Coolant Distribution Units (CDU)

CDU distributes coolant to IT equipment and transfers heat between the IT liquid loop and the facility-side cooling system while maintaining stable temperature, flow, and pressure.

Key functions of a CDU include:

• Supplying coolant to liquid-cooled IT equipment such as cold plates

• Transferring heat from the IT loop to the facility-side cooling loop through a heat exchanger

• Connecting the IT cooling loop with upstream cooling infrastructure such as chillers

• Helping maintain controlled and predictable operating conditions for data center operators

LG approaches the CDU as part of an end-to-end thermal management strategy supported by its HVAC and cooling expertise. Its portfolio includes chillers, air-side cooling, liquid cooling components, and control and monitoring solutions.

As AI servers and GPUs generate higher rack heat densities, data centers are increasingly adopting direct-to-chip liquid cooling to remove heat closer to the source and support more stable thermal management.

Key benefits include:

• Direct heat removal from high-heat components such as CPUs and GPUs

• Reduced dependence on server and room airflow

• Greater cooling capacity for high-density AI environments

• More stable coolant flow and temperature through the CDU

• Low ATD of 4°C (7.2°F) to support efficient heat exchange

• Defined loop conditions and filtration to help protect the cooling system

• Support for scalable hybrid and liquid cooling architectures as data center requirements evolve

LG supports this transition with an integrated portfolio spanning chillers, air-side cooling, liquid cooling components, and control solutions.

CDUs are commonly used in liquid cooling architectures that require controlled coolant distribution and heat transfer between the IT and facility cooling loops.

Common CDU applications include:

• Direct-to-Chip cold plate cooling

• Two-phase Direct-to-Chip architectures in selected system designs

• Immersion cooling configurations where a CDU is integrated with the facility loop

• Hybrid cooling systems combining air and liquid cooling

CDU sizing starts with the required cooling capacity and then considers flow rate, pressure, piping, temperature, redundancy, and future expansion. In hyperscale AI data centers, CDUs are typically deployed as repeatable cooling blocks that can scale with rack density and total IT load.

To size a CDU:

• Estimate the required capacity based on IT load and rack density.

• Define the number of racks supported per CDU and future growth.

• Set the target coolant temperature and operating ΔT.

• Confirm flow rate, pressure drop, pipe size, and pump configuration.

• Plan for redundancy and multi-unit expansion.

LG’s Direct-to-Chip CDU lineup includes 600 kW, 1,000 kW and 2,600kW configurations.

Redundancy helps maintain cooling operation when an individual component fails or requires maintenance, reducing the risk of overheating and system downtime.

Key redundancy features of LG CDUs include:

• N+1 pump and inverter redundancy

• Redundant sensors for continued system monitoring

• AI virtual sensors that estimate backup values when a physical sensor experiences an issue

• Hot-standby inverter operation in multi-CDU configurations

• Fast changeover to help maintain stable flow and temperature

• Service-oriented design and operator monitoring for ongoing system management