Engineered Solutions for Modern Data Center Infrastructure
Reliable Performance for Next-Generation Compute Workloads
- Home
- Electronics Assembly
- Data Centers
- Direct-to-Chip Liquid Cooling for Data Centers
Direct-to-Chip Liquid Cooling for Data Centers
The rapid growth of artificial intelligence (AI), Internet of Things (IoT), and high-performance computing workloads is driving a steady increase in power density in modern data centers. As server performance scales to support these workloads, power consumption rises and heat generation increases at both the rack level and, more critically, at the processor and accelerator level. Today, most data centers rely on air-based datacom equipment cooling systems (DECS) as the primary means of thermal management.
While air cooling remains widely deployed due to its simplicity, maturity, and compatibility with existing data center infrastructure, its practical limits become increasingly evident as rack-level and component-level heat flux continue to rise. These constraints define where air cooling is most effective today, and where alternative cooling architectures begin to emerge.
~80% Market Share
Approximately 70–80% of data centers worldwide continue to rely on air cooling as their primary thermal management approach.
Low Power Density Applications
Air cooling is most commonly used in enterprise, colocation, edge, and legacy data center environments operating at lower rack power densities.
20–35 kW per Rack
Typical operation remains below ~20 kW per rack, with advanced airflow management and containment extending practical limits to ~30–35 kW.
Today’s high-performance server racks increasingly exceed 20, 30, and even 50 kW. As rack densities rise and heat becomes more concentrated at the processor and accelerator level, the limitations of air-based cooling become more pronounced. In response, data center operators are adopting liquid cooling to manage higher heat flux more efficiently. This shift complements, rather than replaces, air cooling and is commonly implemented through three primary architectures: direct-to-chip cooling, immersion cooling, and rear-door heat exchangers.
Direct-to-Chip Cooling
Direct-to-chip cooling removes heat by bringing liquid coolant into direct contact with cold plates mounted on high-power processors, accelerators, and GPUs. By extracting heat at the source, D2C cooling significantly reduces thermal resistance compared to air-based approaches.
This architecture is widely adopted in high-density data centers and represents the dominant liquid cooling deployment today, particularly for AI training, HPC, and advanced accelerator platforms where rack power densities exceed the practical limits of air cooling.
Immersion Cooling
Immersion cooling removes heat by submerging servers or individual components in a dielectric liquid. Heat is absorbed uniformly across exposed surfaces, eliminating the need for traditional heat sinks and server fans.
Systems are designed to operate entirely within the liquid environment, offering high heat flux capability and simplified airflow management, but requiring purpose-built hardware and fluid compatibility considerations.
Rear-Door Heat Exchangers (RDHx)
Rear-door heat exchangers remove heat from server exhaust air at the rack level using liquid-cooled heat exchangers mounted on the rear of racks. Heat is transferred from hot air to a liquid loop while servers remain air-cooled internally.
RDHx systems serve as a hybrid approach, extending the viability of air-cooled servers by reducing room heat load, but do not address chip-level thermal bottlenecks directly.
How Direct-to-Chip Cooling Works
In a direct-to-chip (D2C) cooling system, heat is removed through a closed-loop liquid path that begins at the processor and ends at the facility cooling system. By extracting heat directly at the chip level, this approach minimizes thermal resistance and enables significantly higher power densities compared to air-based cooling.
Step 1: Heat Capture at the Processor
Modern CPUs and GPUs generate high heat loads concentrated at the chip surface due to increasing power density. Heat flows from the processor, through a thermal interface material, and into a liquid-cooled cold plate mounted directly on the CPU or GPU.
Step 2: Heat Absorption and Transport
Coolant flows through channels inside the cold plate and absorbs heat directly from the processor. Heat is absorbed by the coolant or heat transfer fluid.
In single-phase D2C cooling, the coolant remains in liquid form as it absorbs heat, and its temperature rises.
In two-phase D2C cooling, the coolant absorbs heat by partially changing phase (from liquid to vapor), allowing large amounts of heat to be removed at nearly constant temperature.
Step 3: Heat Rejection to Facility Cooling
The warmed liquid (or liquid–vapor mixture in two-phase systems) leaves the cold plate and flows through the direct-to-chip cooling loop. Flow is controlled using pumps, manifolds, and valves. Heat carried by the coolant is transferred to the facility cooling system through a heat exchanger.
In single-phase systems, heat is rejected as the liquid is cooled back down. In two-phase systems, vapor condenses back into liquid during heat rejection before being recirculated.
Main Direct-to-Chip Liquid Cooling Components
At the center of direct-to-chip liquid cooling is the heat transfer fluid, or coolant, which governs how effectively heat is managed once it leaves the processor. Beyond enabling heat removal at the cold plate, coolant properties determine thermal performance, materials compatibiity, and long-term reliability across the entire liquid cooling loop.
Key Heat Transfer Fluid Requirements for Direct-to-Chip Cooling
The primary function of a heat transfer fluid is to efficiently remove heat from high-power processors. Key thermal properties include high specific heat capacity to absorb large heat loads, sufficient thermal conductivity to minimize temperature gradients, and stable performance across the operating temperature range of the cooling loop.
These properties directly influence junction temperature, temperature uniformity across cold plates, and overall cooling efficiency at elevated rack power densities.
Viscosity and flow behavior determine how easily the coolant can be circulated through cold plates, manifolds, and heat exchangers. Fluids with excessively high viscosity increase pumping power requirements and system energy consumption.
Optimized flow properties are essential for maintaining uniform coolant distribution, minimizing pressure drop, and enabling scalable cooling architectures as system complexity increases.
Heat transfer fluids must be chemically compatible with a wide range of materials, including metals, elastomers, polymers, seals, and thermal interface materials used in direct-to-chip cooling systems.
Incompatible fluids can lead to corrosion, swelling, leaching, or degradation of system components, ultimately compromising reliability and increasing maintenance requirements.
Direct-to-chip cooling fluids must maintain stable physical and chemical properties over extended operating lifetimes. Resistance to oxidation, thermal degradation, and contamination is critical for long-term system performance.
Stable fluids reduce the risk of fouling, particulate formation, and performance drift, supporting predictable operation and minimizing downtime in mission-critical data center environments.
Heat Transfer Fluids for Direct-to-Chip Cooling
DOWFROST™ LC 25
DOWFROST™ LC 25 is a ready-to-use 25 vol% propylene glycol (PG25) heat transfer fluid for HVAC, data-center cooling loops, process cooling, and direct-to-chip liquid cooling systems. The fluid contains a corrosion-inhibitor package designed to slow glycol oxidation, control pH, and protect common loop materials, including copper, brass, carbon steel, cast iron, and aluminum.
DOWFROST™ LC 25 is formulated using Dow PURAGUARD™ USP/EP-grade propylene glycol (>99.8% purity), reducing contaminants that contribute to odor, discoloration, and accelerated fluid degradation. High base-fluid purity improves chemical stability beyond what inhibitors alone can provide. The fluid is dyed fluorescent yellow-green for leak detection. Gradual color darkening may occur during service and is acceptable if the fluid remains clear and free of suspended solids.
Corrosion Protection Performance
Visual comparison of metal corrosion after exposure to water, uninhibited propylene glycol, and DOWFROST™ LC 25. Inhibited formulation provides significantly improved corrosion protection.
Typical Applications
DOWFROST™ LC 25 is designed for closed-loop liquid cooling systems requiring long-term thermal stability, corrosion protection, and low maintenance.
- Datacom and telecom equipment cooling loops
- Direct-to-chip liquid cooling for high-performance processors
- Data center thermal management and facility cooling systems
Suitable for systems where fluid cleanliness, material compatibility, and long-term corrosion control are critical
Typical Properties (Engineering Reference)
Key thermophysical and chemical properties relevant to direct-to-chip and data center liquid cooling system design.
Thermal Performance
Thermal conductivity @ 50 °C:
0.485 W/m·K
Specific heat @ 50 °C:
4.13 kJ/kg·K
Boiling point (760 mmHg):
101.4 °C
Flow Characteristics
Viscosity @ 20 °C:
2.72 mPa·s
Viscosity @ 50 °C:
1.15 mPa·s
Volume expansion (−10 to 90 °C):
5.1 %
Chemical Stability & Compatibility
Propylene glycol concentration:
25 vol%
pH:
8.0–10.5
Reserve alkalinity:
>6.0 mL 0.1 N HCl
Sulfate: <10 ppm
Chloride: <5 ppm
Total hardness (as CaCO₃): <20 ppm
Typical properties are provided for informational purposes only and are not guaranteed specifications. Refer to the product datasheet for full details.