Water consumption has become one of the most debated aspects of the data center industry, as concerns mount over usage.

As pressure heats up on companies to cater to AI demand, so too does their environmental responsibility. Tuan Hoang, who leads the product development and innovation team for Schneider Electric’s cooling business, is well-versed in water and cooling discussions being a firm priority.

“Water usage for cooling has been around and leveraged for decades to provide efficient cooling, but that has shifted,” he explained during a recent DCD>Talks interview. “AI and accelerated compute have brought it to the forefront.”

It’s important to understand all aspects of water usage and efficiency in terms of what it means, how much the industry is consuming, what it’s being used for, and how new technologies and new compute workloads can reduce reliance on water and energy consumption.

For Hoang, that balance must be taken into account as the industry looks at data center efficiency. Particularly as these conversations come during a time of immense public pressure and resistance to data centers.

“For the general public, transparency and data‑driven decision‑making should be more relevant,” he added. “The industry is demanding that everyone in the ecosystem is responsible.”

Water usage for cooling has been around for a long time, with cooling towers and evaporative cooling often being used to remove heat without relying on mechanical cooling.

“The decision around how to cool a data center has two main aspects: the climate you are operating in and the design conditions that you need to operate at,” Hoang said.

While cooling enables lower energy consumption, more advanced mechanical cooling equipment enables the gap to close between water usage and energy efficiency.

“The set point is important when cooling data centers. As we move the design condition up to 20°C or even 30°C (68-86°F), it allows us to operate the data center with more economization hours, basically free cooling where you’re not consuming water and you’re not spending energy on mechanical cooling,” Hoang says. “That changes the percentage and the ratio between economization versus mechanical cooling and water consumption.”

In other words, the more hours you have rejecting heat from dry coolers or radiators, the less energy is consumed.

Maximizing efficiency throughout the data center

Compared to traditional air-cooled data centers, techniques like closed-loop liquid cooling used to cool AI workloads enable data center operators to reduce water consumption by 40-60 percent. This works by changing the set point to a warmer water temperature.

“AI workloads can allow us to raise the temperature much higher than traditional data centers that leverage air-side cooling,” he said. “Liquid-cooled servers actually allow us to raise the temperature so we can consume less energy – and if water is chosen to remove the heat, we can reduce that too.”

Perhaps one of the more efficient ways to reduce water consumption within the data center is closed-loop cooling, a method that recirculates liquid or refrigerant through sealed pipes to absorb server heat.

Hoang explained that these systems work by enabling the use of dry coolers, so the operator has a configuration where coolant isn’t exposed to the atmosphere and water doesn’t need to evaporate.

“You can operate the cooling system dry, even if evaporative cooling is available when needed. It also allows you to use mechanical cooling for trimming,” he said. “When conditions are right, you consume zero water because of this dry mode.”

When temperatures exceed what free cooling can handle, operators can turn on the minimum amount of water needed to evaporate and cool, therefore bringing the temperature down.

“You can also quickly use mechanical cooling if you choose to avoid water consumption altogether,” Hoang added. “Closed loop really allows you to manage and maximize your operation while reducing and mitigating the risk of exposing your critical cooling to the atmosphere.”

Servers can take coolant at 45°C (113°F), much higher than the ambient condition in most locations. It means that the data center only has a few hours of the day during the warmest period of the year when evaporative cooling with water or mechanical cooling has to be used.

“This changes the balance between water usage and mechanical cooling. By raising the temperature, you’re minimizing both – and the economic trade-off between saving energy, versus water consumption, becomes very narrow,” Hoang added.

A leader in the industry, Schneider Electric provides cooling solutions that consider the entire end-to-end system of the data center. Its solutions focus on options for the customer where water consumption can be almost completely avoided, while maximizing and increasing unit efficiency.

“We’re developing high‑efficiency air‑cooled chillers that operate at warmer temperatures, which allows us to be extremely efficient while avoiding water consumption,” Hoang explained. “Operators today can review where they are with their customers, what their commitments are, and then look at the workload.”

He added: “Overcooling can feel safe, but we’re not in an age where we have the luxury to do that,” he said. “It’s about paying attention, having awareness and telemetry in the system so that you can operate at higher design temperatures and set points while still managing your risk.”

Click here to listen to the full DCD>Talks interview with Tuan Hoang.