The Potential of 3D Printing in Cooling Solutions and Industry Progress
1. The surge in AI computing power demands accelerates the transition of server cooling to liquid cooling.
Upgrading cooling technology is an inevitable trend: With the explosive growth in AI server computing power and the vigorous promotion of green energy policies, server cooling, represented by NVIDIA, is rapidly transitioning to liquid cooling. From traditional air cooling to cold plate liquid cooling, and now to the current microchannel cold plate mode, there has been a qualitative leap in cooling performance and requirements. Currently, NVIDIA is validating silent and cold plate technology routes internally, and core suppliers of micron-level waterway designs are actively conducting supporting R&D and usage work.
Core suppliers showcase their strengths: Qihong Technology, as a major supplier of overall thermal solutions, played an important role when NVIDIA’s GB200 first adopted liquid cooling; Shuanghong Technology, as a global cooling solution provider, has greatly benefited from the promotion of liquid cooling modules; Zhongshi Technology, as a leading supplier for NVIDIA’s A100 and H100, occupies 30-40% of the market share; Yingwei, Jingyan Technology, and others are the main suppliers for OEM liquid cooling plate modules; Xiangxin Technology has achieved a microchannel precision of about 0.15, while Jingyan Technology’s microchannel water cooling precision has reached about 0.1.
2. Microchannel water cooling technology R&D faces obstacles, and overcoming challenges is urgent.
R&D companies face technical bottlenecks: Jingyan Technology, Xiangxin Technology, and others are actively developing microchannel MLCP water cooling technology. However, traditional liquid cooling plate heat dissipation uses machining discharge for precision processing, making it difficult to ensure micron-level (e.g., 0.1-0.15) microchannel precision. Additionally, traditional processes have many brazed joints, with a leakage rate as high as 5%, and traditional materials like aluminum cannot meet high-performance cooling requirements, while the process performance of copper alloys is also unsatisfactory.
Technological innovation points the way forward: To control leakage within 1% and achieve higher precision, auxiliary processes need to be introduced into the liquid cooling plate industry to meet high-performance cooling requirements.
3. Deficiencies in 3D printing processes hinder applications, and multiple measures seek breakthroughs.
Shortcomings of 3D printing processes: 3D printing technology has certain application potential in micron-level liquid cooling plate technology, but currently, there are issues with poor density. The density of traditional powder stacking 3D printing processes is uncontrollable, while the 3D printing of connection technology has slightly better density but is prone to cracking under high thermal cycling. Therefore, a mask forming process is adopted to shape copper-tungsten alloy plates, replacing traditional materials.
Process comparison seeks solutions: The density of binder processes is superior to that of Sim processes, and both processes require heat treatment (sintering process) to enhance density and strength. The precision of 3D printing in connection technology can meet the requirements of 3C and liquid cooling plates, with no stress, and appearance defects can be compensated through machining, while strength and mechanical properties can be restored through the sintering process. Currently, during the R&D process, both precision and leakage rates can be effectively controlled, but the yield in small batch production is low, which is related to the precision of printing equipment, such as Bolite and Huashu’s 3D printing equipment undergoing upgrades.
Yield improvement requires multiple measures: The yield of printed parts is about 85%, while the yield of finished products after sintering is about 65%. Post-processing of binder printing faces issues such as deformation, flow channel hole deformation, or cracking. By controlling powder stacking precision and density, and increasing tempering processes to adjust the bonding of material elements during sintering hardening, these issues can be effectively resolved.
Mainstream printing methods have significant advantages: The mainstream printing method theoretically employs a dual-track approach of laser melting and laser sintering, which can enhance product independence before post-processing, allowing metal elements to fuse adequately. After high-temperature tempering, it can effectively alleviate alarm rates, outperforming traditional powder stacking and binder printing processes.
Upgrading printing equipment aids mass production: Currently (for more real-time summaries, add WeChat: aileesir), the R&D phase uses 450 small printing devices for one-to-one printing, with a printing time of about 5 minutes for one liquid cooling plate, using a binder-type powder stacking machine. In the future, mass production will adopt multi-head printing equipment, with Huashu, Bolite, and other companies possessing large multi-head printing devices.
4. 3D printed liquid cooling plates have both advantages and disadvantages, and application prospects are promising.
3D printed liquid cooling plates have clear pros and cons: 3D printed liquid cooling plates have advantages such as fast speed, no stress, and lower manufacturing costs compared to machining, but they also have disadvantages such as complex processes and potential cracking and deformation during post-processing.
Performance improvement and customer recognition are pending: The efficiency of 3D printed liquid cooling plates can be improved by 30%, but there is currently no specific data on the enhancement of cooling performance. Ultimately, cooling performance is less related to the process and more dependent on the cost of the liquid cooling plate. Currently, 3D printing and dense processes are better choices for achieving micron-level flow channels, but cracking issues need to be overcome, as machining is difficult to achieve micron-level flow channels.
Process selection is based on actual needs: Achieving micro-hole precision with aluminum materials through machining is relatively easy, but it is more challenging to achieve micron-level microchannel precision with copper alloys or copper-tungsten alloys through machining. Additionally, for hundreds of microchannels, machining efficiency is low, making 3D printing and dense processes better choices.
Significant advantages compared to traditional processes: The full lifecycle of assembled 3D printed liquid cooling plates can save about one-third of the labor time compared to traditional CNC processes. CNC primarily consumes time in processing, while 3D printing is fast, but assembly takes time, and CNC employs integrated processing with fewer split parts.
5. The competitive landscape among enterprises is emerging, and traditional process precision meets standards.
Progress in enterprise-client connections varies: Mainstream enterprises have more connections with foreign clients, while domestically, apart from Jiyan Technology’s cooperation with Huawei, the application progress is relatively slow compared to overseas, such as NVIDIA’s faster progress.
Nanfeng Co., Ltd. focuses on material supply: Nanfeng’s subsidiary, Nanfeng Additive, mainly provides materials, primarily pure copper.
3D printing process applications are promising: The 3D printing process is expected to achieve small batch production applications by 2026, with a major overseas client potentially using 30% of this process next year. Verification began last year, and the plan is basically confirmed, using a binder jetting copper-tungsten alloy process.
Supplier roles are clearly defined: Jingyan Technology participates in OEM, with final assembly handled by Foxconn, and companies like Foshan Yiquan are also actively involved. Feirongda aims to enter the 3D printed liquid cooling plate field, but lacks significant advantages in materials and 3D printing technology.
Equipment selection and process determination: Equipment selection is basically determined to be Huashu’s, with the process adopting the binder jetting copper-tungsten alloy process.
Traditional process precision meets requirements: Traditional machining for fin processes to create microchannels can meet requirements, with diamond machine precision around 0.02mm.
Q&A
Q1: What is the situation of NVIDIA’s server cooling, the internal technology routes used, and the potential future technology routes?
A1: With the explosive growth in AI server computing power requirements, green energy policies, and the promotion of liquid cooling modes in the industry, NVIDIA’s cooling development is accelerating its transition to liquid cooling, evolving from traditional air cooling to last year’s cold plate liquid cooling phase, and now to the microchannel cold plate mode. The current internal validation phase technology routes are silent and cold plate; several core suppliers are in the supporting R&D usage phase for micron-level waterway designs.
Q2: What is the situation of NVIDIA’s core suppliers?
A2: Qihong Technology is a major supplier of overall thermal solutions, transitioning from air cooling to Vc plate cooling, and then to cold plate liquid cooling, being one of the suppliers for NVIDIA’s GB200 liquid cooling mode; Shuanghong Technology is a global cooling solution provider that has greatly benefited from the promotion of liquid cooling modules; Zhongshi Technology is a leading supplier for NVIDIA’s A100 and H100, occupying 30-40% of NVIDIA’s market share; Yingwei, Jingyan Technology, and others are the main suppliers for OEM liquid cooling plate modules; Xiangxin Technology is a leading supplier, with microchannel precision around 0.15, and Jingyan Technology around 0.1.
Q3: What are the challenges and advantages of microchannels, water cooling plates, and MLCP technology, which domestic companies are developing new technologies, and what changes can we expect in the future?
A3: Jingyan Technology and Xiangxin Technology are developing microchannel water cooling technology (MLCP technology). The advantage lies in meeting the cooling demands after the increase in server computing power, addressing the challenges of traditional liquid cooling plates in high-precision microchannel processing. Challenges include: traditional machining struggles to ensure microchannel precision; the leakage rate of traditional processes is high, making it difficult to control within 1%; traditional materials like aluminum perform poorly under high-temperature operation, while the process performance of copper alloys is also difficult to achieve. In the future, some auxiliary processes will penetrate the liquid cooling plate industry to improve precision and control leakage rates.
Q4: What processes might be used, and what role does 3D printing play in this technology?
A4: 3D printing technology has applications in micron-level liquid cooling plate technology, but currently has defects and cannot be introduced into mass production. Traditional 3D printing processes have poor density, leading to increased leakage; the 3D printing of connection technology has relatively better density but may crack under high thermal cycling. A mask forming process will be used to shape copper-tungsten alloy plates, using 3D printing technology to meet relevant needs. 3D printing technology plays a dominant role in this technology but needs to resolve issues of density and high-temperature cracking.
Q5: After the sintering process, do the densities of the two 3D printing processes meet the requirements, and how does the company solve related issues? In which areas will the powder stacking process be used?
A5: The density of the powder stacking process after sintering does not meet the requirements because the powder stacking method has gaps, and during the stress release process of the post-sintering treatment, if the stress is not fully released, it can lead to leakage when installed on the server. The company’s solutions include: controlling the precision and density of powder stacking; and adding tempering processes during sintering to adjust the bonding of material elements. The powder stacking process is suitable for 3C appearance parts (where strength is not mechanically required) and the medical beauty field (where density and strength requirements are not high).
Q6: How does the current process solve the issues faced by 3D printing, can the precision of 3D printing meet customer requirements, and what is the current yield level? How is yield defined?
A6: The 3D printing using connection technology can meet the precision requirements for 3C and liquid cooling plates, with no stress, and appearance defects can be compensated through machining; strength and mechanical properties can be restored through the sintering process. Currently, precision can meet requirements, and leakage rates can also be controlled, but the yield in small batch production is low, related to the precision of printing equipment. The yield for printed parts is about 85%, while the yield of finished products after sintering is about 65%. Yield definition involves the qualification of printed and post-sintering products, with issues such as deformation, flow channel hole deformation, or cracking affecting yield.
Q7: Is the current printing method a green light 3D printing method, and what is the dual-track emission process of laser melting and laser sintering?
A7: The mainstream printing methods theoretically include laser melting and laser sintering, which are dual-track emission processes. Traditional powder stacking processes involve a melting phase, while binder 3D printing uses binder stacking, resulting in poor initial mechanical properties. Laser melting and laser sintering occur simultaneously, enhancing the relative independence of the product before post-processing, allowing metal elements to fuse together, and subsequent high-temperature tempering can alleviate alarm rates, preventing metal elements from cracking in high-temperature and high-heat environments.
Q8: What is the landscape of domestic 3D printers, procurement situation, and user experience? How efficient is the current printing, how many times can it print in a day, how many can it print at once, and how long does it take to print one liquid cooling plate with a single head?
A8: Domestic companies like Huashu High-Tech and Bolite are relatively good. Bolite has a more comprehensive one-stop service for 3D printing than Huashu, while Huashu has made faster progress in liquid cooling than Bolite, which focuses on the 3C field. Currently, in the R&D phase, a 450 small single-head printing device is used, taking about 5 minutes to print one liquid cooling plate. Future mass production will use multi-head printing devices.
Q9: What are the benefits, drawbacks, efficiency, functionality, performance improvements, and customer recognition of using 3D printed liquid cooling plates? What is the rationale for adopting 3D printing technology?
A9: The benefits include fast speed, no stress, and lower manufacturing costs compared to machining; drawbacks include complex processes, post-processing requirements, and potential cracking and deformation issues. In terms of efficiency, there is an improvement compared to traditional processes, with a liquid cooling plate taking about 5 minutes to print. Regarding functionality and performance improvements, there is currently no specific data on the enhancement of cooling performance, which mainly depends on the cost of the liquid cooling plate rather than the process. Customer recognition levels have not been mentioned. The rationale for adopting 3D printing technology is that achieving micron-level microchannel precision with materials like aluminum through machining is relatively easy, but it is challenging with copper alloys or copper-tungsten alloys, and machining is inefficient when handling hundreds of microchannels, making 3D printing a better choice for achieving that precision and form.
Q10: Have the current printing routes and materials fully converged or stabilized? What are the prices of copper alloys and copper-tungsten alloys, and are the printed impeller plates made up of small pieces?
A10: The current printing routes and materials have not fully converged or stabilized, with representative materials being copper alloys and copper-tungsten alloys. The printed impeller plates are assembled from dozens of small pieces, not formed as a single piece. The price of copper-tungsten alloy is about 1500 per kilogram, but the specific price is uncertain.
Q11: How much labor time is required to produce a complete liquid cooling plate cabinet or a set of liquid cooling plates using existing processes, and how does it compare to the full lifecycle of traditional CNC processes?
A11: Compared to traditional CNC processes, producing a complete liquid cooling plate cabinet or a set of liquid cooling plates using existing processes can save about one-third of the labor time. The main reason is that the existing process has fast printing speeds, with time mainly spent on assembly; while CNC primarily consumes time in processing. CNC employs integrated processing, using mechanical control precision, with relatively fewer split parts, and does not use a splicing method.
Q12: Compare the product technology routes, material choices, and client connection progress of several domestic enterprises, including Nanfeng Co., Ltd. and its subsidiary Nanfeng Additive.
A12: Currently, mainstream enterprises have more connections with foreign clients, while domestically, progress is relatively slow. Apart from Jiyan Technology’s cooperation with Huawei, domestic enterprises are lagging behind in applications compared to overseas companies like NVIDIA. Nanfeng Additive mainly provides materials, primarily pure copper.
Q13: Does the five-minute single-key printing refer to SLM technology or binder jetting technology? Which company performs binder jetting technology better, and does the binder jetting process use metal powder combined with binder to form shapes before sintering?
A13: The five-minute single-key printing refers to binder jetting technology, which is a dual-track process combining laser melting and laser sintering. Huashu and Bolite perform relatively well in binder jetting technology, and the 3C field will combine 3D printing with this process. The binder jetting process uses metal powder combined with binder to form shapes, which are then sintered.
Q14: Does the assembly process of the composed parts require welding, and which generation of products can the 3D printing process be used for the fastest?
A14: The assembly process of the composed parts may involve welding, but traditional processes using welding can easily lead to leakage issues, while the 3D printing process forms the flow channels as a whole in one go, not through welding. The 3D printing process may be applied as early as 2026, but only for small batch production, not full mass production. Just like last year, liquid cooling was not fully introduced, combining air cooling and liquid cooling.
Q15: When did overseas major clients start validating this process, and will 30% of them use 3D printing technology next year? Is the plan confirmed to use the binder jetting copper-tungsten alloy process, who is the OEM supplier, and has the equipment selection been finalized, using which company’s equipment? Is it correct that ASUS supplies equipment to Jingyan, and Jingyan produces products for Foxconn?
A15: Overseas major clients started validating this process last year, and the technical plan is basically confirmed, using copper-tungsten alloy materials, with an expected 30% share using 3D printing technology next year. The OEM suppliers include Jingyan Technology, Foxconn, and others. The equipment selection is basically determined to be Huashu’s, as Huashu focuses on advanced service fields, while Bolite primarily focuses on the 3C field and is bound to Apple. ASUS supplying equipment to Jingyan and Jingyan producing products for Foxconn is correct.
Q16: What is the progress of Han’s Laser in liquid cooling 3D printing?
A16: Han’s Laser entered the liquid cooling 3D printing field relatively late, mainly focusing on the 3C field, with relatively slow progress in liquid cooling.
Q17: Can traditional machining methods for making fins meet the requirements for microchannels, and what is the current precision requirement in microns?
A17: Traditional machining methods for making fins can meet the requirements for microchannels, with diamond machine precision meeting the requirements. The current precision requirement is about 0.02mm, not at the micron level.
Disclaimer: The above content does not constitute investment advice, and any losses arising from using it as an investment basis are not the responsibility of the author.
Essential investment tools are available, such as the knowledge community, which updates daily (research summaries, meeting recordings) and other real-time information. Over 50,000 articles annually, join quickly by scanning the code.
If you want to become a qualified investor, a supportive tool is particularly important. This community will be a quality choice for you. Since its operation began, it has attracted some friends’ attention. For more investment information, research summaries, meeting minutes, and industry analysis reports, please pay attention to the QR code above for daily updates, a professional investment research tool that will ease your stock selection worries. A gathering place for fundamental enthusiasts; if you are a fundamental enthusiast and want to grasp real-time information across various industries, this may be suitable for you.