


# Liquid Cooled PCB: Cooling Methods and Design Considerations

_Published: 2026-11-17_

Table of Contents

Toggle

- What Is a Liquid Cooled PCB?
- PCB Liquid Cooling: Cold Plates, Immersion, and Embedded Channels
- How Does PCB Design Impact Thermal Performance?
- How Do Coolant Temperature and Flow Affect PCB Cooling?
- What Are the Reliability Risks of a Liquid Cooled PCB?
- Where Is PCB Liquid Cooling Used?
- EBest Circuit’s PCB Manufacturing and Assembly Capabilities
- FAQs About Liquid Cooled PCB

A [liquid cooled PCB](https://www.bestpcbs.com/blog/2026/11/liquid-cooled-pcb/) uses a liquid-based thermal path to remove heat from electronic components, either through a separate cold plate, a compatible immersion fluid, or cooling structures integrated into the board assembly. Its performance depends on where heat enters that path, the resistance between the component and coolant, and the operating conditions. A lower coolant temperature alone cannot compensate for a poor thermal interface or a concentrated hotspot.

EBest Circuit combines FR4 and metal-core PCB manufacturing with component sourcing, SMT/THT assembly, and inspection. For electronics that interface with cooling hardware, coordinating the board stack-up, component placement, and assembly requirements helps reduce mismatches between the electrical board and its mechanical installation. To discuss your PCB or PCBA requirements, visit [Contact Us](https://www.bestpcbs.com/contact/index.htm) and complete the form at the bottom of the page, or email **sales@bestpcbs.com**.

![liquid cooled PCB](https://www.bestpcbs.com/blog/wp-content/uploads/2026/11/liquid-cooled-PCB-1.jpg)

## What Is a Liquid Cooled PCB?

A liquid cooled PCB is a board or assembled electronic module whose heat is transferred to a liquid cooling system. The liquid does not necessarily enter the PCB or touch the circuit. In a cold-plate assembly, it can remain entirely inside a separate metal heat exchanger mounted against the device package or a board-side thermal surface.

The distinction matters because total power and local heat flux describe different problems. In an illustrative calculation, 100 W passing through a 10 × 10 mm thermal contact area produces an average heat flux of **100 W/cm²**. The same 100 W spread over 20 × 20 mm produces **25 W/cm²**. Both assemblies have the same heat load, but the smaller contact concentrates it fourfold. These are area-average calculations, not measured die hotspot values.

Likewise, a 140 A current rating does not establish the cooling requirement. At a constant assumed resistance of 1 mΩ, conduction loss is I²R = **19.6 W**; at 2 mΩ, it is **39.2 W**. Switching losses and other heat sources must be added separately. Cooling specifications therefore need heat loss in watts, not just circuit current or output power.

## PCB Liquid Cooling: Cold Plates, Immersion, and Embedded Channels

PCB liquid cooling can use three substantially different arrangements. Their main difference is the location of the coolant relative to the electrical assembly.

Arrangement
Main heat-transfer path
Does coolant touch the circuit?
Principal design constraint

Separate cold plate
Package or board thermal surface → interface material → cold plate → coolant
Normally no
Contact area, interface thickness, mounting load

Dielectric immersion
Component surfaces → compatible dielectric fluid
Yes
Compatibility of all wetted materials and components

Embedded cooling structure
Heat source → integrated metal structure or channel wall → coolant
Normally isolated from conductors by the designed fluid boundary
Channel construction, sealing, electrical isolation, repairability

A [metal-core PCB](https://www.bestpcbs.com/blog/2026/09/mcpcb-stackup/) can spread heat into an attached cold plate without containing any fluid channels. Similarly, a conventional [copper coin](https://www.bestpcbs.com/blog/2026/09/buried-copper-coin-pcb/) is a **solid heat-conduction insert**. It only becomes part of a fluid-carrying structure when channels, tubing, or another sealed coolant passage are deliberately integrated with it.

Cold plates themselves can use tubing embedded in a plate, brazed constructions, or microchannels. Their sealing process belongs to the specific metal and fluid-boundary design; a PCB dielectric layer is not automatically a pressure-rated seal. This distinction prevents a common specification error: treating ordinary metal-core fabrication as equivalent to manufacturing a sealed liquid-cooled assembly.

![liquid cooled PCB](https://www.bestpcbs.com/blog/wp-content/uploads/2026/11/liquid-cooled-PCB-2.jpg)

## How Does PCB Design Impact Thermal Performance?

PCB design affects how far heat must travel, how widely it spreads, and how consistently the assembly contacts the cooling surface. Copper planes and thermal vias are important when heat travels through the board. For a top-cooled package, however, the main heat path may bypass most of the PCB thickness.

A concrete layout example comes from the Micron 9650 SSD. Its liquid-cooling-oriented architecture places approximately **90% of heat-generating components on one side of the PCB**, compared with roughly **60% in the typical designs described by Micron**. Concentrating those components toward the cold-plate interface reduces the need to conduct heat from the opposite side through the board. This is a layout comparison, not a claim of 30% better cooling.

Interface thickness can be just as consequential. Consider an idealized, uniform thermal interface with conductivity k = 3 W/(m·K), a 20 × 20 mm contact area, and 100 W passing through it. The calculation uses R = t/(kA).

Assumed interface thickness
Calculated interface resistance
Temperature drop across this layer at 100 W

0.10 mm
0.083 K/W
8.3°C

0.50 mm
0.417 K/W
41.7°C

The thicker layer has five times the conduction resistance under these assumptions. These calculated values exclude contact resistance, spreading resistance, surface roughness, and material compression. They are not predictions for a finished board. They show why component height variation, board flatness, and cold-plate clearance deserve attention alongside the advertised conductivity of a thermal pad.

For board-side cooling, via construction, copper distribution, dielectric thickness, and electrical isolation must work together. Adding copper helps only when it forms a useful path to the cooled surface; it does not remove a bottleneck elsewhere in the assembly.

![liquid cooled PCB](https://www.bestpcbs.com/blog/wp-content/uploads/2026/11/liquid-cooled-PCB-3.jpg)

## How Do Coolant Temperature and Flow Affect PCB Cooling?

Coolant inlet temperature sets the thermal starting point, while flow determines how much the fluid warms as it carries heat away. The component-to-fluid temperature difference also depends on the cold plate and interfaces.

For a single-phase liquid at steady state, the energy balance is given below.

**Q = ṁ × cp × (Tout − Tin)**

Here, Q is heat transferred into the coolant in watts, ṁ is mass flow in kg/s, and cp is specific heat in J/(kg·K). Consider the following engineering example using approximate room-temperature water properties. It represents water inside a closed cold-plate circuit, not water contacting a powered PCB.

Parameter
Baseline
Lower-flow case
Higher-heat case

Heat entering coolant
200 W
200 W
400 W

Coolant inlet temperature
25°C
25°C
25°C

Volumetric flow
1.0 L/min
0.5 L/min
1.0 L/min

Assumed density
1,000 kg/m³
1,000 kg/m³
1,000 kg/m³

Assumed specific heat
4,180 J/(kg·K)
4,180 J/(kg·K)
4,180 J/(kg·K)

Calculated mass flow
0.0167 kg/s
0.00833 kg/s
0.0167 kg/s

Calculated coolant rise
2.87°C
5.74°C
5.74°C

Calculated outlet temperature
27.87°C
30.74°C
30.74°C

These are calculated operating examples, not recommended flow specifications or test results. They assume all stated heat reaches the coolant, negligible heat loss elsewhere, and no boiling. A water-glycol mixture or dielectric fluid requires its own density and specific heat.

The table predicts **coolant temperature rise, not semiconductor junction temperature**. For a separate simplified example, assume an effective junction-to-inlet thermal resistance of 0.20 K/W at a specified operating flow. A 200 W device with a 25°C inlet would then reach 65°C. Raising the inlet to 35°C gives 75°C if the heat load and effective resistance remain unchanged. Because the resistance is referenced to the inlet, the outlet rise is not added again.

More flow may reduce thermal resistance, but it also changes pressure drop and pumping demand. The operating flow must come from the pump-and-loop combination; a pump’s unrestricted flow rating does not establish the flow through a restrictive microchannel plate.

## What Are the Reliability Risks of a Liquid Cooled PCB?

The main risks are loss of fluid containment, deterioration of wetted materials, and mechanical stress introduced by the cooling hardware. A low operating temperature does not by itself demonstrate reliable solder joints or seals.

Cold-plate mounting adds load to the package, board, and supports. Differential thermal expansion then changes that load during heating and cooling. As an illustrative calculation, two constrained structures with a **7 ppm/K difference in expansion coefficient**, a **100 mm span**, and a **60°C temperature change** have a free-expansion mismatch of **42 µm**. This is not a prediction of solder-joint displacement or an acceptance limit; actual deformation depends on mounting geometry, stiffness, and compliance.

For immersion assemblies, compatibility extends beyond the PCB laminate to connectors, labels, adhesives, seals, and thermal interface materials. Initial electrical operation does not establish long-term fluid compatibility. For closed cold-plate loops, the fluid, tubing, seals, and metal surfaces must also be compatible with one another.

The useful reliability comparison is therefore the complete assembly before and after defined thermal, mechanical, and fluid exposure—not simply a cooler infrared image of the board.

## Where Is PCB Liquid Cooling Used?

PCB liquid cooling is relevant where concentrated heat, limited airflow, or continuous operation makes the thermal path difficult to manage. The board’s electrical power rating alone does not determine whether it needs liquid cooling.

- AI and HPC assemblies. NVIDIA specifies a configurable maximum TDP of up to 700 W for the H100 SXM. That provides a concrete scale for accelerator thermal design, but it is not a universal PCB power rating or proof that every H100 installation uses the same cooling arrangement. Power-delivery components and memory still need their own thermal paths.
- Dense storage systems. Micron’s 9650 example shows that controller, DRAM, and NAND placement can be reorganized around a cold-plate interface. Liquid cooling affects board layout as well as the enclosure.
- Power converters and motor drives. An illustrative 10 kW-output converter operating at 98% efficiency dissipates approximately 204 W, calculated as 10,000 × (1/0.98 − 1). The cooling load is the conversion loss, not the full 10 kW output. Device-level losses determine how unevenly that heat is distributed.
- RF amplifiers and laser electronics. Localized heat and temperature stability can make a short thermal path valuable, particularly where airflow access is restricted. The relevant limits remain those of the specific devices and assembly.

These applications have different electrical isolation, packaging, and service requirements. A cooling structure suitable for a processor board is not automatically suitable for a high-voltage power stage.

## EBest Circuit’s PCB Manufacturing and Assembly Capabilities

EBest Circuit offers the following PCB fabrication and assembly capabilities for the electronic portion of liquid-cooling projects.

Item
Capabilities

FR4 layer count
Standard: 1–10 layers; special process: 10–32 layers; high-Tg material for 8 layers and above

FR4 inner-layer copper
Standard: 0.5–5 oz; special process: 5–20 oz

FR4 outer-layer copper
Standard: 1–5 oz; special process: 5–20 oz

FR4 processing thickness, ENIG or OSP
Standard: 0.4–3.5 mm

FR4 minimum finished hole diameter
Standard: 0.20 mm; special process: 0.15 mm

FR4 maximum through-hole aspect ratio
Standard: 8:1; special process: 10:1

MCPCB processing thickness
Standard: 0.8–3.0 mm; 4.0 or 5.0 mm with custom material procurement

MCPCB dielectric-layer thickness
Standard: 100–110 μm; 150 or 200 μm subject to confirmation

MCPCB inner-layer copper
Standard: 0.5–3 oz

MCPCB outer-layer copper
Standard: 1–3 oz

MCPCB outline tolerance
Standard: ±0.15 mm; special process: ±0.10 mm

Minimum SMD component
01005

Minimum BGA pitch
0.25 mm

Assembly processes
SMT, through-hole, and mixed assembly

Specifications depend on the board construction and process combination. These PCB/PCBA capabilities do not specify cold-plate fabrication or coolant-channel sealing.

## FAQs About Liquid Cooled PCB

**Can ordinary water directly cool a powered PCB?**

Ordinary water should remain inside a properly isolated cooling circuit. Direct immersion requires a suitable dielectric fluid and compatible equipment. Deionized water is not a permanent substitute for electrical isolation because contamination can change its conductivity.

**Will a liquid-cooled board still need fans?**

It may. A cold plate can remove heat from selected devices while leaving inductors, capacitors, connectors, and other components dependent on airflow or another thermal path. Fan requirements depend on the remaining heat sources, not just the main processor temperature.

**Does liquid cooling eliminate condensation?**

No. Condensation can occur when an exposed surface is colder than the local air’s dew point. Lowering the coolant temperature can therefore introduce moisture problems unless humidity, insulation, and temperature control are addressed.

**Can an existing PCB be converted to liquid cooling?**

Sometimes, but adding a cold plate requires suitable mounting support, clearance, and contact surfaces. A board with uneven component heights or unsupported mounting loads may need mechanical changes or a layout revision before the conversion is practical.

**Are smaller cooling channels always better?**

No. Smaller passages can improve heat transfer in a suitable design, but they can also increase pressure drop and sensitivity to contamination. Channel size must be considered together with the coolant, flow distribution, filtration, and available pumping pressure.

**Is an infrared image enough to verify cooling performance?**

No. It shows accessible surface temperatures, affected by emissivity and reflections; it does not directly establish every device’s junction temperature. A meaningful assessment also records electrical losses, coolant inlet and outlet temperatures, flow, and the installed mechanical configuration.

A successful [liquid cooled PCB](https://www.bestpcbs.com/blog/2026/11/liquid-cooled-pcb/) project needs a repeatable electronic build as well as an effective heat exchanger. EBest Circuit brings FR4 and metal-core fabrication, component procurement, assembly, and inspection into one coordinated workflow, helping reduce handoffs as the board moves from fabrication to populated assembly. Send your PCB or PCBA requirements through [Contact Us](https://www.bestpcbs.com/contact/index.htm), using the form at the bottom of the page, to discuss the board construction and assembly scope for your cooling architecture.
