What is free cooling?
What Is Free Cooling for AI Data Centers?
Cooling has become a major energy and infrastructure challenge as AI servers generate more heat within each rack. Free cooling can reduce this burden by using suitable outdoor conditions to reject heat with little or no compressor-based refrigeration.
However, its effectiveness depends on more than outdoor temperature. This article explains how free cooling works, the conditions it requires, and how liquid cooling technologies such as CreAlien’s W2PC can help extend its use in high-density AI data centers.
What Is Free Cooling in a Data Center?
A data center cooling system performs two basic tasks. First, it removes heat from servers and their components. It then transfers that heat through the facility and releases it into the outdoor environment.
Free cooling applies to the second task. It is a method of rejecting data center heat with little or no compressor-based refrigeration. When outdoor conditions are suitable, the facility can use outdoor air or water-side heat-rejection system to remove heat instead of relying entirely on a chiller.
Despite its name, free cooling still consumes electricity. Pumps, fans, heat exchangers, and control systems remain necessary. “Free” refers specifically to reducing or avoiding energy used by refrigeration compressors.
Free cooling also does not always mean that a data center is completely chiller-less. Some facilities use free cooling when outdoor conditions permit and switch to a chiller when they do not. A fully chiller-less design must be able to reject the required heat throughout its intended operating conditions without conventional mechanical refrigeration.
How Does Heat Move Through a Data Center Cooling System?
The two tasks above take place on different sides of the cooling system:
IT-Side Cooling System
IT-side cooling collects heat from CPUs, GPUs, and other server components. Depending on the server design, it may use air cooling or liquid cooling.
Facility-Side Cooling System
Facility-side heat rejection carries the collected heat out of the data center and releases it into the outdoor environment. This is where compressor-based refrigeration, free cooling, or a combination of both is used.
What Are the Main Types of Data Center Free Cooling?
Once heat has been collected from the IT equipment, the facility-side cooling system must release it into the outdoor environment. Free cooling can perform this task in two main approaches: airside free cooling, which uses outdoor air conditions to cool the data center air, and waterside free cooling, which rejects heat from a facility water loop.
Airside Free Cooling
Airside free cooling uses cool outdoor air to remove heat from the data center. Outdoor air may enter the facility directly after filtration, or it may cool indoor recirculated air through an indirect heat exchanger.
This method can reduce air-conditioning energy when outdoor temperature and humidity remain within acceptable limits. However, filtration, humidity control, contamination risks, and the large airflow required by high-density AI servers must still be considered.
Waterside Free Cooling
Waterside free cooling transfers heat through a liquid loop and rejects it outdoors through equipment such as a dry cooler or cooling tower. When outdoor conditions provide a sufficient temperature difference, the system can reduce or bypass compressor-based refrigeration.
Its effectiveness depends partly on the required coolant temperature. A system that can transport heat at a higher water temperature can use outdoor heat rejection under a wider range of conditions. This makes waterside free cooling particularly relevant to high-density AI servers using liquid cooling.
How Does IT-Side Cooling Connect with Free Cooling?
Air cooling and liquid cooling describe how heat is collected from IT equipment. Airside and waterside free cooling describe how the heat is rejected using suitable outdoor conditions. The same IT-side cooling method may connect with different facility-side free-cooling systems.
IT-Side Cooling | Facility-Side Free Cooling | How They Work Together |
Air Cooling | Airside free cooling | Cool outdoor air either enters the facility directly or removes heat from recirculated data center air through a heat exchanger that keeps the indoor and outdoor air streams separate. |
Air Cooling | Waterside free cooling | Room air transfers heat to a facility water loop, which then carries it to outdoor heat-rejection equipment. |
Liquid Cooling | Waterside free cooling | A liquid loop collects heat from IT components and transfers it to a facility water loop for outdoor heat rejection. |
Liquid Cooling | Airside free cooling | This is not a typical direct pairing because the heat from the IT equipment has already been collected by a liquid loop. |
How Does Liquid Cooling Support Waterside Free Cooling?
Liquid cooling collects heat close to CPUs, GPUs, and other high-power components. The heated coolant then carries this energy toward the facility water system, which ultimately releases it through outdoor heat-rejection equipment.
For waterside free cooling to work, the facility water must be sufficiently warmer than the outdoor environment for heat to flow outdoors. When an IT-side liquid cooling system can transport heat at a higher temperature, the facility water can also operate at a higher temperature. This allows equipment such as a dry cooler to reject heat without compressor-based refrigeration under a wide range of outdoor temperatures.
The required water temperature is therefore an important link between liquid cooling and waterside free cooling. A system requiring low-temperature coolant may depend on a chiller during warmer weather, while a system designed for higher coolant temperatures can extend free-cooling operation into warmer conditions.
How Does W2PC Achieve Year-Round Free Cooling?
W2PC circulates DI water through a vacuum negative-pressure loop. The reduced pressure lowers the boiling point of water to 45ºC+. As the water absorbs heat from CPUs and GPUs, it changes into vapor and carries the heat to a condenser. The vapor then releases the heat, returns to liquid, and recirculates through the system.
Because the heat is transported at a relatively high temperature, it can be transferred to the external heat-rejection system without first producing low-temperature chilled water. This allows W2PC to reject heat without operating a compressor-based chiller.
The 45ºC boiling-point design also supports free cooling during hot weather. Even when the ambient temperature reaches 35ºC+, the system retains the temperature difference required for external heat exchange. W2PC can therefore continue operating without a chiller during summer, enabling year-round free cooling within its specified ambient conditions.
What Are the Benefits of Free Cooling for AI Data Centers?
By removing compressor-based refrigeration, free cooling can affect both facility energy use and the economics of AI deployment. Its three main benefits are lower cooling energy, less cooling infrastructure, and more facility power available for computing.
Lower Cooling Energy
Chiller compressors consume substantial electricity to produce low-temperature cooling water. Free cooling avoids this refrigeration process and uses lower-energy equipment such as pumps, fans, and heat exchangers to transfer heat outdoors.
By combining two-phase heat transfer with compressor-free cooling, CreAlien’s W2PC can reduce cooling-system energy consumption by up to 97%.
Lower Infrastructure Costs and TCO
Conventional chiller-based cooling may require chillers, cooling towers, mechanical rooms, chilled-water piping, and related control systems. Free cooling can reduce chiller operating hours, but this infrastructure remains necessary if mechanical cooling is still required under hot conditions.
By supporting year-round free cooling at ambient temperatures of 35ºC+, CreAlien’s W2PC removes the need for conventional chillers and cooling towers. This reduces equipment investment, cooling energy, and maintenance requirements, lowering the data center’s total cost of ownership.
Lower PUE and More Power Available for AI Computing
Power Usage Effectiveness, or PUE, compares the total energy consumed by a data center with the energy delivered to its IT equipment. A lower PUE means that less facility power is used by cooling, power distribution, and other supporting systems.
This difference becomes particularly important when a facility has limited power capacity. For example, a data center with 1MW of available power can deliver approximately 667kW to IT equipment at a PUE of 1.5. At a PUE of 1.1, approximately 909kW can support servers within the same facility power limit.
By reducing cooling overhead, CreAlien’s W2PC can support a PUE below 1.1 and leave more of the available electricity for CPUs, GPUs, and other AI hardware. This allows operators to deploy more computing capacity and generate more AI output without increasing the facility’s total power supply.
Plan Your AI Cooling Infrastructure with CreAlien
CreAlien’s W2PC extends free cooling beyond cool-weather operation. Its 45ºC+ DI-water phase-change design supports chiller-less heat rejection at ambient temperatures of 35ºC+, helping AI data centers reduce cooling energy and infrastructure requirements.
Contact CreAlien to discuss your AI cooling requirements.