Definition
A CRAH (Computer Room Air Handler) is a data center cooling unit that uses chilled water flowing through a cooling coil to remove heat from air and maintain suitable environmental conditions for IT equipment.
Unlike a CRAC (Computer Room Air Conditioner), which typically uses a compressor and refrigerant-based refrigeration cycle, a CRAH relies on a central chilled water system to provide cooling. This architecture allows CRAH units to integrate with centralized chillers and support scalable cooling across larger data center facilities.
How CRAH Works
A typical CRAH cooling process includes four basic steps:
- Warm air from the data center is returned to the CRAH unit.
- Fans move the air across a chilled-water cooling coil.
- Heat transfers from the air into the chilled water.
- Cooled air is supplied back into the data center while the warmed water returns to the chilled water plant.
The CRAH therefore functions as the air-side cooling interface between the data center space and the facility’s chilled water infrastructure.
CRAH in Data Centers
CRAH units are commonly used in data centers that have centralized chilled water infrastructure, including:
- Enterprise data centers
- Colocation data centers
- Hyperscale data centers
- Large server rooms
- High-capacity computing facilities
CRAH systems are particularly suitable where cooling capacity needs to scale across multiple rooms or zones and where centralized cooling infrastructure can provide efficient heat removal.
For higher-density AI and HPC environments, CRAH units may operate alongside other thermal management technologies, including liquid cooling, rear door heat exchangers (RDHx), and direct-to-chip cooling.
CRAH vs. CRAC
| Feature | CRAH | CRAC |
|---|---|---|
| Full Name | Computer Room Air Handler | Computer Room Air Conditioner |
| Primary Cooling Method | Chilled water | Refrigerant / mechanical refrigeration |
| Compressor | Typically no internal compressor | Typically includes a compressor |
| Cooling Source | Central chilled water plant | Integrated or dedicated refrigeration system |
| Typical Deployment | Larger data centers and centralized cooling systems | Smaller or distributed cooling environments |
| Scalability | Well suited to centralized, scalable cooling | Well suited to independent or localized cooling |
The key distinction is the cooling mechanism: CRAH uses chilled water supplied by a central cooling plant, while CRAC typically uses an integrated refrigeration cycle.
Why CRAH Matters
CRAH units provide an important connection between data center air cooling and centralized facility cooling infrastructure.
Their use can support:
- Scalable cooling capacity
- Centralized cooling plant integration
- Efficient airflow management
- Zoned temperature control
- Redundant cooling architectures
As rack power density increases, CRAH systems may also be combined with liquid cooling technologies to create hybrid thermal architectures that match different IT loads.
CRAH and Modern Data Center Cooling
CRAH is primarily an air-based cooling technology. It removes heat from the data center environment by transferring heat from air to chilled water.
For conventional IT loads, CRAH can provide the primary room-level cooling system. For higher-density AI and HPC workloads, however, air cooling may need to be supplemented or replaced by liquid-based technologies depending on rack density and thermal requirements.
This makes CRAH an important component of the broader data center cooling and thermal management ecosystem.
Related Terms
- CRAC (Computer Room Air Conditioner)
- Precision Cooling
- Air Cooling
- Chilled Water System
- Data Center Cooling
- Thermal Management
- Liquid Cooling
- Direct-to-Chip Cooling
- Rear Door Heat Exchanger (RDHx)
- High-Density Computing
- AI Data Center
Related ATTOM Solutions
ATTOM provides data center cooling solutions covering precision air cooling and high-density liquid cooling architectures for modern data center environments.
Explore ATTOM’s Precision Cooling and Liquid Cooling solutions for applications ranging from conventional enterprise infrastructure to high-density AI computing.
