Data center cooling: liquid cooling, hybrid systems, and new opportunities for refrigeration

The growth of high-density data centers , driven particularly by artificial intelligence and high-performance computing, is rapidly changing the cooling needs of digital infrastructures. Thermal management is no longer an ancillary building service: it is becoming one of the elements that directly influence the facility's computing capacity, operational continuity, and future expansion.

For the HVAC-R sector, this means dealing with more complex architectures, in which chillers, hydronic circuits, liquid cooling, control systems and air cooling must work in a coordinated manner.

At the same time, there is a growing demand for professionals capable not only of installing equipment, but also of commissioning, regulating, diagnosing and maintaining increasingly complex systems .

More power in the racks means bringing cooling closer to the chips

Increasing the computational density first changes the point at which heat must be removed.

In traditional data centers, much of the cooling is done by air, which absorbs the heat produced by the IT equipment and then transfers it to the refrigeration system.

However, with racks featuring higher thermal densities, this approach may become insufficient or less efficient.

Hence the growing interest in liquid cooling , which allows the cooling fluid to be brought much closer to the electronic components. Direct-to-chip configurations, for example, use dedicated circuits to extract heat directly from processors and other high-dissipation components.

The main advantage comes from the thermal properties of liquids, which allow them to transfer large amounts of heat using smaller flow rates and volumes than air.

This doesn't mean, however, that air cooling is destined to disappear.

The direction is towards hybrid thermal architectures

One of the most interesting trends is in fact that of hybrid systems , in which air, water and dedicated liquid circuits coexist within the same infrastructure.

Chillers continue to play a key role in managing the overall heat load, while liquid cooling primarily intervenes in the areas of highest density, bringing cooling closer to the servers.

A data center can therefore integrate:

• chiller for refrigeration production;
• primary and secondary hydronic circuits;
• direct-to-chip liquid cooling systems;
• CDU, Cooling Distribution Unit;
• air handling units;
• dry cooler;
• free cooling;
• centralized control systems.

Design thus becomes less and less focused on the single machine and more and more on the coordinated management of the entire thermal chain .

A change in IT load can in fact rapidly propagate from the rack to the main refrigeration circuit, making it necessary to adapt flow rates, temperatures, pumps, compressors and dissipation systems.

Chiller and CDU become part of the same ecosystem

In liquid cooling, the Cooling Distribution Unit , or CDU, plays a particularly important role.

The CDU represents the interface between the data center's cooling circuit and the one that directly reaches the IT equipment. It must manage parameters such as flow rate, temperature, and pressure, while keeping the different circuits separate.

For refrigeration professionals, this evolution brings the world of data centers closer to the typical skills of industrial refrigeration and hydronic systems .

Heat exchangers, pumps, valves, temperature control, redundancy, and flow management become central elements in ensuring reliability.

The main difference is the continuity requirements: a cooling interruption can quickly compromise the functioning of the IT infrastructure.

For this reason, commissioning and preventive maintenance also take on a particularly high importance.

Modular systems to accommodate data center growth

Another trend concerns modularity.

AI-dedicated infrastructure can scale rapidly, but sizing the refrigeration system for maximum future load from the outset can lead to high investments and oversized equipment in the early stages of operation.

The solution could be represented by modular cooling plants , partly built in the factory and progressively added as the installed IT power increases.

This approach allows you to:

• progressively increase the cooling capacity;
• reduce the work carried out directly on site;
• standardize part of the installation;
• speed up the implementation;
• avoid immediately installing unused capacity.

Modularity therefore concerns not only servers, but increasingly also the thermal infrastructure that supports them.

Energy and water become design constraints

However, the growth of data centers encounters significant physical limits.

Among the main ones are the availability of electricity, water, space, and local infrastructure capacity . In the coming years, these factors could increasingly influence the choice of sites and the cooling technologies used.

An evaporative system can offer energy benefits under certain climate conditions, but requires the availability of water.

A completely dry solution reduces water consumption, but may present more challenging operating conditions during high outside temperatures.

Free cooling is also highly dependent on the climate and the operating temperatures acceptable to the IT system.

Design must therefore find a balance between:

• PUE and energy efficiency ;
• water consumption;
• external design temperature;
• continuity of service;
• available surface area;
• energy cost;
• possibility of future expansion.

Data center cooling thus becomes an energy and infrastructure problem as well as a refrigerator problem.

Intelligent controls and diagnostics become indispensable

The amount of data available in the plants also allows for the development of more advanced control systems.

Temperatures, flow rates, pressures, power consumption, and operating conditions can be analyzed in real time to identify inefficiencies or deviations from expected operation.

Technologies developed for high-density environments are enabling the use of real-time diagnostics, predictive optimization, and smart controls —tools that may later find application in traditional commercial buildings.

The goal is to reduce the margin between available cooling capacity and the capacity actually required, avoiding keeping equipment operating at levels higher than required.

In data centers, even small percentage improvements can become significant, considering the high installed power and continuous operation.

The skills required of technicians are also growing

More technology does not necessarily mean less maintenance.

The increased complexity of systems is increasing the need for specialized technicians capable of working on refrigeration equipment, hydronics, electronic controls, and liquid cooling.

In addition to traditional HVAC-R skills, knowledge relating to:

• control systems;
• liquid circuits for servers;
• commissioning;
• regulation;
• digital diagnostics;
• redundancy;
• typical data center operating procedures.

This opens up a market for refrigeration engineers and service companies that not only involves the construction of new infrastructure, but also the maintenance, updating and expansion of existing systems .

Data center cooling could therefore become one of the areas in which professional refrigeration will evolve most in the coming years: not simply by increasing the available cooling power, but by developing more modular systems, close to the load, connected and capable of rapidly adapting to growth in computing capacity .