As India races to build the digital infrastructure that will power its AI driven future, an important question remains largely absent from public discourse: what resources are quietly sustaining this technological revolution? Behind every AI query, cloud service, and data intensive application lies a vast network of data centres that consume not only energy but also significant volumes of water, often in regions already grappling with water stress.
The challenge is particularly relevant for a country balancing rapid economic growth, digital expansion, and climate resilience. Decisions made today about where and how digital infrastructure is built will have lasting implications for communities, natural resources, and sustainable development.
Drawing on his experience in business transformation and sustainable industrial development, Vipul Kumar, Partner and Managing Director India at Xynteo, examines the often overlooked water footprint of AI and data centres. He argues that India has a unique opportunity to embed sustainability into its digital growth story from the outset, ensuring that technological progress does not come at the cost of long term resource security.
India's data centres consumed an estimated 150 billion litres of water in 2025, a figure projected to more than double by 2030, underscoring where smarter infrastructure decisions, better policy, and coordinated investment can matter most. Now is India's chance to figure this out early and build digital infrastructure for the long haul.
A conversation with a high-performance AI model uses the approximate equivalent of a large glass of drinking water to power and cool the servers processing your query—pulled from municipal supplies, groundwater, or local reservoirs, depending on where the data centre sits. Most of us remain blithely unaware of both the scale and intensity of the issue. And that invisibility is a challenge we must solve for.
The footprint, and what it tells us
Cooling technology shapes both the power and water footprint of a data centre. Many facilities rely on evaporative systems—releasing water vapour into the air to dissipate heat. A single 100 megawatt facility on this model draws roughly two million litres a day. Closed-loop alternatives circulate water in a sealed circuit and use a fraction of that volume. The engineering is not experimental or expensive to operate—it exists, it works, and choosing it is largely a matter of where operators set their priorities.
Research published in Nature Sustainability puts AI's total water footprint at between 312 and 764 billion litres last year. India's four largest data centre markets, Mumbai, Bengaluru, Chennai, and Hyderabad, are all water-stressed by international classifications. Bengaluru experienced its worst water crisis in five centuries in 2024; Hyderabad faces a projected daily deficit of 870 million litres by 2027. The data centre boom is heading straight into these realities. That is precisely why the conversation about where and how to build needs to happen now, before patterns set in that are expensive to undo.
The opportunity to get it right
Cornell University researchers found that combining smart siting, grid decarbonisation, and operational efficiency can reduce a data centre's water consumption by up to 86%, reframing the entire discussion towards building better.
Cities such as Pune, Coimbatore, Jaipur, and Bhubaneswar don't get talked about enough in this conversation. Cooler temperatures, less pressure on water supplies, available land, and better economics than the overcrowded metros—the case for shifting some of India's data centre growth to Tier II and III cities practically makes itself. The jobs follow, regional infrastructure catches up, and the whole system gets a little less fragile. What's interesting is that the investment logic and the sustainability logic land in the same place.
When it comes to cooling technology, the opportunity is similarly concrete. Closed-loop systems recirculate water rather than consuming it, but efficient designs go unrewarded without a standard way to measure them. The industry already does this for power: Power Usage Effectiveness (PUE) is reported, scrutinised by investors, and competed on by operators. Water Usage Effectiveness (WUE) exists as a metric but is rarely disclosed with the same rigour. Treating it as a first-class benchmark—reinforced by something like a BEE-style rating—would give developers a clear target and investors a reliable way to compare assets. In fact, the Council on Energy, Environment and Water has already proposed frameworks along these lines.
Coordination to unlock systemic action
No individual operator, however well-intentioned, can solve a structural problem through internal policy alone. Siting standards, disclosure requirements, cooling benchmarks need to be shared frameworks, not voluntary commitments made in isolation. In fact, only five of India's 15 state data centre policies contain any sustainability parameters at all, leaving significant room to build better defaults into the policies still being written.
This is exactly the kind of challenge that rewards coordinated action across investors, operators, and governments. One such example is PoweringAI, a spin-off of Xynteo developing brownfield industrial sites across Europe—former factories and legacy industrial zones with existing power, land, and water infrastructure—into de-risked data centre developments. The principle is straightforward: design power, water, land use, and community impact into the brief from the start, rather than solving for them after the concrete is poured. Applied to India, that logic could anchor a new generation of development on sites where water and grid capacity are already accounted for—infrastructure built around local realities, not despite them.
India can set the standard
Microsoft, Amazon, Google, and the Adani Group together committed over $167 billion to Indian data centre development in the past 18 months. That capital would be most effective in building infrastructure designed for the country as it is—water-stressed in parts, rich with untapped Tier II potential, and publicly committed to its climate targets.
India now has a genuine opportunity to define what responsible digital infrastructure looks like as a model for others to follow. The investment wave is large, and it is early. The standards written now will shape the next two decades of AI infrastructure across the Global South.
x
.png)
.jpg)
