When Bytes Meet Water: How India's Data Centres Are Drinking the Planet Dry

It’s a quiet drama playing out across India’s tech hubs – rows of blinking servers, humming cooling units, and, oddly enough, a lot of water being poured down the drain. As artificial intelligence spreads its roots, the country’s data centres are suddenly finding themselves at the centre of a growing water‑stress conversation.

Think about it: a typical data centre needs to keep its hardware cool, and in most Indian climates that means relying on water‑based cooling systems. One megawatt of IT load can easily soak up 1,000 to 1,500 litres of water every hour, depending on the design. Multiply that by the dozens of mega‑scale facilities sprouting up in Bengaluru, Hyderabad and Delhi, and you’ve got a water‑use figure that rivals that of a small town.

Why does AI make the problem worse? The answer is simple – AI workloads are thirsty. Training a large language model can demand a handful of GPUs running flat‑out for weeks, and each GPU rack is a heat‑generator on steroids. The more complex the model, the more cooling power – and consequently water – you need.

India is already feeling the squeeze. According to the Ministry of Jal Shakti, per‑capita water availability dropped from 5,177 cubic metres in 2000 to just 1,521 cubic metres in 2023. Some regions, especially the arid zones of Rajasthan and Gujarat, are classified as “water‑scarce” and are seeing groundwater levels dip alarmingly.

So what does this mean for a country that’s racing to become a global AI hub? In the short term, data centre operators may face higher operating costs as water pricing climbs. In the long term, the mismatch could force companies to rethink where they locate new facilities – perhaps moving toward cooler, more humid coastal areas where air‑side cooling can cut water use dramatically.

Fortunately, a handful of innovative solutions are bubbling up. One approach gaining traction is liquid‑cooling at the chip level, which uses a closed‑loop system and recirculates a coolant rather than constantly pulling fresh water. Companies like Dell and Nvidia are already shipping servers with these systems, and early adopters in India report up to a 40% reduction in water consumption.

Another avenue is harnessing renewable energy‑driven “dry” cooling technologies – using advanced heat exchangers and phase‑change materials to dissipate heat without water. While the upfront cost can be higher, the long‑term savings – both financial and environmental – are compelling, especially as the country’s carbon‑pricing mechanisms evolve.

Policy also plays a starring role. The Indian government recently rolled out the “Sustainable Data Centre Initiative,” which offers tax incentives for facilities that meet stringent water‑efficiency benchmarks. On the ground, a pilot program in Pune is pairing data centre operators with local municipalities to recycle grey‑water from nearby treatment plants for cooling purposes.

All said and done, the story isn’t all doom and gloom. The very technology that’s amplifying water demand – AI – also holds the keys to smarter water management. Predictive analytics can flag leaks, optimise irrigation, and even guide real‑time cooling adjustments in data centres, shaving off litres that would otherwise be wasted.

In a nutshell, India stands at a crossroads where digital ambition meets physical limitation. The choices made today – from embracing cutting‑edge cooling tech to crafting clever policy levers – will dictate whether the nation’s AI dreams stay afloat or run dry.