What is a Chiller?

What Is a Chiller? How It Works and What Chiller-less Means

 

Chillers are widely used in factories, commercial buildings, and data centers that require stable temperature control. In data centers, they form part of the facility cooling system that removes heat generated by IT equipment.

 

As AI servers generate more heat, cooling systems must handle greater loads. This can increase energy consumption and require additional cooling capacity and supporting infrastructure. Data center operators are therefore exploring ways to reduce their dependence on chillers.

 

The article explains what a chiller is, how it works, and how chiller-less cooling can support high-density AI systems. It also introduces CreAlien W2PC as one approach to chiller-less AI cooling.

 

What Is a Chiller?

 

A chiller is a refrigeration machine that removes heat from water or another circulating liquid. The cooled liquid is then delivered to equipment or air-conditioning systems that require temperature control. After absorbing heat, it returns to the chiller to be cooled again.

 

In a data center, chilled water may be supplied to air-handling equipment that cools the server room. It may also support a liquid-cooling system through heat exchangers and facility water loops. In both cases, the chiller keeps the cooling liquid below the temperature that could be achieved through direct heat rejection alone.

 

Most mechanical chillers produce this cooling through a vapor-compression refrigeration cycle. This cycle uses four main components: an evaporator, a compressor, a condenser, and an expansion valve.

 

How Does a Chiller Work?

 

A typical chiller uses two circulating fluids. Chilled water travels between the chiller and the equipment that needs cooling. Refrigerant circulates inside the chiller and transfers the collected heat to the outside environment.

 

The cycle follows four steps:

 

• Evaporation

The refrigerant absorbs heat from the returning water. This cools the water and turns the refrigerant from liquid into vapor.

 

• Compression

The compressor raises the pressure and temperature of the refrigerant vapor.

 

• Condensation

The hot refrigerant releases its heat to outdoor air or cooling water. It then changes back into liquid.

 

• Expansion

The refrigerant passes through an expansion valve, which lowers its pressure and temperature. It can then absorb heat again.

 

The chilled water and refrigerant continue circulating through their separate loops. The chilled water carries heat away from the equipment, and the refrigerant transfers the heat out of the chiller.

 

What Are the Main Types of Chillers?

 

Chillers are commonly classified by how they release heat from the refrigerant at the condenser. The two main types are air-cooled and water-cooled chillers. Both produce chilled water, but they reject heat differently.

 

 

Air-cooled chillers are generally easier to install because they do not require a separate condenser-water loop or cooling tower. However, their efficiency may decrease as outdoor temperatures rise. Water-cooled chillers require additional pumps, piping, and cooling-tower infrastructure. This increases installation and maintenance requirements, but they are often more efficient for large, continuously operated cooling systems. The appropriate choice depends on the cooling load, climate, available space, water supply, and long-term operating requirements.

 

How Are Chillers Used in Data Center Cooling?

 

A chiller usually does not cool servers directly. It supplies chilled water to the room-cooling equipment or facility water loop that carries heat away from the IT equipment. Two common configurations are:

 

• Air-Cooled Servers with Chilled-Water Room Cooling

 

Fans transfer heat from the server components into the room air. A computer room air handler, or CRAH, draws in the warm air and passes it across a coil containing chilled water. The water remains inside the CRAH and does not enter the servers. After absorbing heat from the air, it returns to the chiller to be cooled again.

 

The heat follows this path:

Chips → room air → CRAH coil → chilled water → chiller

 

• Direct-to-Chip Liquid Cooling

 

Coolant flows through cold plates attached to CPUs and GPUs, allowing it to collect heat directly from these components. A coolant distribution unit, or CDU, transfers the heat from the server coolant to the facility water loop. The facility water may then return to a chiller for cooling.

 

The heat follows this path:

Chips → server coolant → CDU → facility water → chiller

 

Chillers allow both configurations to maintain controlled water temperatures when outdoor conditions cannot remove enough heat on their own. However, the compressor and support equipment also consume energy and require additional infrastructure. As AI servers generate more heat, the chiller system must handle a larger cooling load. This increases energy use and may require additional cooling capacity, leading operators to consider whether chiller operation can be reduced under suitable conditions.

 

Why Are AI Data Centers Reconsidering Chiller-Dependent Cooling?

 

Chillers remain useful when a facility needs water temperatures below what outdoor conditions can provide. However, as AI servers increase the cooling load, chiller-dependent systems may create three challenges:

 

• Higher Cooling Energy Consumption

The compressor requires electricity to maintain low chilled-water temperatures. Pumps, condenser fans, and cooling towers may add further energy use, depending on the system. This increases Power Usage Effectiveness, or PUE, which compares the total energy used by a data center with the energy delivered to its IT equipment.

 

• More Supporting Infrastructure

A larger cooling load may require greater chiller capacity, additional electrical supply, and expanded chilled-water piping. Water-cooled systems also require condenser-water pumps, piping, cooling towers, and water treatment. These requirements

 

• More Installation and Maintenance Work

Chiller systems require refrigerant management, compressor maintenance, controls, and regular inspection. Water-cooled systems also require cooling tower cleaning and water quality management. Expanding this infrastructure can increase both deployment time and long-term maintenance requirements.

 

These challenges do not mean that chillers are unsuitable for AI data centers. They show why operators are considering whether some cooling loads can be handled without compressor-based refrigeration.

 

How Can AI Cooling Operate Without a Chiller?

 

A chiller is needed when the cooling system must produce water below the temperature available through direct heat rejection. If the cooling loop can operate at a higher temperature, the collected heat may be released directly to the outdoor environment without compressor-based refrigeration.

 

Liquid cooling makes this more achievable by capturing heat close to CPUs and GPUs. The heated coolant transfers this heat to a dry cooler, cooling tower, or another outdoor heat exchanger. Fans and pumps are still required, but the chiller compressor is removed from the process.

 

The heat path becomes:

Chips → liquid-cooling loop → outdoor heat-rejection equipment → outdoor environment

 

A fully chiller-less design must be able to reject the required heat even during the site’s warmest expected conditions. If the outdoor environment cannot provide enough cooling, the facility may use a chiller temporarily or combine mechanical cooling with free cooling.

 

Chiller-less therefore does not mean that no cooling equipment is required. It means that the system can remove and reject heat without using a compressor-based refrigeration cycle. The operating temperature of the liquid loops and the site’s climate determine whether this is possible throughout the year.

 

How Does CreAlien W2PC Enable Chiller-less Cooling?

 

CreAlien W2PC uses DI water in a sealed, negative-pressure loop to remove heat directly from high-power CPUs and GPUs. Lowering the pressure also lowers the boiling point of water, allowing the DI water to begin changing into vapor at approximately 45ºC.

 

Inside the cold plates, the DI water absorbs heat from the server components and changes into vapor. The vapor travels to the condensation side, where it releases the heat and returns to liquid. The liquid then circulates back to the cold plates, and the process continues.

 

Many cooling systems use a chiller to lower the temperature of the circulating water before it returns to collect more heat. W2PC does not require this refrigeration step. Its DI-water loop transfers heat at a higher operating temperature, allowing the system to release the collected heat through heat exchange without activating a compressor-based chiller.

 

The 45ºC phase-change temperature maintains a usable temperature difference even when the surrounding environment reaches 35ºC. As a result, W2PC can continue releasing heat in hot weather and support year-round chiller-less cooling within its specified operating conditions.

 

Planning a new AI facility or cooling upgrade? CreAlien can help evaluate whether W2PC fits your site conditions and cooling load.