The four questions India must ask before its AI data centre boom outpaces its grid

For decades, data centres have done their work quietly, largely out of sight and, until recently, out of mind. They powered digital banking, ecommerce, streaming and enterprise technology. During Covid, they kept the country's digital spine intact while millions worked and learned from home.

That quiet phase is over. Large investment announcements have brought data centres firmly into public view, and AI has brought even greater scrutiny. Four questions require answers.

Do data centres in India really consume more water than comparable industrial or commercial facilities?


Also read: Data centres are coming. Where will India find the people to run them?

If they use less water, does that shift the burden to electricity, and what is being done about efficiency?

If they use more electricity than the most efficient facilities elsewhere, typically benchmarked against the Nordic countries where cold climates make cooling almost free, how is that gap compensated for?

Does this infrastructure affect surrounding communities through grid capacity, availability or cost?

Water usage is 1.8-1.9 litres a kWh of electricity consumption, annualised to roughly 15-16 mn litres per MW. In contrast, a closed-loop air-cooled system, which rejects heat-using fans and refrigerant loops rather than evaporating water consume less than 0.2 litre per kWh. India's nationwide operation consumed 1.1 mn litres per MW, about 7% of the technology referred to raise concerns.

Compare this to commercial facility. A 50 MW data centre is built on roughly 4 lakh sq ft. A commercial space with similar area will employ at least 4,000 people who, with a 45 litre per person a day limit, will consume 65.7 mn litres in a year, compared to 55 mn in a 50 MW data centre.

If the heat transfer is done using a mechanical refrigeration system, does it consume more electricity? Does running compressors harder to save water require additional electricity? Saving water while raising energy consumption would not be a sustainability victory. This is where grey and recycled water use matters as an efficiency lever.

A hybrid cooling system - closed-loop air-cooled system combined with evaporative grey water - reduces electricity consumption by 10%. An air-cooled data centre running at 1.45 PUE (power usage effectiveness; the lower the better) is expected to improve to less 1.3 PUE with grey water.

But is that kind of PUE enough in a climate as hot and humid as many parts of India? The approach industry has adapted is to decarbonise the electricity used through renewable capacity.

As for impact on community, a large data centre typically drawing tens of MW continuously. This raises the question of whether it crowds out capacity meant for homes and local businesses. Most large facilities in India connect through dedicated high-voltage substations rather than the local distribution network, and a growing share of their power comes from renewable generators under long-term contracts, adding new generation capacity rather than simply drawing down what exists. Predictable industrial loads can also improve utilities' load-factor economics. So, well-planned clusters can actually also catalyse grid strengthening that benefits a region well beyond the facility's own walls.

AI will, undoubtedly, require significant physical infrastructure. But growth can't be an excuse for careless resource use. The task is to build AI infra and manage its footprint at the same time, with the same rigour already applied to uptime and reliability, publishing data, acknowledging gaps, and investing continuously in improvement.

India can stay on the sidelines, and later ask why the next generation of technology was built elsewhere. Or it can build, with RE as a backbone, efficiency embedded into design from the outset, and transparency to prove it is working.

(The writer is CEO, Sify Infinit Spaces)