Ecolab: Smart Water Conservation for Future Data Centres

The energy requirements of generative AI have created pressure on data centre cooling infrastructure. A query processed by a generative AI system consumes 2.9 watt-hours of electricity – 10 times the power used by a standard search engine query.  

Scale that across millions of daily queries and by 2030 computing infrastructure for large-language models will require 327 gigawatts of electricity by 2030, 70% of the United States’ total electricity consumption in 2024. 

Power has become the primary bottleneck for data centre growth, but as facilities expand to meet AI workloads, water requirements are an equally pressing concern. The response from some operators has been to adopt waterless cooling methods, though this approach can miss a fundamental point about the relationship between water and energy. 

“In today’s climate, data centre expansion faces a number of obstacles, most notably around reduction of water and energy footprints,” says Mike Obradovitch, Vice President, Data Center Global Accounts at Ecolab. “At Ecolab, we’re working with operators to turn some of those obstacles into opportunities to demonstrate data centres as net contributors to communities as opposed to resource consumers.” 

The company argues that water remains a highly efficient cooling medium, even when broad water consumption is factored in. Natural gas electricity generation requires 570 to 1,100 litres of water per megawatt-hour, creating a water footprint before cooling systems begin operation. Water-cooled data centres then use 10% to 30% less energy than air-cooled chiller applications. The physics are straightforward: water has 3,500 times the heat-carrying capacity of air and transfers heat 23.5 times more effectively. 

Cooling demand fluctuates hourly regardless of weather conditions 

A 100-megawatt data centre uses less water annually than an 18-hole golf course when properly optimised, but the comparison only goes so far. Where golf courses respond to weather patterns, data centre water consumption fluctuates hourly based on cooling load, water quality, equipment efficiency, temperature and humidity. 

“Data centre water consumption can fluctuate hourly, regardless of weather, driven by factors such as cooling load, water quality, equipment efficiency, ambient temperature, and humidity,” Mike explains. “Adapting to these demand fluctuations is the key to maximising system efficiency.” 

Development pace has exceeded what individual operators can manage across design, operation and maintenance. Ecolab has built its business around handling cooling management from site selection through ongoing operations, using digital monitoring to identify efficiency gains that might otherwise go unnoticed. 

Ecolab 3D TRASAR Technology saves US$11m for data centre customers 

The company’s approach is anchored by its 3D TRASAR Technology platform, which includes versions for adiabatic cooling, cooling water systems and direct-to-chip liquid cooling. The direct-to-chip variant uses IoT sensors to track pH, conductivity and glycol concentration in real time as coolant circulates through servers. The cooling water version provides operational control for towers and chillers, while the adiabatic version optimises direct-evaporative cooling through smart equipment and real-time monitoring. 

In 2023, 3D TRASAR Technology for Cooling Water delivered more than US$11m in annualised savings for data centre customers while reducing water consumption, energy use and CO2 emissions. A pilot project with Digital Realty is now testing a related AI-driven water conservation solution across 35 US data centres, targeting a 15% reduction in water use. The deployment is expected to avoid withdrawing up to 126 million gallons of potable water from local watersheds annually while extending equipment life. 

Site selection drives cooling efficiency gains 

Ecolab's methodology starts before ground is broken. Site selection evaluates climate, water availability, power sources and watershed impacts, factors that determine which cooling approaches will work most efficiently. 

“To meet the cooling demands of digital infrastructure, data-centre cooling topology must then be designed with versatility and flexibility in mind, not only for today but for the future,” says Mike. 

When direct-to-chip cooling handles more thermal load, facility systems like cooling towers and chillers operate under reduced demand, cutting water and power consumption. Design incorporating adiabatic and liquid cooling reduces energy use, while water recycling projects decrease potable water consumption. The approach requires ongoing monitoring to maintain performance as conditions change. 

Ecolab has operated in water treatment for more than a century across food safety and environmental management sectors, now applying this experience to data centres requiring liquid-cooling expertise and technical services.