity will become a serious financial risk for the industry soon. If local communities or governments impose restrictions on water use, data centre expansion may slow or shift locations. In effect, the availability of water may end up determining the geography of the digital economy.

That leads to a fascinating reversal of history. For centuries, civilisations formed around rivers because humans needed water to survive. In the 21st century, rivers may again shape the map; this time because computers do.

Responsible Future

None of this suggests that AI or digital infrastructure should be abandoned. The technologies powering the Internet have brought undeniable benefits: from telemedicine and financial inclusion to scientific discovery and climate modelling. The challenge is not deciding whether the digital revolution should continue, but how it should unfold responsibly.

Some steps are clear, provided we accept reality and begin implementing. One, transparency must improve. Data centre firms disclose electricity use but seldom provide details on water consumption, especially indirect use through energy supply chains; truthful reporting will allow an understanding of the actual impact. Two, tech innovation must accelerate. Cooling systems that need less freshwater are being developed, such as closed-loop liquid cooling, immersion cooling and air-based systems. These require higher upfront investment, but can dramatically reduce water demand. Three, location. Building data centres in regions with ready water resources or cool climes will lower ‘green’ pressure. Conversely, putting up computing clusters in water-scarce areas is asking for trouble.

The policy implications are clear. India’s digital ambitions are legitimate, but they must be balanced against the reality of water security. Pushing the use of treated wastewater, mandating water-efficient cooling tech and integrating data-centre planning with urban water management are essential steps.

The lesson is simple yet profound. Technology does not exist outside nature. Every algorithm, every cloud server, every AI prompt is anchored in the physical world. It draws electricity, minerals and water from a finite planet. The illusion that digital progress floats above ecological limits may be convenient, but it is nonsense; it is dangerous too. The future of AI depends not only on faster chips and smarter code, it also banks on something far older and simpler. And sweeter. Water.

The writer can be reached on [email protected]. Views expressed are personal

The writer is a veteran journalist and communications specialist