As the power consumption of AI servers continues to rise, liquid cooling has shifted from an optional choice to a mandatory requirement for data centers. This transition left a complete trace in the disclosure windows of the 2025 annual reports and the 2026 semi-annual reports: who is investing in R&D, where the investments are directed, and how far the progress has reached. Taking 16 A-share manufacturers with existing liquid cold plate R&D or products as samples, and based on the R&D project lists in the 2025 annual reports, financial data in the 2026 semi-annual reports, and the latest disclosures from investor interaction platforms, this article reviews the situation in three parts: the first part focuses on R&D investment directions, the second part examines the liquid cold plate layout, and the third part evaluates R&D capabilities and market realization.
*The 16 A-share manufacturers with existing liquid cold plate R&D or products are Envicool, Goaland, Shenling Environmental, Tongfei Shares, Frd Science, Suqun New Material, Jones Tech Plc, Yinlun Co., Ltd., JSCC Group, Cotran New Material, Jonhon Optronic Technology, Yd Electronic Technology, CRUN, Hongfuhan Technology, LINGYI iTECH, and Speed Wireless Technology.
01. Investment Directions: R&D Directions of the 16 Manufacturers Are Determined by Their Main Businesses
The R&D investments of the 16 manufacturers vary, and the differences are directly related to their respective main businesses. From the R&D projects in the 2025 annual reports, there are four common layout directions: control systems, in-house material development, complete cabinets, and testing and verification. The following text will first sort out these four commonalities by direction, and then analyze the sources of differences by main business.
First, control systems have become a key R&D focus for multiple manufacturers. Goaland has five ongoing control-related projects, ranging from embedded controllers to liquid cooling AI application architectures; Shenling Environmental has completed the in-house development of programmable logic controllers; Envicool has turned AI energy-saving algorithms into pure software modules. The reasons are written in the project proposals: Shenling Environmental's goal is to replace external purchases and reduce costs, while Envicool's goal is to deploy upgrades through software without adding new hardware. The value of control systems is directly reflected in energy efficiency indicators—the PUE targets disclosed by various companies have been reduced from 1.3 to 1.15, and these indicators must be achieved through control systems.
Second, projects for in-house material development are increasing, with the goal of reducing costs. Frd Science is developing brazable die-cast aluminum alloys, and the project proposal specifies that they are used to replace silver-based and nickel-based materials; the standard plate development of CRUN is described as the "implementation of the component localization strategy"; Yinlun Co., Ltd. has built a complete chain from die-casting materials to brazing processes. The wording in the project proposals of the three companies consistently points to two words: cost and independence. The investment in in-house development of materials and components corresponds to the part of the processing link in product costs that cannot be compressed further.
Third, products are expanding from single components to complete cabinets and entire stations. Tongfei Shares is developing integrated liquid cooling cabinets that can run immediately after server racking; Shenling Environmental is developing prefabricated cooling stations and integrated cooling sources; CRUN is building an engineering module library for CDUs and Manifolds. At the same time, Frd Science, Shenling, Goaland, and Tongfei are all developing air-liquid hybrid products, and the goals specified in the project proposals are to adapt to the retrofitting of traditional waterless computer rooms and operation in water-scarce regions. While expanding product forms, the four companies have unanimously retained air cooling compatibility, corresponding to the existing market of operating computer rooms.
Fourth, the progress of projects among various companies has diverged, and leading manufacturers have begun to build their own long-term operation testing platforms. Single-phase cold plates are advancing the fastest; JSCC Group has sent samples, and Suqun New Material has entered trial production; two-phase cold plates and immersion cooling are still in the early stages of R&D. Emerging simultaneously with the progress gap is the construction of testing capabilities: Envicool has built a long-term operation testing platform for cooling system fluids; Frd Science is researching performance testing methods for liquid cooling modules; Goaland is researching data center liquid cooling purification technology and coolant corrosion mechanisms. Liquid cooling equipment entering a customer's supply chain requires long-term operation verification, and testing platforms are the threshold for entering customers' shortlists.
The above is the part where directions converge. For the part where outcomes diverge, the answers must be found in the main businesses of each company. Comparing R&D projects with main businesses, the 16 companies present three paths.
Status of Ongoing R&D Projects in Computing Liquid Cooling Directions of the 16 Manufacturers in 2025
Source: Financial reports of each company, compiled by MIR
After comparison, the 16 companies can be divided into three situations.
The first type: the main business is thermal management, and liquid cooling is an extension of existing business. The four companies, Envicool, Goaland, Shenling, and Tongfei, have R&D projects covering the entire process of complete machines, components, control, and working fluids. In the common trends, the control system track is mainly driven by these companies—control systems are originally part of their complete machine business. In terms of progress, these companies are also at the forefront: Tongfei's data center manifold and cold water backplane have completed achievement transformation; Envicool's quick-connect joints have entered pilot testing; Tongfei's cabinets and Shenling's cooling stations are under R&D; Goaland has as many as five control-related projects. The reason they can do the full chain lies in the main business structure: they originally delivered system-level thermal management equipment, and complete machine integration and control are inherently their existing capabilities.
The second type: the main business is thermal dissipation materials or precision parts, extending to liquid cooling components. The six companies are Frd Science, Suqun New Material, Jones Tech Plc, JSCC Group, Yd Electronic Technology, and Cotran New Material. The trend of in-house material development mainly emerges in this group—materials are originally their main business, and R&D is just applying existing formulas and processes to liquid cooling scenarios. The progress gap in this group is the most obvious: Jones Tech Plc's liquid cooling system integration and two-phase flow thermal management have both achieved mass production, making it the company among the six whose annual reports disclose that all R&D liquid cooling projects have been mass-produced; Suqun New Material's 3DVC is in trial production, and JSCC Group's cold plates have been sampled; some of Cotran New Material's liquid cooling products have entered mass production, but there are few subsequent R&D pipelines. The three companies with advanced progress have a common point: their main business processes are directly related to cold plates—Jones Tech Plc's graphite materials plus microchannel processes, JSCC Group's MIM injection molding processes are directly used for cold plate structural parts, and Suqun New Material's VC vapor chamber processes are directly used for 3DVC.
The third type: the main business is relatively far from liquid cooling, entering in the form of single products or capacity. Jonhon Optronic Technology has only one cold plate component under R&D, expanding adjacently around its connector main business; CRUN transformed from hydraulics and boilers, and all its liquid cooling projects are in the development stage, ranking last in progress among the 16; LINGYI iTECH and Speed Wireless Technology already have liquid cold plate products, but there are no liquid cooling directions in their ongoing R&D projects. From the perspective of R&D projects, the entry of such companies mainly relies on existing manufacturing capacity and customer channels, and R&D investment has not yet become the main driving force.
02. Liquid Cold Plate Layout: Competitive Barriers Are Shifting from Design to Processes and Working Fluids
Distribution of ongoing R&D project investments among 9 companies directly developing cold plates
Source: Financial reports of each company, compiled by MIR
Among the 16 manufacturers, 9 take the cold plate as an independent R&D object, 4 focus on process and material supporting around the cold plate, 2 have the cold plate as only a part of their system projects, and 1 has its layout on the adjacent component of the heat dissipation substrate. For the 9 companies directly developing cold plates, the progress ranges from mass production to the development stage, and the technical routes are concentrated on three paths: single-phase, two-phase, and 3DVC. The above table organizes cold plate-related projects by manufacturer.
Four points can be seen from this table.
First, the competition for the cold plate body has been stratified. Ranked by progress, Jones Tech Plc and Cotran New Material are in the mass production layer; Yinlun Co., Ltd.'s chip cold plate has completed the project; JSCC Group has reached the sampling stage; Suqun New Material is in trial production; Jonhon Optronic Technology and CRUN are still in development. The cold plate is the component that most prominently presents the "full progress band from mass production to development" among the 16. In the future, this segment is expected to face sufficient competition. The technical development difficulty of liquid cold plates is not high; the key lies in implementing protective measures such as leak prevention. Therefore, most manufacturers with precision processing capabilities will choose to enter.
Second, single-phase cold plates are the current main force, and two-phase is a clear next-generation direction. Among the 9 companies directly making cold plates, 6 are on the single-phase route (Jones Tech Plc, Cotran New Material, JSCC Group, Suqun New Material, Jonhon Optronic Technology, CRUN), and 4 are researching the two-phase route (JSCC Group, Envicool, Goaland, plus CRUN's common research also covers immersion cooling). In terms of progress, single-phase has reached sampling and mass production, while two-phase is in the early stages of design or development. A noteworthy detail: while sampling single-phase cold plates, JSCC Group has already started the design of two-phase cold plates, with two generations of routes advancing in parallel within the same company.
Third, 3DVC is the common choice for material-based manufacturers to enter chip-level thermal dissipation. Frd Science (project completed and applied at scale), Suqun New Material (trial production), and Envicool (trial production) have all invested in R&D for 3DVC. The commonality of the three is that they were already making VC vapor chambers or thermal conductive materials. For them, 3DVC is a direct extension of existing processes without needing to cross over to liquid systems.
Fourth, the competition for cold plates is expanding towards both processes and working fluids. On the process side, Frd Science is working on vacuum brazing and friction stir welding; Envicool is researching high-frequency induction brazing; Jones Tech Plc is optimizing the cold plate microchannel process. On the working fluid side, Goaland is researching the corrosion mechanism between coolant and copper plates. After the design of the cold plate body converges, welding quality and working fluid compatibility have become new differentiators.
03. R&D Generally Exceeds Profitability Support, and Market Realization Speed Is Diverging
Comparison of R&D Capabilities of the 16 Manufacturers in the First Half of 2026 (2026 Semi-Annual Report)
Source: Financial reports of each company, compiled by MIR
In the first half of 2026, the R&D investments of all 16 companies increased year-on-year, with a total R&D investment of approximately CNY 3.87 billion and a total net profit of approximately CNY 3.24 billion—the sum of net profits is insufficient to cover R&D investments, and the gap must be filled by the existing funds of each company. Profit performance is split down the middle: 8 companies grew, and 8 companies declined or incurred losses. Overall status: R&D is being increased across the board, but output results are diverging.
R&D generally exceeds profitability support, with 9 companies having R&D investments higher than net profits. Among the 16, 9 have R&D investments exceeding net profits, and R&D expansion has gone beyond the support range of current profitability. Subsequent investments must be maintained by existing cash or external financing. More than half of these 9 companies have negative net cash simultaneously: the manufacturers with the most aggressive R&D investments happen to be those with the tightest room for fund allocation. Although all manufacturers seem not far from liquid cooling, the migration of capabilities still requires evaluating whether their own technical capabilities can provide support.
Looking at the realization of R&D from the market side, domestic liquid cold plate projects are currently undergoing a transition from sampling to mass production implementation. In terms of delivery forms, Suqun New Material's CSP orders are delivered on a rolling monthly basis, and Hongfuhan Technology's winning bid orders are placed monthly from June 2026 to January 2027. The two have different customer types but consistent ordering methods. The liquid cooling penetration of overseas NVIDIA chips is faster than that domestically, and the growth rate of liquid cooling demand is correspondingly faster: Envicool has passed Intel's test certification and entered the complete machine certification system packaged with quick-connect joints and CDUs; Goaland has obtained NVIDIA GB300 certified supplier qualification and Google shortlist trial production orders.
Market Dynamics of Liquid Cold Plates of Various Companies
Source: Financial reports of each company, compiled by MIR
Conclusion
The review of the 16 manufacturers shows: the distance between the main business and liquid cooling affects whether R&D is invested in liquid cooling and in which segment; the progress of cold plate R&D has been stratified; and the across-the-board increased R&D investments are diverging into different market realization speeds. Looking ahead 12 to 18 months, the pipelines of the 16 companies will be concentratedly transformed into products. The cold plate segment will simultaneously see three states: capacity release, stratification of certification lists, and monthly order delivery. The penetration speed of the overseas chain and the time point when more manufacturers cross the mass production line are points worth observing next. We will continue to track order and shipment data to observe the trend of the liquid cold plate market.