Data Centre Cooling - What are the Solutions?

The image of the data centre industry is transforming. With mass-adoption of AI underway and expansion planned for the sector around the world, concerns on water consumption and wider environmental impact are on the rise.

For part two of Techbuyer’s webinar series on data centres and water consumption, in partnership with the Data Centre Alliance, we brought industry experts together for a panel discussion on the cooling of data centres. We were joined by Matthew Rutherford from Evapco, Jon Summers from RI.SE, and Mark Acton, a Consultant and Technical Advisor for data centre lifecycles.

Why Liquid Cooling Now?

The discussion began with a simple question – Why are we even talking about liquid cooling now?

Liquid cooling isn’t a new concept and has been used to cool data centres for a long time from a heat removal perspective. The problem lies in that data centres and IT equipment are incredibly inefficient – they consume a lot of power and this power then tends to be converted to heat.

Up to a certain point, air cooling is a perfectly good method for heat extraction – for temperatures of between 20-40 kilowatts a cabinet, air cooling is a reasonably effective method for heat removal in data centres. This was highlighted by Jon Summers who stated that: “You can reject a certain amount of heat in a seven‑metre by seven‑metre air duct. The same amount of heat can be rejected in a liquid pipe of about 200 millimetres in diameter.”

However, as density increases and equipment is being moved closer and closer together, it becomes much more difficult to extract heat. This is when liquid cooling can play a much bigger role. As Summers explains, “Tthe reason we’re using liquids is densification. We’re putting more IT closer together, and air just isn’t up to the job.”

What Liquid Cooling Actually Means

Despite the attention it’s received recently, liquid cooling itself isn’t new. Data centres have used liquid - typically chilled water - in mechanical plants for decades. What is new is where the liquid is going.

“When I talk about liquid cooling, I say it’s when liquids invade the IT space,” said Summers. Rear‑door heat exchangers, for example, still rely on air as the primary medium, even though liquid helps remove the heat. “Rear‑door heat exchangers bring liquid into the white space, but air is still the prime mover. Liquid cooling is when liquid breaks that barrier and goes to the IT.”

Direct‑to‑chip cooling represents that shift most clearly, with liquid delivered straight to the processors themselves. As Acton summarised: “We’ve had chilled water loops in data centres for decades. What’s different now is taking the liquid to the chip.”

The Temperature Trade‑Off

One of the most misunderstood aspects of liquid cooling is temperature. Many assume the goal is to drive temperatures down as far as possible, but the reality is more nuanced.

Summers compared temperature to voltage: “If you don’t have a difference, you don’t have movement. Heat moves naturally from high temperature to low temperature.” Operating at higher temperatures can make heat rejection far easier, often without mechanical cooling, improving overall facility efficiency. “If you can operate at higher temperatures, you can reject heat without using extra energy,” he said.

However, this comes with a trade‑off on the IT side. Acton warned that “As chip temperatures increase, the chips themselves run less efficiently and demand more power.” In other words, “Your PUE might look fantastic, but you’re actually consuming more power on the IT side.” Finding the optimal balance between cooling efficiency and compute efficiency is critical.

Why Hybrid Cooling is the Reality

Even in highly liquid‑dense environments, air cooling isn’t going away. “You’re always going to have elements in the data centre that need air cooling - power equipment, networking, cables. You can’t get away from that,” Acton said.

As a result, most real‑world designs are hybrid. “The designs I see typically have 15 to 25 percent air cooling,” Acton noted. “It’s completely hybrid.” Liquid may handle the bulk of the thermal load, but air still plays a vital supporting role.

Water, Energy and the Bigger System

Discussions about liquid cooling often turn quickly to water usage, but the speakers were clear that this is frequently misunderstood. “A chilled water loop is a closed loop,” Acton explained. “You’re not consuming water - you’re just moving heat.”

Water consumption becomes a factor only when evaporative or adiabatic systems are used for heat rejection - usually to reduce electricity consumption. As Matthew Rutherford summed up: “If you use water, you use less electricity. If you don’t use water, you use more electricity.”

Summers framed it as an optimisation problem: “You can’t look at water or energy in isolation.” Geography matters deeply. “A data centre in the Nordics is fundamentally different to one in Southern Europe,” he said, a point Acton reinforced: “There’s nothing wrong with using water - unless you’re in a water‑scarce area.”

Looking Forward

While liquid cooling is now essential, the speakers agreed it is ultimately a response to increasingly inefficient compute. High performance computing (HPC) has relied on liquid cooling for decades, and AI is now following the same path.

Longer‑term, more radical shifts may change the equation entirely. “Electrons are heavy. Photons aren’t,” Summers observed. “If we keep everything in the photonics domain, we won’t need anywhere near the power we use today.”

Until then, liquid cooling is not an optional upgrade - it is the enabling technology that allows modern compute to exist at all.

If you are interested in rewatching this webinar, you can watch the session on the Data Centre Alliance website here.

This webinar is 2nd in a series of three, conducted following the publication of the DCA Data Centre Water Usage Guide in 2025.