By Hari Prasad, Managing Director, India, Gradiant
India’s data-centre build-out is accelerating, with 1,789 MW of operational capacity across its top seven markets in H1 2026 and another 3,860 MW in the pipeline through 2030. AI is further driving this growth. AI workloads now directly account for nearly 20% of total leading demand, and AI-related colocation surged to 348 MW in 2025, more than doubling.
This growth is changing what data centres require from their surroundings. While power has long dominated conversations around capacity and site selection, cooling is becoming equally critical, bringing water certainty into focus.
We have seen this firsthand in India’s semiconductor sector. At a leading U.S. chipmaker’s new fab in one of the country’s most water-stressed regions, freshwater was limited, and a zero liquid discharge mandate was applied from day one. Rather than treating water as a utility connection, the plant’s ultrapure water production, wastewater recycling and zero liquid discharge were designed and delivered as one integrated system. The fab now recovers 96% of its water and has cut disposal volumes by 98%, leaving only solid salts. Data centres have different water needs from fabs, but the lesson carries over: when water is planned as core infrastructure from the start, it becomes a source of certainty rather than risk.
For infrastructure designed to operate for decades, reliable water access is no longer just a utility consideration. It is increasingly linked to operational continuity, permitting, expansion, regulatory compliance and long-term resilience.
AI is not only rewriting the thermal equation but also transforming the water equation for data centres.
AI workloads are pushing rack densities beyond what conventional air cooling was designed to handle, with AI racks reaching 100 kW or more. At these densities, managing heat requires more precise thermal management.
Liquid cooling, including direct-to-chip and immersion systems, is increasingly part of the solution. But liquid cooling does not automatically mean lower water consumption. Heat still needs to be rejected through chillers, cooling towers, dry coolers or other systems, each with different energy and water implications. Reducing one part of the cooling footprint can increase another if the overall system is not designed holistically.
Water strategy must therefore be part of the cooling strategy. The question is no longer simply whether to move from air to liquid cooling, but how to design the thermal and water ecosystem to ensure reliable water availability, quality and treatment capacity throughout the facility’s operating life. This assessment needs to begin early, alongside power and site planning.
Water certainty is becoming a competitive advantage
Water has historically been treated as a utility input. For data centres, it is becoming a key factor that affects site selection, expansion, operating costs, permitting, and community acceptance.
For large AI facilities, operators need to assess not just how much water is used, but how reliably it can be sourced, treated, recycled and reused over the life of the asset. A reliable supply today does not guarantee continuity over a 20- or 30-year asset life. Seasonal variability, regulatory changes, competing demand, infrastructure constraints and future expansion can all affect water availability.
An early water strategy can address these risks by assessing water sources, treatment needs, discharge requirements and regulatory obligations before they become constraints. It can also support smoother permitting and reduce pressure on freshwater resources shared with surrounding communities, regardless of whether a facility is located in a traditionally water-stressed region. The objective should therefore move beyond water conservation to water resilience and certainty.
From water efficiency to water security
Water Usage Effectiveness (WUE) is a useful measure of water intensity, but efficiency alone does not guarantee operational continuity. A facility can have strong WUE and still face disruption due to supply constraints, infrastructure changes, regulatory shifts or water-quality issues. Water security requires a broader approach, built on three strategies.
Three strategies to ensure water certainty in AI data centres
- Diversify water sources
Freshwater should not be the default. Evaluate treated municipal wastewater, reclaimed industrial water, seawater where appropriate, and other non-potable sources to reduce dependence on a single supply.
- Maximise water circularity
Recover, treat and reuse water wherever feasible. Advanced treatment, optimised cooling cycles and, where appropriate, high-recovery or zero-liquid-discharge systems can reduce freshwater intake and manage discharge.
- Design for future resilience
Cooling and water systems should be designed for changing conditions, future expansion and evolving regulations. Build in treatment capacity, alternative sources and flexibility from the outset to ensure long-term reliability.
Measure water like power
Data-centre resilience has traditionally focused on redundant power, backup generation, network availability and physical security. Water management is now becoming part of that equation. Reliable cooling depends on more than water availability. It requires the right sources, treatment, quality, recovery and discharge management. Integrating advanced treatment, reuse and circular water systems into facility planning can turn water from an operational risk into a dependable utility.
Water requirements should also be considered alongside power requirements. Data-centre capacity is typically measured in MW and GW because power has historically been a defining constraint. As AI increases cooling requirements, water needs should become equally visible in infrastructure planning. A future data-centre proposal may need to answer not only “How many megawatts will this facility require?” but also “How many million litres of water per day will it require, from which sources, at what quality and with what recovery rate?”
Building for the next phase of AI growth
India’s data-centre decisions made today will shape resource requirements for decades. Cushman & Wakefield estimates $29.9 billion in development capital will be required through 2030, creating an opportunity to build water resilience into new capacity from the outset. As AI increases computing density and cooling needs, water, energy and cooling must be planned as an integrated system, considering source reliability, treatment, recovery, discharge and future expansion.
The most resilient data centres will not simply use the least water but achieve the greatest water certainty through intelligent sourcing, treatment, reuse and recovery. As AI scales, water security will become as fundamental to reliability as power security.
The next competitive advantage may not be more power or better cooling, but greater certainty about where the water will come from.