With thermal management such a critical part of modern operations, the effectiveness of cooling technology is being examined more closely than ever.

Where the boundaries between the chiller plant, the cooling distribution system, and the rack are becoming increasingly blurred, what were once treated as separate layers of data center infrastructure are now being designed as an interconnected cooling ecosystem.

This shift highlights the importance of understanding how proven and emerging technologies can be integrated to meet the demands of next-generation computing.

To explore this perspective, a recent DCD>Broadcast saw three industry experts examine cooling from different points in the stack.

Housing compute equipment directly integral to revenue generation, Bob Hicks, chief revenue officer at DDC Solutions, sees advanced rack containment enabling higher-density cooling while isolating IT assets, improving the efficient use of cooling resources and fundamental operational resilience properties.

When it comes to managing the actual thermal methodology, Dr. Steve Harrington, founder and CTO of Chilldyne, shares how the company’s CDU technology reduces risk, while enabling efficient cold plate cooling directly at the server level.

Where both of these equipment classes are necessary in capturing and transferring heat locally, they do not neutralize or expel it. They rely heavily on the chiller plant to function as the ultimate heat rejection endpoint, required for continuous cooling.

Here, Greg Jeffers, vice president of data center solutions at Daikin Applied, explains how the company’s decades of expertise in HVAC and refrigeration evolved its thermal management technologies to address the challenges of today’s high-density data centers.

Representing the cabinet, CDU, and chiller layers respectively, these three industry leaders have combined expertise to offer a holistic solution for modern data center cooling – demonstrating how solutions from chiller to chip can work together as a coordinated system rather than as isolated technologies.

A different set of rules

For years, data center cooling has been divided into separate disciplines, with solutions only integrated at the point of deployment. As AI workloads drive rack densities into the hundreds of kilowatts, that sequential approach is increasingly difficult to sustain.

“Historically, the industry looked at data center design in layers – at the chiller plant, the data hall, the white space, and oftentimes those have been engineered independently,” says Jeffers, adding:

“That worked at those lower densities, but now that we’re up to very high densities driven by AI and high-performance compute, those decisions need to be made much further in advance so you can optimize the entire system.”

The last 18 months have seen immense uptake of advanced cooling solutions, but this has introduced challenges for data centers that weren’t designed to handle that degree of cooling.

Not only do data center owners require enhanced support for unfamiliar technologies, but they are grappling with the physical preservation of high-value equipment when introducing the likes of direct-to-chip liquid cooling.

“We’ve taken that to a different level from an asset protection perspective because racks of today are stacked with millions of dollars worth of servers – you now have a concentrated location of an extremely expensive stack of IT gear, which adds a different dynamic and challenge that we address with a rack containment solution,” says Hicks.

Beyond just a high-performance vendor, Harrington reflects on another shift in customer expectations to handle their high-value investments:

“One of the most important considerations is that they want a liquid cooling vendor they can trust. Someone that will provide support, because they don’t want someone to just come by, drop off the CDU, leave and never talk to them again.”

He explains that in many cases, customers are not too familiar with liquid cooling and don’t necessarily know what issues they will come across, or how to recover quickly and effectively in the event something goes wrong.

Data center cooling now needs to be considered as an ecosystem decision, fundamentally changing design approaches. By combining the management and servicing of these three entities, customers are provided with a single-stop solution.

“The Chilldyne CDU addresses the liquid right to the chip. The DDC containment solution is a rack enclosure that acts as a protection play because of its isolation. Both the rack and the CDU require chiller support from Daikin. So bringing it all together is just a really smooth solution,” explains Hicks.

The thermal stack and its environment

The concept of the thermal stack is becoming an increasingly useful way of describing modern data center cooling, capturing the complete thermal pathway.

Designing across the entire stack allows operators to optimize the individual components and the interfaces between them. In turn, matching the performance characteristics of each layer can significantly improve efficiency and simplify operation.

“Chiller plants like to work at a delta T, a temperature differential of around 10°C (50°F),” explains Harrington. “If the server works around the same temperature differential, then they work pretty well together. That’s just one example of something you want to think about when designing your liquid cooling system, to make sure the chiller and the server are both expecting the same temperature differential.”

Beyond the cooling equipment itself, the physical environment surrounding the IT infrastructure plays an important role in protecting increasingly valuable compute assets.

“Racks were originally designed to address density, but the byproduct as the IT assets grow in value is containment as an asset protection envelope – isolating the equipment from dust, unexpected water discharge, and fire – solving a problem regardless of the cooling benefits,” says Hicks.

By combining rack containment with liquid cooling technologies, operators can address both thermal performance and physical resilience within the same architecture. When cooling, environmental protection, and operational continuity are treated as complementary elements of a single infrastructure strategy, operators can receive a greater, more seamless level of support from their infrastructure partners.

A realistic retrofit path

For many operators, building a new AI-ready facility is neither practical nor necessary. Instead, the focus is shifting toward retrofit strategies that allow higher-density infrastructure to be introduced alongside existing deployments, without disrupting the wider data hall.

Modular, rack-based solutions are making that possible. By combining containment with integrated cooling, operators can create high-density environments within legacy facilities, while allowing conventional racks to continue operating unchanged.

“It’s not as if you have to have an entire dedicated data hall. We can just fit right in,” says Hicks, adding:

“All of a sudden our racks can do 100kW per rack, while the traditional racks around it are tapping out at ten or 15kW per rack. We don’t affect them. They don’t affect us. We don’t affect the room. We’re very neutral in that regard.”

The approach also changes the economics of expansion. Rather than investing upfront in an entirely new white space, operators can deploy prefabricated infrastructure incrementally, making use of available floor space without relying on raised floors or perimeter cooling.

“You can build your way across that room as opposed to having to build the whole room and consume it in little bits at a time,” explains Hicks.

For organizations taking their first steps toward high-density or liquid-cooled infrastructure, this phased approach also reduces design risk. It allows operators to gain operational experience while scaling capacity in line with demand, rather than committing to a full facility transformation from day one.

Harrington believes the same philosophy should extend to the underlying infrastructure. Even where liquid cooling is not an immediate requirement, designing for future capacity – through larger cooling water pipes, additional electricity provision and reserved floor space – can significantly simplify the next stage of a facility’s evolution.

Bringing liquid to the rack

As we see rack power densities continue to rise, the rack itself has become the focal point for data center design. It’s where IT requirements meet physical infrastructure, dictating everything from electrical distribution to cooling capacity – and increasingly, the choice between air and liquid cooling.

“From a design standpoint, as you build data centers, everything is really driven by what is being contained in the rack and the types of components and the power densities that are contained there. That drives the amount of power you need, the amount of cooling you need, et cetera,” explains Jeffers.

For many AI deployments, liquid cooling is now a practical necessity. As processor power continually rises, traditional air cooling alone becomes increasingly difficult to scale, making direct-to-chip liquid cooling an essential part of the thermal strategy.

“You can have the proportion of uptime you expected from air-cooled solutions, but with liquid-cooled systems,” says Harrington.

Introducing liquid into the rack, however, inevitably raises concerns about reliability. With AI servers representing millions of dollars in hardware, operators want assurance that cooling infrastructure will not become a source of operational risk.

To address those concerns, Chilldyne employs a negative-pressure liquid cooling system. Rather than pushing coolant through the circuit under positive pressure, the system operates under vacuum, meaning that if a leak develops, air is drawn into the system instead of coolant escaping into the rack. The approach is designed to minimize the risk of leaks throughout the life of the installation, especially as components age or where condensation could otherwise become a concern.

Rack containment provides an additional layer of protection. Traditional open-frame racks rely on unrestricted airflow, allowing air – and potentially moisture or contaminants – to move freely through the IT equipment. By enclosing the IT environment, containment helps isolate valuable hardware from external hazards while creating a controlled environment for both air and liquid cooling.

“If you’re not using a Chilldyne negative-pressure solution, that water could spray and could invade other racks. Our containment would isolate that and eliminate that risk,” says Hicks.

The two technologies have also been designed to work together. DDC’s cabinets accommodate the manifolds that distribute coolant to individual servers, integrating the liquid cooling infrastructure safely within the rack enclosure.

Together, it is clear that successful liquid cooling is about more than simply replacing air with water. It requires the rack, the cooling distribution system and the containment strategy to be engineered as complementary parts of the same infrastructure.

Future-proofing the next generation of data centers

The future of data center cooling is unlikely to be a single transition point from air to liquid, but rather a phased evolution that allows operators to adapt as workloads and technologies change.

“Gone are the days when you could just cherry-pick equipment and expect it to work. Now it really has to be a full design – a full ecosystem decision for everything to work as expected,” says Jeffers.

For many operators, the challenge is knowing when and how to make that investment. With cooling technologies evolving rapidly, flexibility has become a key design requirement. Modular, rack-based approaches allow facilities to begin with the capacity they need today while creating a pathway for future upgrades, avoiding the risk of being locked into a single cooling architecture.

Similar to retrofit methodology, future-proofing becomes instantly more feasible when a rack designed as a self-contained, adaptable environment can support this transition. Operators can deploy the infrastructure initially without fully populating its capacity, then progressively introduce higher-density systems as requirements develop.

That flexibility is central to the ecosystem approach. By bringing together rack containment, cooling distribution, and facility-level thermal management, DDC, Daikin, and Chilldyne aim to provide operators with a coordinated pathway from existing infrastructure to next-generation cooling.

Ultimately, the advice from the experts is to not wait for the perfect moment or the final technology standard. Building practical experience today is the most effective way to prepare for future requirements.

“Let’s deploy a handful of racks. Nothing major – get yourself situated, get the piping schemes and the approach down, and then work your way into it,” says Hicks.

For operators navigating the shift toward AI-ready infrastructure, future-proofing is less about predicting exactly what comes next and more about creating a cooling strategy flexible enough to evolve with it.

Watch the full DCD>Broadcast here.