The Good News - Water and Digital Growth can coexist A positive pathway from water conservation to resilient Artificial Intelligence (AI) and citizen-centric digital services Data centres are now forming the physical foundation of digital public services, as well as of Artificial Intelligence (AI), banking and payments, healthcare, education, research and industrial automation. The chance exists to make this growth water-smart from the start by using less freshwater, switching from evaporative to closed-loop cooling where suitable, recovering appropriate water streams, and constantly optimising operations. Water stewardship need not hinder innovation; in fact, it can present a technological opportunity by integrating thermal engineering, sensors, advanced treatment, AI control, and circular-water design in such a way that each unit of fresh water generates more digital and social value. A Circular Architecture for water smart compute Figure. Original illustration synthesizing publicly documented practices from the U.S. Department of Energy (DOE), Federal Energy Management Program (FEMP), NITI Aayog, Google and Microsoft. [1–5] CDU = coolant distribution unit. Technology choices that change water equation Technology Options for Water-Efficient Data Centre Cooling and Circular Water Management Technology What it does Water benefit Design note Direct-to-chip liquid cooling Transfers heat from high-density central processing units (CPUs) and graphics processing units (GPUs) directly to a recirculating liquid loop through a coolant distribution unit (CDU). Can eliminate or sharply reduce evaporative cooling water. Best suited to high-density Artificial Intelligence (AI) /high-performance computing (HPC) racks; requires thermal, fluid and reliability design. Rear-door heat exchangers Removes rack heat through a liquid heat exchanger mounted at the rear of the rack. Reduces room-level cooling demand and can support higher rack density. Useful for retrofits and mixed air/liquid environments. Dry/hybrid cooling Uses air-cooled or hybrid heat rejection where climate and workload permit. Reduces dependence on evaporative cooling. Balance water savings against energy use, peak heat conditions and resilience. Water treatment + recovery Uses filtration, side-stream treatment and, where justified, reverse osmosis (RO) to recover suitable blowdown or wastewater streams. Reduces freshwater makeup demand and increases reuse. Assess energy, concentrate/residuals, water quality and operating cost. Condensate + rainwater Captures heating, ventilation and air-conditioning (HVAC) condensate and rainwater for suitable non-potable uses. Offsets potable/freshwater demand for selected applications. Requires storage, treatment, quality monitoring and local compliance. Artificial Intelligence (AI) + digital twins Combines thermal, water, weather and workload telemetry to predict demand and test operating strategies. Cuts avoidable cooling and water use while maintaining thermal limits. Use closed-loop control with safeguards, alarms and human oversight. Measure It: Water Usage Effectiveness (WUE) Is The Water Companion To Power Usage Effectiveness (PUE) Water Usage Effectiveness (WUE) is defined as the ratio of annual site water use to annual Information Technology (IT) equipment energy use, expressed in litres per kilowatt-hour (L/kWh). The U.S. Department of Energy (DOE) and the Federal Energy Management Program (FEMP) regard WUE as a practical metric at the site level and point out temperature/humidity optimization, economizing, cooling-tower practices and direct liquid cooling as possible ways of reducing water usage. The Water-Amart Al Enterprise From responsible cooling to intelligent water management Figure. Original technology roadmap showing how measurement, cooling, recovery, reuse and Artificial Intelligence (AI)-based operations can work together. STP = sewage treatment plant; WUE = Water Usage Effectiveness; PUE = Power Usage Effectiveness. AI Can Optimize the Water System that Sustains AI The same kind of intelligence that is employed in citizen-centric applications can be used to manage the infrastructure supporting them. A water-aware control system for a data centre can integrate data from flow meters, temperature and humidity sensors, pressure sensors, conductivity sensors, and those measuring total dissolved solids (TDS), along with readings from the tanks, weather forecasts, the IT load, GPU utilisation and telemetry from the cooling system. Machine-learning models are able to identify normal operating patterns, detect leaks or fouling, predict cooling requirements and suggest the lowest-water operating mode that meets both thermal and reliability constraints. Digital twins introduce an additional feature which allows operators to virtually test out cooling strategies, water-reuse situations and workload schedules before making any changes to the actual facility. AI is also capable of shifting non-urgent workloads in terms of time or location when this results in reduced cooling stress. Under this approach, sustainability turns into an operational intelligence issue and ceases to be a one-off reporting task. Design the Campus as a Water Cascade Water Management Cascade for Sustainable Data Centre Operations Stage Action Examples Outcome 1. AVOID Reduce heat and cooling demand at source. Efficient IT, server power management, thermal zoning, workload optimization. Less heat to remove; less water needed. 2. REDUCE Tune systems so the same cooling service uses less water. Temperature/humidity setpoints, airflow, economizers, cooling-tower cycles. Lower water intensity and operating cost. 3. REUSE Use fit-for-purpose non-potable water streams. Treated wastewater, condensate, rainwater for suitable uses. Freshwater is reserved for higher-value needs. 4. RECOVER Treat suitable streams for additional recovery. Cooling-tower blowdown treatment, filtration, RO where justified. More water stays within the site water loop. 5. REPLENISH Support local watershed resilience beyond the site boundary. Recharge, restoration, community water-security projects. Digital growth contributes to local water resilience. Proof that the Technology Direction is Already Moving Google states that its water-stewardship programme in 2025 resulted in the replenishment of about 7.7 billion gallons—around 78% of its total freshwater use—and included 165 projects across 97 watersheds.[3] Microsoft has set a water-positive target by the year 2030 and has developed a new data-centre design which requires no water for cooling; according to its 2025 sustainability report, the design can save more than 125,000 cubic metres of water each year per facility under the conditions mentioned.[4] They are merely examples, not general rules. The engineering decision should take as its basis the local water stress, the climate, the water quality, the available energy, the workload density, the reliability requirements, and the total lifecycle impact. Key Water and Energy Performance Indicators for Data Centre Sustainability. WUE PUE Water reuse Water quality Replenishment L/kWh Facility energy/IT energy % suitable demand reused pH, TDS, conductivity m³/local outcome THE POSITIVE STORY - Al and water stewardship can reinforce each other. A water-smart data centre can power digital public infrastructure and business innovation while reducing freshwater demand, recycling suitable water repeatedly, and contributing to healthier local watersheds. References: U.S. Department of Energy (DOE), Federal Energy Management Program (FEMP), “Cooling Water Efficiency Opportunities for Federal Data Centers.” energy.gov/cmei/femp/cooling-water-efficiency-opportunities-federal-data-centers NITI Aayog (National Institution for Transforming India), Government of India, “Reuse of Treated Wastewater in Urban/Peri-Urban Agriculture in India,” June 2023; and “Water Trading Mechanism to Promote Reuse of Treated Wastewater,” July 2023. niti.gov.in Google Sustainability, “2026 Environmental Report” and “2026 Google Water Stewardship Project Portfolio.” sustainability.google Microsoft, “2025 Environmental Sustainability Report,” including direct-to-chip cooling and water-reuse technologies. microsoft.com Microsoft, “2026 Environmental Sustainability Report,” including data-centre water-efficiency improvements. This article is written by Professor Amlan Chakrabarti , Visiting Professor, Department of Artificial Intelligence, IIT Kharagpur | Professor, A.K. Choudhury School of IT, University of Calcutta [Published in NIWAS Vartika - Vol-II, Issue-2 (Jul-Sept'26): A WASH magazine by SPM NIWAS, MoJS, GoI]