Cooling the Cloud: Tackling Data Centres’ Water Use

Data centres, the online core driving everything from social media to artificial intelligence, are facing growing scrutiny for their massive water consumption. Climate change is growing and water shortages expanding as water finds itself more and more in the spotlight.

Water has a central role to play in data centre operations. The three main areas of application are cooling systems, humidification, and generating electricity. Servers produce a lot of heat, and if not properly managed, overheating leads to malfunction or system failure. To avoid this, water-based cooling mechanisms in the form of chillers, cooling towers, and adiabatic economisers are used by most data centres. These methods employ water to cool air or extract heat by evaporation. Though some of these facilities recycle water, activities such as evaporation and water treatment continue to lead to constant losses.

Humidity control is a further component of water use. Drying air can lead to static electricity build-up, potentially harmful to delicate devices. Humidification systems apply water vapour in order to keep humidity within its optimum level. This, however, causes mineral deposits within the system to be flushed and water replenished periodically—increasing the total amount of water used.

Data centers also indirectly add to water use through the power they consume. Fossil fuel-based power, particularly coal-based power, burns enormous amounts of water in generating steam and cooling turbines. In contrast, coal-fired power uses close to 19,185 gallons of water per megawatt hour (MWh), while natural gas plants use approximately 2,803 gallons per MWh. Compared to wind and solar energy sources, which need little water, the future of energy production seems sustainable.

Major tech firms have begun reporting water usage data. Google consumed 8.1 billion gallons in 2024, almost twice its 2021 amount, primarily due to AI-related workloads. Though the company is investing in water replenishment initiatives, they only offset around 64% of total use and don't always have the impact in the regions where the water was initially used.

Amazon boasts a water usage effectiveness (WUE) of 0.15 litres per kilowatt hour (L/kWh), much less than the business average of 1.8 L/kWh, and has made a commitment to become water-positive by 2030. Microsoft, which hosts more than 300 data centres across the world, has a WUE of 0.30 L/kWh and is now piloting closed-loop cooling systems designed to remove evaporative losses. Even with these measures, regional communities are growing more concerned about the effect of these huge withdrawals. In Spain, Amazon's proposed warehouses have been licensed to siphon sufficient water to irrigate hundreds of acres of land—providing cause for concern regarding sustainability and regional resource shortages.

A number of solutions are being pursued in the industry to deal with these issues. These comprise the use of closed-loop and air-based cooling technologies that restrict water loss, and deployment of greywater reuse systems and rainwater harvesting. The shift to renewable energy sources also greatly diminishes indirect water use associated with conventional power generation. Additionally, replacing old infrastructure with present day, water-efficient technology and locating data centres in naturally cooler climates are feasible means of reducing reliance on water for cooling.

Yet another new-school strategy is circular water management, through which plants seek to recycle water within the system and replenish local resources wherever possible. This integrated approach dovetails with overall environmental sustainability and enables long-term operational resiliency.

Yet, the fast development of AI workloads is worsening the situation. Latest research shows that training big AI models can use up to 700,000 litres of water. Left unchecked, AI demand may consume more than 6 billion cubic metres of water every year by 2027. This growth creates a high risk to already stressed ecosystems, particularly in areas already suffering from water stress and biodiversity loss.

The growth of data infrastructure thus needs to be paralleled by strong water management policies and increased transparency. With increasing demands, sustainability practices are no longer a choice but something to be adhered to in order to address environmental concerns while enabling further technological advances.

Source: TechTarget