In-Rack CDU: What a CDU Does and Why It Matters | 2026

As next-generation Artificial Intelligence (AI), Machine Learning (ML), and High-Performance Computing (HPC) workloads accelerate, traditional air cooling reaches its thermodynamic limit. With high-density accelerator architectures pushing rack power budgets beyond 30 kW toward 100 kW+ (such as the NVIDIA Blackwell B200 and H200 clusters), liquid cooling has transitioned from an optional enhancement to an operational necessity.

At the center of this infrastructure shift is the In-Rack Coolant Distribution Unit (CDU). This article delivers a deep technical analysis of how in-rack CDUs manage heat exchange, maintain loop isolation, and lower Power Usage Effectiveness (PUE) within modern enterprise and colocation data centers.

What Is an In-Rack CDU?

An in-rack CDU is a localized fluid-management and thermal-transfer system deployed directly within a standard 19-inch or 21-inch Open Rack (ORv3) architecture. Unlike centralized room-level or in-row CDUs, an in-rack variant offers dedicated, granular thermal management for a single high-density enclosure.

The primary engineering purpose of a CDU is hydraulic isolation, separating the data center into two distinct thermodynamic loops:

The Primary Loop (Facility Water System - FWS): Connects the CDU’s internal heat exchanger to the facility's centralized chilled water plant, dry coolers, or cooling towers. This loop typically uses industrial-grade water or water-glycol mixtures.

The Secondary Loop (Technology Cooling System - TCS): A closed, ultra-pure circuit that circulates specialized coolant directly through server-level cold plates, direct-to-chip (D2C) blocks, or rear-door heat exchangers (RDHx).

By decoupling these two loops via a high-efficiency internal Plate Heat Exchanger (PHE), the CDU protects sensitive IT assets from the high pressures, chemical contaminants, and debris common in facility water networks.

Technical Operational Mechanism: How the CDU Controls Fluid Dynamics

The operational cycle of an in-rack CDU balances precise thermodynamics with fluid mechanics to maximize heat transfer efficiency.

In an In-Rack CDU (Coolant Distribution Unit) architecture, cooling is achieved through two isolated fluid loops connected by a plate heat exchanger (PHE).

Phase 1: Heat Absorption at the Silicon Level

Warm dielectric fluid or treated water-glycol is evacuated from the server cold plates at elevated temperatures. The fluid flows through structured manifolds integrated into the rack frame, converging at the return manifold of the in-rack CDU.

Phase 2: Heat Exchange and Thermal Mitigation

The heated secondary fluid enters the primary Plate Heat Exchanger (PHE). Inside the PHE, a series of corrugated stainless-steel plates create a large surface area for counter-current fluid flow. Heat is transferred across the plates from the secondary TCS loop to the cooler primary FWS loop without any physical mixing of the fluids.

Phase 3: Secondary Fluid Reconditioning

Once cooled, the secondary fluid is pressurized by the CDU’s internal Variable-Frequency Drive (VFD) pumps. It passes through a high-efficiency particulate filter (typically 5 to 50 microns) before being pumped back into the rack's supply manifold at the precise target inlet temperature and flow rate required by the GPUs or CPUs.

How a CDU Cooling Distribution Unit Works

A CDU cooling distribution unit acts as the interface between the facility’s cooling loop and the IT equipment’s liquid loop. Key operational functions include:

Coolant Distribution: Supplies chilled liquid to cold plates, rear-door heat exchangers, or immersion systems

Heat Collection: Returns warm coolant to the facility loop for heat rejection

Flow Regulation: Maintains stable pressure and flow rates

Filtration: Removes particulates to protect sensitive server components

Monitoring & Control: Tracks temperature, pressure, and flow in real time

By placing the CDU at the rack level, heat is removed closer to the source, improving efficiency and responsiveness.

Key Procurement and Design Selection Criteria

When evaluating in-rack CDUs for AI clusters or enterprise infrastructure, design engineers must focus on several critical parameters:

Risk Mitigation: Leak Detection and Fluid Control Systems

Introducing fluids into dense compute environments requires a multi-layered risk mitigation strategy managed directly by the CDU's smart control framework:

Continuous Leak Detection Strings: Multi-zone sensing cables deployed at the base of the CDUS and along the bottom internal rim of the server rack trigger instant alerts and automatic isolation sequences if moisture is detected.

Differential Pressure Monitoring : The CDU constantly calculates the pressure differential between the supply and return manifolds. A sudden drop in signals a downstream pressure loss or line rupture, initiating automatic pump shutdown and closing motorized isolation valves. 

Expansion Tank Integration: Closed secondary loops utilize internal or external expansion bladders to accommodate volumetric thermal expansion of fluids as server workloads swing from idle to peak utilization.

Key Benefits of an In-Rack CDU

1. Supports Higher Rack Density

Liquid cooling managed by an in-rack CDU enables racks to exceed the limits of air cooling, often supporting 30 kW, 60 kW, or more per rack.

2. Improved Energy Efficiency

Shorter coolant paths and reduced reliance on high-speed fans lower power consumption and improve overall PUE.

3. Modular and Scalable Design

Each rack can operate independently, making it easier to scale deployments or mix air- and liquid-cooled systems.

4. Faster Deployment

Pre-integrated rack and CDU configurations reduce on-site plumbing complexity and installation time.

5. Enhanced Thermal Control

Localized cooling minimizes hotspots and ensures consistent inlet temperatures across servers.

Common Applications for In-Rack CDUs

In-rack CDUs are widely used in:

High-performance computing (HPC) clusters

AI and machine learning infrastructure

Colocation data centers

Edge and modular data centers

Enterprise environments with space constraints

Any scenario with high thermal loads and limited floor space can benefit from a CDU cooling distribution unit.

Maintenance and Lifecycle Optimization
To guarantee continuous operation and uphold service level agreements (SLAs), operational teams should implement the following maintenance procedures:

Water Chemistry Analysis: Routinely sample the secondary loop fluid to monitor pH levels, biocides, and corrosion inhibitor concentration to prevent biological fouling and scale formation on micro-channel cold plates.

Filter Elements Refinement: Inspect and replace internal mechanical filters systematically during active operations using parallel dual-filter designs that permit hot-swapping.

Flow-Rate Optimization: Calibrate the VFD pump algorithms against the server cluster’s dynamic telemetry. Pumping fluids at flow rates higher than recommended can cause fluid erosion inside copper cold plates, while under-pumping leads to localized core thermal throttling.

Managing Risks with Liquid Cooling

While liquid cooling is highly effective, it introduces new considerations such as leak prevention and fluid compatibility. These risks are mitigated through:

Integrated leak detection systems

High-quality quick-connect fittings

Redundant pumping and monitoring

Proper commissioning and maintenance procedures

With the right design, liquid cooling with an in-rack CDU is both safe and dependable.

Conclusion

In-Rack Coolant Distribution Units represent the cornerstone of modern thermal management for next-generation data centers. By placing the liquid-to-liquid heat exchanger immediately adjacent to the high-density compute assets, in-rack CDUs ensure maximum heat transfer efficiency, micro-second response times to workload spikes, and excellent overall facility PUE.

As hardware power envelopes continue to scale aggressively, the implementation of highly optimized, redundant in-rack CDUs will remain mandatory for sustainable, reliable AI infrastructure deployment.

FAQ About In-Rack CDU

What is an in-rack CDU and how does it work in a data center?

An in-rack CDU (Cooling Distribution Unit) is a liquid cooling system installed directly inside a server rack to manage heat generated by high-density IT equipment. It circulates coolant through cold plates or rear door heat exchangers, absorbing heat from servers and transferring it to the facility’s cooling loop. By cooling equipment at the rack level, in-rack CDUs improve thermal efficiency, reduce energy consumption, and support higher power densities compared to traditional air cooling.

What does a CDU do in liquid cooling systems?

A CDU regulates, monitors, and distributes coolant within a liquid-cooled IT environment. Its core functions include controlling flow rate, maintaining proper pressure, managing coolant temperature, and providing filtration and leak detection. Acting as the interface between the facility’s cooling infrastructure and the IT equipment, a CDU ensures stable, safe, and efficient thermal performance for modern data centers.

When should you use a CDU cooling distribution unit instead of air cooling?

A CDU cooling distribution unit is recommended when server racks exceed the cooling limits of traditional air systems—typically in high-performance computing (HPC), AI, or GPU-dense environments. Liquid cooling with a CDU becomes essential when rack power densities rise above 20–30 kW, where air cooling becomes inefficient, costly, or unable to maintain reliable operating temperatures.