Schneider Electric says AI Water Waste is a Choice

Share this article
Share this article
Prioritise Us on Google
Tuan Hoang is Head of Cooling Technology and Product Development at Schneider Electric. Credit: Schneider Electric
Tuan Hoang highlighted that closed-loop liquid cooling is now mandatory for high-density AI racks and can cut data centre water use by about half

Is heavy water consumption an inevitable by product of the AI revolution, or simply a design choice? 

Industry leaders at Schneider Electric argue the latter, demonstrating how next-generation closed-loop liquid cooling can support massive AI workloads while drastically reducing reliance on local water supplies. 

ā€œZero water is needed to cool AI data centres,ā€ said Tuan Hoang, Head of Cooling Technology and Product Development at Schneider Electric, speaking to journalists in Buffalo, New York, at the Schneider Electric & TeraWulf Global Press Event.

His comments arrive amid heightened scrutiny of Gen AI’s environmental footprint, particularly around localised water scarcity. 

Tuan drew a sharp distinction between liquid cooling and water consumption, stressing that the former is about efficiently moving heat away from high-density IT loads – not necessarily evaporating or discharging water onsite.

ā€œWater consumption for data centres is a choice,ā€ he said. ā€œIt’s a geographical choice dependent on power, land and what is required.ā€ 

By contrast, for racks pushing 400kW, liquid cooling is no longer optional. 

ā€œLiquid cooling is required but it’s for the load and radiators,ā€ said Tuan. 

Motivair’s coolant distribution unit factory in Buffalo, New York. Credit: Schneider Electric

Projecting water consumption reduction

To illustrate the impact of liquid cooling, Schneider Electric presented two theoretical case studies comparing traditional air-cooled designs with liquid-cooled approaches in Dallas, Texas, and Paris, France. 

The projections suggest roughly a 50% reduction in annual water use after transitioning to liquid cooling:

  • Dallas: from 382,000 cubic meters per year (air-cooled) to 197,000 cubic meters (liquid-cooled), representing a 48% reduction.
  • Paris: from 108,000 cubic meters to 51,000 cubic meters, representing a 53% reduction.

ā€œIt’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,ā€ Tuan said. ā€œWe focus on high-efficiency chilling.ā€

One example is Schneider Electric’s Uniflair XCA, a pre-engineered, air-cooled chiller line designed for data centres

ā€œIt does not use water, just radiates heat out,ā€ Tuan noted. 

Youtube Placeholder

Each unit delivers continuous cooling up to 2.4MW, supports fast restarts and offers dual-feed power redundancy. 

The systems rely on a closed-loop, factory-sealed volume of high-quality cooling fluid – tested in Motivair factories – designed to last the life of the data centre without evaporation or discharge.

Motivair and liquid cooling

Liquid cooling’s role is quickly becoming non-negotiable for AI-scale densities, said Rich Whitmore, CEO of Motivair by Schneider Electric:

ā€œLiquid cooling has been around since the 1980s but now it isn’t an option, it’s mandatory. People don’t have a choice – if you want advanced AI systems going in, you have to cool them.ā€

Rich Whitmore is CEO of Motivair. Credit: Schneider Electric

Framed this way, implementing closed-loop liquid cooling – paired with air-cooled heat rejection – becomes less about water consumption and more about thermal efficiency. 

For operators planning AI buildouts, it offers a path to high-density performance while substantially reducing reliance on external water supplies.

Executives