How AI Data Centres Cut Annual Water Consumption by 50%

Liquid cooling doesn’t have to mean water use, says Schneider Electric. 

“We do not need to consume water to operate data centres,” said Tuan Hoang, Head of Cooling Technology and Product Development at Schneider Electric, addressing recent investigations into the water footprints of Gen AI facilities and the potential for local water stress.

“Zero water is needed to cool AI data centres. Liquid cooling is required but it’s for the load and radiators. 

“Water consumption for data centres is a choice – a geographical choice dependent on power, land and what is required.”

While some operators may still choose water-consuming approaches, Tuan drew a clear line between water use and liquid cooling itself. 

For AI racks approaching 400kW, liquid cooling is now an operational necessity – yet, with the right architecture, it doesn’t have to draw on local water resources.

Cutting water consumption by half

To illustrate the impact of transitioning from traditional air cooling to liquid cooling, Schneider Electric presented two modelled data centre scenarios using projections for Dallas, Texas, and Paris, France. In both cases, transitioning to liquid cooling roughly halved the annual water consumption. 

The projected water usage for the Dallas data centre reduced from 382,000 cubic metres per year using traditional air cooling to 197,000 cubic metres with liquid cooling – a 48% reduction. 

Likewise, the Paris facility dropped from 108,000 cubic metres to 51,000 cubic metres, representing a 53% savings. 

“It’s a choice to how you reject the heat, and the myth that all data centres with liquid cooling are using lots of water isn’t true,” Tuan said. “We focus on high-efficiency chilling.”

Closed-loop engineering: liquid cooling without water consumption

Schneider Electric pointed to its Uniflair XCA line – pre-engineered, air‑cooled chillers designed for data centre environments – as an example of a heat-rejection strategy that avoids onsite water use. 

“It does not use water, just radiates heat out,” Tuan noted. 

Key features include continuous cooling capacity up to 2.4MW per unit, fast restart capabilities, dual-feed power redundancy and closed-loop design using a factory-sealed volume of high-quality cooling fluid. 

According to Schneider Electric, the fluid formulation is tested at Motivair factories and remains contained within the system for the facility’s operational life. 

With no evaporation or discharge, these closed-loop systems are designed to eliminate routine water consumption tied to cooling.

Why liquid cooling is now mandatory for AI

Liquid cooling has existed for decades, but its role in AI-scale computing has shifted from optional to essential, said Rich Whitmore, CEO of Motivair by Schneider Electric. 

Motivair has been developing liquid cooling systems for 400kW racks for more than a decade.

“People don’t have a choice – if you want advanced AI systems going in, you have to cool them,” Rich said.

As AI clusters densify, closed-loop liquid cooling reframes the sustainability conversation. 

Rather than equating liquid cooling with water use, Schneider Electric argues it is a high-efficiency thermal approach that can significantly reduce or eliminate reliance on external water supplies.