But AI infrastructure is not abstract. It’s physical, resource-intensive and tied to local ecosystems. This raises the question: what happens when this rapid expansion collides with a country facing severe water stress?
Global tech companies are committing investments to expand cloud capacity in India. Domestic players are doing the same. The goal is to make AI cheaper, faster and more accessible.
Data centres, which power everything from search engines to genAI models, require cooling to operate. A single large facility can consume as much water in a day as a small village. Even in the US, the strain is becoming evident, from groundwater pressures in Texas to rising withdrawals in Virginia’s ‘Data Centre Alley’.
In India, this trend intersects with a fragile reality. The country is home to nearly 18% of the world’s population, but has access to only about 4% of its freshwater resources. Most cities are already dealing with falling groundwater levels, infra pressures, and growing competition between domestic, agricultural and industrial use.
These same cities are now becoming hubs of data centre expansion. Hyderabad, Bengaluru and parts of Maharashtra are attracting significant investment in AI infra even as they face recurring water shortages. In Pune, concerns over water access have already sparked public pushback.
Unlike electricity, which can be transmitted, water is local. When data centres draw from urban supplies, they compete with households, farmers and small businesses, often relying on potable, drinking-grade water. At the same time, AI’s water footprint extends well beyond data centres. Power generation itself, including some RE sources, can be water-intensive, and production of advanced semiconductor chips requires vast quantities of ultra-pure water. The true resource demand of AI infra is far greater than what is immediately visible.
India’s AI ambitions are designed to address pressing domestic challenges. From improving agricultural productivity, expanding access to healthcare and widening educational opportunities, AI is being positioned as a tool for inclusive growth. But the infra that enables these solutions may also be placing additional strain on the systems that sustain everyday life.
More computing requires more energy. More energy often requires more water. In a country where water is already scarce, this creates a feedback loop that cannot be ignored. And, yet, policy frameworks have not fully caught up with this reality.
While several states have introduced incentives to attract data centre investments, only a handful meaningfully incorporate sustainability criteria. As a result, facilities can be built in water-stressed regions without standardised reporting or clear accountability to local communities. A clear unified national framework governing water use or disclosure is needed.
Things can get worse as demand scales. More than 60% of India’s existing data centre capacity is expected to operate in high water-stress regions this decade. Corporate sustainability commitments offer some reassurance. But their impact remains uneven. Major tech companies have pledged to become ‘water positive’, investing in watershed restoration and replenishment efforts. Yet, these interventions often operate at a broader regional level and don’t necessarily offset localised strain, particularly in cities already facing shortages.
Globally, Google alone reported using over 5.6 bn gallons of water for its data centres in 2023, a figure that continues to rise alongside AI demand. Infrastructure is also proving vulnerable to environmental stress. During a 2022 heatwave in Britain, cooling failures forced temporary shutdowns of major data centres during demand peaks.
Despite this, incentives driving expansion remain largely economic. Tax benefits, subsidised land and fast-track approvals continue to prioritise rapid growth, often without integrating long-term water and energy constraints into planning decisions.
None of this suggests that India should slow down its AI ambitions. The real challenge is not whether India should invest in AI infra, but how.
Location-specific planning must become central to decision-making.
In water-scarce regions, less water-intensive cooling systems may need to be prioritised, even at the cost of higher energy use.
Standardised reporting on water and energy use would enable better oversight and more informed public debate.
India has demonstrated global leadership in building DPI. The next phase will require applying that same rigour to physical systems that support it. Success of a digital future will depend not only on data and compute, but on whether natural resources that sustain it are managed with equal ambition and care.
The writers are co-founders, Leherum