In This Article
High-power PCS designs adopt liquid cooling not simply to replace fans with coolant, but to control IGBT or SiC junction temperature and module-to-module temperature variation while power density rises, enclosures become more sealed, and outdoor environmental requirements become more demanding. On the manufacturing side, the critical issues are the power-module thermal interface, cold-plate flow distribution, post-joining flatness, leak and pressure integrity, cleanliness, and long-term corrosion reliability.
Why PCS thermal management deserves a closer look
A power conversion system, or PCS, sits between a battery system and the AC grid and performs bidirectional power conversion. Grid-forming PCS designs can also provide functions such as voltage support, frequency support and black start.
The key heat-generating components are the power semiconductors and associated power-electronic assemblies, especially IGBT and SiC power modules. Conduction and switching losses ultimately become heat, making junction temperature and thermal cycling important reliability factors. A 2025 IGBT thermal-management review identifies junction temperature, thermal resistance and temperature gradient as major design targets. Renewable and Sustainable Energy Reviews
Commercial platforms already demonstrate liquid cooling in higher-power PCS designs. Inovance’s IES600 400 kW string PCS uses fully enclosed liquid cooling for core components, IP65 protection and full-capacity operation at 2,000 m altitude. Kehua introduced a 500 kW liquid-cooled string PCS in 2026. HyperStrong’s HyperBlock M uses a self-developed SiC PCS and a dual-channel liquid-cooled TMS for batteries and PCS. Inovance Kehua HyperStrong
These examples do not mean every PCS will become liquid-cooled. They do show that liquid cooling is an established engineering option when high power, enclosure protection and environmental capability must be achieved together.
What is the actual thermal path?
A simplified power-module path is:
IGBT / SiC die -> power-module baseplate -> TIM -> cold plate -> coolant -> manifold -> thermal-management loop
The cold plate is part of the total thermal-resistance chain rather than an isolated accessory.
| Layer | Primary function | Manufacturing or assembly risk |
|---|---|---|
| IGBT / SiC module | Power switching | Device loss and junction temperature |
| Module baseplate | Spreads heat | Warpage and local hot spots |
| TIM | Fills microscopic gaps | Thickness variation and contact resistance |
| Cold-plate surface | Receives module heat | Flatness and roughness |
| Internal channels | Transfer heat to coolant | Flow, pressure drop and dead zones |
| Manifold | Distributes coolant | Flow imbalance |
| Cooling loop | Rejects heat | Pump, heat exchanger and coolant |
The module baseplate - TIM - cold-plate surface interface is often underestimated. A sophisticated channel cannot compensate for a large interface gap caused by cold-plate distortion.
Why liquid cooling is about more than heat-transfer coefficient
Air cooling can remain appropriate for lower-power systems with enough space and benign environments. Liquid cooling becomes more attractive when several constraints occur together.
| PCS design change | Air-cooling challenge | Potential liquid-cooling value |
|---|---|---|
| Higher unit power | More airflow and larger heat sinks | Greater heat removal per package volume |
| Higher power density | Air ducts consume space | Cooling close to heat source |
| Thinner enclosure | Harder airflow routing | Heat transported through liquid lines |
| IP65/sealed core | Large external airflow is difficult | Supports sealed core components |
| High altitude | Lower air density reduces cooling | Liquid-side transfer is less dependent on air density |
| Dust and salt exposure | Contamination can enter with airflow | Reduces exposure of core electronics |
| Multiple power modules | Temperature spread can increase | Manifold/channel design improves flow balance |
A more useful statement is:
PCS liquid cooling simultaneously supports thermal performance, power density and environmental protection.
If SiC is more efficient, why is liquid cooling still relevant?
SiC can reduce some conduction and switching losses and enable higher switching frequency, voltage and packaging density. It is inaccurate to say that SiC simply runs hotter than IGBT.
The system-level trend is often:
higher device efficiency -> higher feasible power or smaller package -> higher system power density -> thermal management remains a design constraint
HyperStrong publicly combines SiC PCS with PCS liquid cooling in HyperBlock M and combines SiC PCS with an all-liquid-cooled architecture in HyperBlock IV. HyperStrong HyperBlock M HyperBlock IV
Which parts are precision-manufacturing opportunities?
| Part | Typical material | Typical process | Main manufacturing concerns |
|---|---|---|---|
| Cold-plate base | 6061/6063 or similar aluminum | CNC, extrusion plus machining | Flatness, channels, distortion |
| Cover plate | Aluminum | Machining and joining | Joint quality and distortion |
| Manifold | 6061 aluminum | CNC deep/cross holes | Flow distribution, burrs, cleanliness |
| Connector block | Aluminum/stainless steel | CNC | Threads and sealing |
| Power-module base | Aluminum/copper | Precision machining | Flatness and location |
| Copper connection block | Copper | CNC/forging plus machining | Conductivity, heating and burrs |
| Mounting frame | Aluminum/steel | CNC/sheet metal | Assembly stack-up |
A supplier that combines material control, CNC, joining, cleaning, surface treatment, leak/pressure testing and assembly can extend from a single cold plate to a cold plate + manifold + connector + subassembly package.
Why PCS cold plates cannot simply copy GPU cold plates
| Comparison | AI GPU cold plate | PCS IGBT/SiC cold plate |
|---|---|---|
| Main heat source | GPU/CPU/HBM | IGBT/SiC power modules |
| Heat-source geometry | Small, concentrated | Larger modules, multiple positions |
| Channel trend | Fine/microchannels increasingly common | Conventional millimeter-scale channels and distribution networks |
| Mounting interface | Package interface | Large power-module face |
| Primary thermal issue | Local high heat flux | Module temperature spread and area uniformity |
| Environment | Indoor data center | Outdoor ESS and power stations |
| Corrosion life | Important | Strong long-life requirement |
| Protection level | Inside server system | PCS often needs environmental protection |
| Product cycle | Faster platform iteration | Longer maintainability expectations |
PCS cold plates may not require the finest channels, but often place greater emphasis on large-area flatness stability, long-term sealing, coolant compatibility and branch flow balance.
How should the cold-plate manufacturing route be selected?
| Manufacturing route | Suitable use | Advantages | Main risks |
|---|---|---|---|
| CNC channels + joined cover | Prototype, medium volume, complex channels | High design freedom | Machining time and joining distortion |
| Extrusion + CNC | Standardized medium/high volume | Good material efficiency | Channel geometry constrained |
| Vacuum brazing | Multi-layer or large-area joints | Complex internal structure | Furnace distortion and braze control |
| Friction stir welding | Aluminum plate channels | Solid-state joining | Weld path, clamping and distortion |
| Embedded tube | Simpler flow paths | Mature and cost-effective | Interface consistency and thermal resistance |
The team should first ask whether complex channel geometry is truly required, whether annual volume justifies near-net-shape blanks, whether final machining allowance remains after joining, whether plate size amplifies distortion, and what working/proof pressure is required.
Why mounting-surface flatness is a first-level CTQ
Power modules are normally clamped to the cold plate through a TIM. The TIM should fill microscopic gaps, not compensate for significant macroscopic warpage.
An unstable mounting surface can create:
local gap increase -> TIM thickness variation -> higher contact thermal resistance -> local temperature rise -> larger module-to-module temperature spread
Flatness must therefore be reviewed together with module baseplate condition, TIM type and target thickness, fastener locations, tightening sequence, cold-plate wall thickness and channels, joining distortion and operating-temperature deformation.
This is a CTQ governed jointly by thermal design, mechanical stack-up and assembly process.
Why final machining before joining often creates trouble
A common failure chain is:
final machining -> joining thermal cycle -> residual-stress release/local shrinkage -> flatness change -> rework
A more robust sequence often looks like:
blank -> rough/semi-finish channels -> join cover -> stabilize -> final machine mounting face -> finish interfaces -> clean -> leak/proof test
Joining must be included in the dimensional stack and machining-allowance plan rather than treated as an independent post-process.
Why a manifold can be more demanding than it looks
A modular PCS can use a manifold with supply/return galleries, multiple branch holes, deep and cross-drilled passages, threaded ports, O-ring grooves, plug ports and sensor/valve interfaces.
| CTQ | Failure consequence |
|---|---|
| Branch cross-section | Unequal module flow |
| Deep-hole drift | Thin wall or incorrect intersection |
| Cross-hole burrs | Particles can block downstream channels |
| O-ring groove | External leakage |
| Plug sealing face | Long-term seepage |
| Internal cleanliness | Particles reach pumps, valves or cold plates |
| Material consistency | Corrosion and strength variation |
For a long-life circulating system, cross-hole deburring and internal cleanliness can be as important as dimensional tolerance.
Leak test, proof pressure and burst test are different
| Verification | Purpose | Main focus |
|---|---|---|
| Leak test | Detect measurable leakage | Joints, fittings, O-rings and porosity |
| Proof pressure | Confirm no permanent damage/leakage at specified pressure | Structural margin |
| Burst test | Establish destructive pressure limit | Ultimate structural capability |
Production pressure values should come from the customer’s design specification and applicable validation requirements rather than an arbitrary multiplier chosen by the supplier.
Why corrosion is a first-class power-electronics cooling issue
Passing an initial leak test does not guarantee years of reliable operation. Risks include aluminum-coolant compatibility, galvanic corrosion between copper and aluminum, coolant conductivity and ionic contamination, joint-area corrosion, residual cleaning chemistry, particles/flux/machining debris and surface-treatment suitability inside the fluid path.
Material selection should therefore consider:
thermal performance + strength + joinability + machinability + coolant compatibility + long-term corrosion
How should the process change from prototype to mass production?
| Stage | Typical strategy | Main objective |
|---|---|---|
| Prototype | Machined-from-solid, flexible channels | Fast validation |
| EVT/DVT | Stable datums and joining parameters | Freeze thermal/hydraulic/structural design |
| Pilot | Dedicated fixtures and process gates | Demonstrate repeatability |
| Mass production | Extrusion/near-net shape plus CNC, automated test | Reduce unit cost and variation |
Scale-up should reduce CTQ variation, joining distortion, rework, leak failures, cleanliness risk, inspection time and material variation, not only machining time.
Why PCS, SVG and other high-power equipment can share manufacturing capabilities
Liquid cooling also appears in SVG/STATCOM, wind converters, central inverters, medium/high-voltage drives and other high-power electronic systems.
Hopewind publicly offers multiple 10 kV and 35 kV water-cooled SVG products, showing that a power module + cold plate + cooling loop architecture is not unique to battery storage. Hopewind
The more useful manufacturing-platform definition is:
High-Power Electronics Liquid Cooling Manufacturing
The same aluminum/copper machining, CNC channels, manifolds, joining, cleaning, leak/pressure testing and subassembly capabilities can serve multiple power-electronics applications.
RFQ: what should be defined before design and quotation?
| RFQ input | Why it matters |
|---|---|
| IGBT/SiC module type and layout | Heat-source and mounting geometry |
| Module and total heat loss | Thermal boundary |
| Coolant type and concentration | Material compatibility |
| Inlet temperature range | Thermal calculation |
| Target flow rate | Channel design |
| Allowable pressure drop | Pump/system matching |
| Working pressure | Structural design |
| Proof/burst requirements | Validation planning |
| Mounting-face flatness/roughness | TIM interface control |
| Connector specification | Sealing and assembly |
| Material and surface treatment | Corrosion/process route |
| Annual volume and ramp plan | CNC versus near-net-shape decision |
| Joining restrictions | DFM/equipment matching |
| Cleanliness requirement | Cleaning and acceptance planning |
| Leak-test specification | Production quality gate |
Conclusion
Liquid-cooled high-power PCS should not be viewed as AI-server cold-plate technology simply moved into an energy-storage cabinet.
The two applications share a manufacturing foundation, but PCS has different heat-source geometry, module interfaces, outdoor environments and lifetime expectations.
The manufacturing logic must connect:
heat-source layout -> TIM interface -> mounting-face flatness -> channel/manifold distribution -> joining distortion -> leak and pressure integrity -> cleaning -> corrosion compatibility -> production control
The manufacturing opportunity is therefore not merely machining an aluminum plate. It is managing the cold plate, manifold, connector and critical assembly interfaces as one thermal-fluid-mechanical system.
FAQ
Why are more high-power PCS designs adopting liquid cooling?
Liquid cooling is not used only for higher heat-transfer capability. As PCS power density increases, packaging becomes more compact, core components require sealed protection, and systems must operate in outdoor, high-altitude, dusty or corrosive environments, liquid cooling can help control junction temperature and module-to-module temperature spread while reducing dependence on large volumes of external cooling air.
How is a PCS cold plate different from an AI server GPU cold plate?
Both require control of thermal resistance, flatness, flow, sealing and cleanliness, but the heat-source geometry is different. GPU cold plates increasingly focus on localized high heat flux and fine channels, while PCS cold plates often cool multiple IGBT or SiC power modules and place greater emphasis on large-area temperature uniformity, flow distribution, long-term pressure integrity, corrosion and outdoor equipment life.
What are the most important manufacturing CTQs for an IGBT or SiC power-module cold plate?
Typical CTQs include mounting-surface flatness and roughness, mounting-hole position, channel geometry and pressure-drop consistency, manifold flow distribution, joining-induced distortion, internal cleanliness, leak integrity, and long-term material compatibility with the selected coolant.
What information should be included in a PCS cold-plate RFQ?
The RFQ should at minimum define the power-module type and layout, device heat loss or thermal load, coolant, inlet temperature, target flow rate, allowable pressure drop, working and proof pressure, mounting-surface flatness, connector design, surface treatment, annual volume, joining method, leak-test criteria and cleanliness requirements.
