By Vaheed Syed (Research Scholar) & Dr Saikat Sinha Ray (Senior Assistant Professor)
Department of Environmental Science and Engineering, SRM University-AP, Amaravati
AI is seen as a digital revolution, but each AI model ultimately rests on a physical paradigm of electricity, computing hardware, land, energy, and water.As India races to embrace AI, a much less heralded resource challenge is emerging alongside this boom: soaring demand for water to cool the data centres that power the backbone of AI workloads.
The scale of the problem
Cooling is the reason. A data centre routinely produces large amounts of heat, which it may dissipate using water or evaporative cooling systems. As AI workloads become more compute-intensive, demand for cooling resources will likely increase. A rapidly expanding base of cloud computing and AI infrastructure has pushed India’s installed data-centre capacity to about 1.5 GW, with around 6.5 GW expected by 2030. Data centres in India used around 150 billion litres of water a year in 2025, a figure expected to more than double by 2030. CHW requirements can vary greatly at the facility level based on cooling technology, climatological factors, IT loads, and overall facility design. According to the Government of India’s Economic Survey 2025-26, Individual AI data centres need as much as 20 lakh litres of water every day, presenting a looming threat to freshwater resources due to the continuing expansion of AI infrastructure. Importantly, this number should not be interpreted as an average rate of consumption across the globe: data-centre water use varies widely based on factors including workload characteristics, server efficiency, cooling technology employed, climate, energy source, and infrastructure design.
This geographic distribution makes the issue particularly important for India. Water availability varies sharply across states and districts, while many regions already face seasonal or chronic water stress. Consequently, the siting of large data centres should consider not only electricity availability and network connectivity but also local water availability, competing municipal and agricultural demand, cooling technology, and the potential use of non-potable or seawater resources. This is particularly relevant as India expands data-centre infrastructure into coastal regions where alternative water sources may be technically accessible.
A case unfolding at home: Visakhapatnam
Andhra Pradesh offers a case unfolding close to home. Google has announced that they will have a $15 billion (2026-2030) five-year investment in Visakhapatnam to create an AI hub, including gigawatt-scale data-centre infrastructure. The project entered its construction phase in 2026, with the development described as a 1-GW hyperscale AI data centre, while approximately 600 acres were allocated across Tarluvada, Rambilli, and Adavivaram. The scale of this development makes water management an important part of the infrastructure discussion. Instead of treating cooling water as an afterthought, designing such facilities allows consideration of seawater cooling, recycled water, closed-loop cooling, and other water-efficient technologies from day one.
Case study: how Google has already used seawater cooling
Interestingly, one of the biggest software industries has developed a technology that seems to be part of an answer to this challenge long ago. Google has a data centre in Hamina, Finland, whose cooling system draws seawater from the Gulf of Finland. The system is energy efficient and built to harness the site’s coastal setting and minimise the energy required for traditional cooling. Google has adopted numerous water-management measures across its data-centre portfolio, including seawater cooling, industrial canal-water cooling, recycled or grey-water cooling, stormwater capture and reuse, rainwater harvesting, and dry-cooling facilities where possible.
The lesson is not that seawater cooling has this universally transportable process from Finland to India. On the contrary, the Hamina example shows that cooling strategy is essentially an engineering decision unique to a given site. With appropriate intake systems, heat-sink corrosion and biofouling prevention, environmental discharge compatibility, thermal impacts, and pretreatment requirements, coastal data centres may be able to use seawater as a low-cost (compared to river water) heat sink. Coastal cities like Visakhapatnam should assess these considerations before locking in freshwater-dependent cooling infrastructure.
That is where desalination comes in
Seawater cooling certainly cannot solve the broader nexus between AI growth and water security in India. This calls for a broader approach with water conservation at its core, supported by wastewater reuse, desalination, and cooling technologies working in unison.
An instructive example of longer-run diversification comes from Singapore. An exploration of how Singapore integrates water across the Four National Taps that underpins its water-security strategy: collection from local catchments, imported water, NEWater and desalinisation. As we need more water, NEWater and desalination will become even more crucial; in 2060, these two weather-resilient sources could meet as much as 85% of Singapore’s annual water needs. Singapore therefore demonstrates a simple yet important lesson: water security improves when sources are developed before scarcity becomes a crisis. This principle would translate for India’s emergent AI infrastructure into increased use of treated wastewater, seawater desalination, water-efficient cooling, closed-loop systems, and site-specific water-management strategies..
The membrane science behind the scale-up
Membrane experts say that membrane distillation’s ability to scale up will depend not simply on increasing membrane area or system capacity, but also on maintaining membrane stability during long-term operation, particularly anti-wetting, antifouling, and self-cleaning performance. Membrane wetting and biological fouling remain important challenges for highly saline and industrial wastewaters because they can progressively affect vapour transport, flux stability, and salt rejection. From this perspective, membrane surfaces should be designed as multifunctional interfaces in which hydrophobicity, hierarchical surface structure, anti-wetting behaviour, antibacterial activity, and self-cleaning characteristics work synergistically. A high-water contact angle alone, therefore, should not be regarded as sufficient evidence of membrane durability. Long-term resistance to pore wetting, biological adhesion, flux decline, salt-rejection deterioration, and surface degradation should instead form part of the standard evaluation of next-generation membrane-distillation materials.
When AI becomes part of the water solution
The relationship between AI and water treatment is also not one-directional. Membrane researchers say that AI can make desalination and membrane-based processes more predictive, adaptive, and resource-efficient. Their work has explored the application of AI to water treatment and seawater desalination, including artificial neural networks, genetic algorithms, prediction, optimisation, modelling, monitoring, and decision-making.
So, AI should not be used solely to predict which plants will perform well. It can be harnessed to create adaptive, smart water-treatment systems that respond dynamically in real time to changing feed characteristics, fouling behaviour, energy demand, and operating conditions. This integration could enhance process monitoring, optimisation, predictive maintenance, and operational decision-making.This presents a fascinating paradox: the same technological revolution that is intensifying pressure on water resources might also equip us with tools to use water more intelligently. Technologies for AI-driven optimisation, advanced membrane materials, seawater desalination, wastewater reuse and water-efficient data-centre cooling need not be independent agendas. They may combine into a single water-energy-digital infrastructure strategy.
Overview
India’s AI ambitions must therefore grow alongside itswater-security aspirations. The issue is not whether India should adopt AI infrastructure, but how to design such systems without intensifying pressure on this resource-poor country’s water resources. Coastal sites allow for a re-examination of the traditional reliance on freshwater for cooling through a combination of seawater-based heat rejection, solar desalination and treated wastewater, membranes and smart process control. Countries like Singaporeshow that water resilience should be planned for in long-term infrastructure, not after a crisis. For India, mainstreaming water security into an AI economy would thus help to determine if its digital expansion is a vehicle for resource conflict or provides enabling opportunities to enhance the efficiency and resilience of India’s water-technology system.
