随着AI算力芯片功率密度突破1200 W，其局部热流密度高达300-500 W/cm²，这对热界面材 料(TIMs)提出严苛要求：总热阻需≤1 mm²K/W才能将界面温升控制在5 °C内。传统导热硅 脂、相变材料等热阻普遍＞4 mm²K/W，而且这些材料的导热率不足，厚度稍大则热阻快速增 大，难以满足需求。为了适应当前和未来AI芯片的散热需求，潜在解决方案包括液态金属和 碳基复合材料两大技术路线：液态金属虽热阻≈1 mm²K/W，但存在氧化、泄漏等问题；垂直 取向碳材料虽本征热导率＞1000 W/mK，但因材料刚性导致热阻＞10 mm²K/W。本研究突破 性地开发了超声-抛光两步处理工艺：首先通过超声处理使石墨片在介观尺度可控断裂，形成 兼具高热导率和柔性的微观结构；再经精密抛光优化表面平整度，显著提升界面接触性能。 最终制备的石墨-硅胶复合材料展现出极低总热阻（50 psi：1.8 mm²K/W），高体导热率，优 良压缩顺应性以及优异的热循环稳定性。该材料成功解决了传统碳材料压缩性与界面接触性 能的矛盾，其导热特性和应用水平极为接近液态金属的水平，为AI芯片散热提供了理想解决 方案。 文章亮点 1. 经超声 ...

随着AI芯片功耗迈入"千瓦时代"，一场围绕散热的革命正悄然加速，液冷正从一个"可选项"迅速变为数据中心的"必选项"。 综合瑞银与东吴证券近期发布的深度报告，市场似乎仍未完全消化这场变革的颠覆性。报告明确指出，市场普遍低估了向液冷过渡的速度 和规模，这并非一次普通的硬件升级，而是由AI芯片激烈竞争驱动的"即时、战略性必要之举"。 这一结构性转变不仅将催生一个高达311.91亿美元（约合人民币2205.23亿元）的超级周期，更在重塑全球供应链格局，特别是为国产厂商 打开了进入核心体系的大门。 "热墙"来袭：当风冷走到尽头，液冷即将迎来爆发 AI算力的爆炸式增长，正将芯片推向一道"热墙"。据瑞银及东吴证券报告梳理，AI加速器的热设计功耗（TDP）正以惊人的速度攀升。 英伟达的GPU产品路线图清晰地展示了这一趋势：从H100的700W，到Blackwell B200的1200W，再到预计明年推出的VR200的1800W- 2300W，以及2027年可能超过3600W的VR300，直至未来Feynman平台可能高达5000W-7000W的功耗。 芯片功耗的飙升直接导致机柜功率密度急剧攀升。一个GB200 NVL72机柜 ...

Core Insights - The liquid cooling market in China's intelligent computing centers is projected to reach 18.4 billion yuan in 2024, representing a 66% year-on-year growth, and is expected to further expand to 130 billion yuan by 2029, indicating a potential market explosion [1] - Cold plate liquid cooling is anticipated to become the mainstream cooling solution for data centers due to its higher efficiency compared to traditional air cooling methods, especially as GPU power design continues to increase [1][2] - 3D printing is expected to emerge as the optimal manufacturing technology for liquid cooling plates, offering advantages over traditional machining methods [2][4] Market Trends - The cold plate liquid cooling method is the most widely used, utilizing a closed cavity made of copper or aluminum to conduct heat without direct contact with the liquid, thus requiring less redesign of existing data center equipment [1] - The historical data indicates that cooling can account for up to 40% of a data center's energy consumption, highlighting the importance of efficient cooling solutions [1] Technological Advancements - 3D printing allows for complex flow channel designs that enhance cooling performance, overcoming limitations faced by traditional manufacturing methods [2][3] - Microchannel liquid cooling plates are becoming a new trend, with companies like Jinfu Technology successfully developing microchannel designs that improve heat dissipation [3] - The industry is transitioning from traditional machining methods to 3D printing for liquid cooling plate production, driven by the need for intricate designs and improved performance [3][4] Industry Developments - Companies are developing advanced 3D printing technologies to overcome the challenges of printing with copper, which has high reflectivity in the red light spectrum [4] - Several firms have successfully launched 3D printed liquid cooling plate products, demonstrating the feasibility and advantages of this technology in the market [4]

Core Insights - The report from Nomura Securities highlights the urgent need for advanced cooling solutions in AI chips due to rapidly increasing thermal design power (TDP) levels, with projections indicating that TDP for mainstream AI chips will rise from 600-700W in 2023 to potentially over 3500W by 2027 [3][4][10]. AI Chip Cooling Demand - The TDP of AI chips is expected to escalate significantly, with Nvidia's Blackwell series reaching 1000-1400W by 2025 and the Rubin series potentially hitting 2300W in 2026, and 3500W in 2027 [3][4]. - Traditional single-phase liquid cooling solutions are nearing their limits, necessitating new technological breakthroughs to handle TDPs above 2000-3000W [4]. Microchannel Cold Plates (MCL) - MCL is identified as the most practical solution for cooling chips exceeding 3000W post-2027, integrating heat spreaders and cold plates to reduce thermal resistance [5][7]. - MCL maintains compatibility with existing supply chains, utilizing current cooling fluids and components, unlike two-phase liquid cooling which requires extensive redesign [7]. - There are three main challenges to MCL mass production: design precision of microchannels, manufacturing capabilities, and supply chain coordination [8][9][10]. Thermal Interface Materials (TIM) - Upgrading TIM is crucial, with current materials like graphite films being insufficient for future TDPs; alternatives like indium TIM show promise but face challenges in assembly and interface treatment [10][11]. Other Technologies - Emerging technologies such as TSMC's Si integrated microcoolers and Microsoft's embedded microfluidics are considered less likely to be implemented in the short term due to scalability issues [11]. Market Opportunities for Traditional Cooling Manufacturers - Traditional cooling manufacturers like AVC and Auras are expected to see growth due to overlooked liquid cooling demands for non-core chips and the overall acceleration of liquid cooling adoption in AI servers [12][13]. - The market for liquid cooling components in AI servers is projected to grow from $1.2 billion to $3.5 billion between 2025 and 2027, with a compound annual growth rate exceeding 60% [12]. Investment Targets - Jentech is highlighted as a leading player in the microchannel market, with expected revenue from MCL contributing significantly to its overall earnings by 2028 [15]. - AVC and Auras are also recommended for investment, with AVC being a key supplier for Nvidia and Auras having advantages in manifold components [15].

Summary of Liquid Cooling Materials Industry Report Industry Overview - The report focuses on the liquid cooling materials industry, particularly in the context of increasing power consumption in AI chips and the demand for hybrid liquid cooling solutions [1][2]. Key Points and Arguments - **Server Cooling Demand Upgrade**: The power consumption of GPUs has increased from 1.2 kW to 1.8 kW, leading to total cabinet power consumption reaching 250-300 kW. Traditional single-phase cold plates are inadequate, necessitating an upgrade to liquid cooling solutions, with hybrid liquid cooling likely becoming mainstream, especially for high-power products like Rubin Ultra [1][2]. - **Advantages of Hybrid Liquid Cooling**: Compared to silent liquid cooling, hybrid solutions retain cold plates covering CPUs/GPUs, removing 60%-70% of heat, reducing the requirements for silent loops, enhancing system stability, allowing for single-point maintenance, and significantly lowering the amount of fluorinated fluids used, thus reducing costs [1][4]. - **Market Potential for Fluorinated Materials**: If hybrid liquid cooling solutions can be scaled, they could significantly drive demand for fluorinated materials, with a potential market size reaching 20 billion yuan, compared to the current demand of around 1,000 tons and a market capacity of 3 billion yuan [1][5]. - **NVIDIA Cooling Architecture Optimization**: The GB200 architecture uses one liquid cooling plate for two GPUs and one CPU, while the GB300 architecture provides individual cooling plates for each GPU and CPU, achieving more precise cooling and controlling total cabinet power consumption to approximately 132 kW. Higher power products like the Shadow series will require upgrades to bidirectional or phase change medium solutions [1][6]. Additional Important Insights - **Liquid Medium Choices**: Different cooling solutions utilize various liquid mediums. Single-phase cold plates typically use ethylene glycol, while bidirectional systems require low-boiling-point fluorinated fluids like R134A, hydrofluoroethers, and hexapropylene dimer [2][10]. - **New Zobon’s Liquid Cooling Strategy**: New Zobon is actively expanding in the liquid cooling sector, planning to add nearly 12,000 tons of capacity, including about 5,000 tons of perfluoropolyether capacity. The company aims to capture 50% of the semiconductor etching and temperature control market share [2][18]. - **Future Prospects for New Zobon**: The company expects its main business revenue to reach around 1.4 billion yuan by 2025-2026 and potentially 1.8-1.9 billion yuan by 2027. If the liquid cooling business materializes, the long-term market potential could reach 20 billion yuan, positioning the company for a valuation exceeding 20 billion yuan, with a short-term target market cap close to 50 billion yuan [2][20]. - **Challenges in Silent vs. Hybrid Solutions**: Silent solutions face high costs and reliability issues, while hybrid solutions enhance reliability and maintainability, significantly reducing the amount of medium required and overall costs [2][7]. - **Market Potential for Liquid Cooling Solutions**: The liquid cooling market has substantial potential, with estimates suggesting that if 100,000 cabinets are sold, the market space could approach 20 billion yuan, indicating significant growth elasticity compared to the current demand and market capacity [2][17]. This summary encapsulates the critical insights and projections regarding the liquid cooling materials industry, highlighting the evolving technology landscape and market opportunities.

Core Viewpoint - Nvidia is driving upstream suppliers to develop a new type of water cooling component called MLCP (Microchannel Liquid Cooling Plate) to address the increasing heat generated by AI GPU chips as technology advances [1] Industry Summary - The computing power and chip cooling industry is shifting from "technological breakthroughs" to "scale applications" [1] - As the computing power of AI chips increases, their power consumption grows exponentially, making cooling technology a necessity rather than an option [1] - The demand for cooling components is driven by the need for new data centers to achieve a Power Usage Effectiveness (PUE) of ≤1.25 as part of China's "East Data West Computing" initiative [1] - Cooling systems account for over 20% of total energy consumption in data centers, making them the largest energy source after IT equipment itself [1] Technology Summary - Microchannel water cooling technology is an advanced and efficient cooling solution, offering revolutionary advantages over traditional air cooling and conventional liquid cooling, especially for high-power, small-sized heat sources [1] Company Summary - Relevant A-share concept stocks mentioned include Aled and Plittech [1]

Summary of Liquid Cooling Materials Conference Call Industry Overview - The conference focuses on the liquid cooling materials industry, particularly in the context of AI chip cooling and data centers [1][2][4]. Key Points and Arguments Types of Liquid Cooling Media - Liquid cooling media are categorized into three types: water-based, oil-based, and fluorinated liquids, each with distinct advantages and disadvantages [1][3][5]. - **Water-based Media**: High thermal efficiency, low cost (around tens of yuan per kilogram), but prone to leakage and microbial growth [5]. - **Oil-based Media**: Cheapest option (around 10 yuan per kilogram), but poor fluidity and potential fire hazards [5]. - **Fluorinated Liquids**: High performance but expensive (high-end options like 3M perfluoropolyether cost 400-500 yuan per kilogram) and face regulatory risks regarding PFAS [1][6]. Trends in AI Chip Cooling - The trend in AI chip cooling is shifting from traditional air cooling to liquid cooling solutions due to high power density requirements [2][4]. - Single-phase cooling plates are being replaced by dual-phase cooling plates, which are expected to be widely adopted by 2024 [2]. Cost Considerations - The cost of cooling media is significant in system design. For example, a single cabinet requiring 120-150 kW can cost up to 1 million yuan when using high-end fluorinated liquids [1][10]. - Using alternative media like triol or silicone oil can significantly reduce costs [10][11]. Future Trends - The combination of full immersion cooling and dual-phase cooling plates is seen as a future trend to reduce costs and address high heat flux density issues [1][13]. - The industry is moving towards designs that minimize liquid usage and optimize heat dissipation through techniques like zoning and partitioning [1][13]. Market Dynamics - Major data centers in China, such as Alibaba and ByteDance, are currently using single-phase silent liquid cooling systems, indicating a trend towards advanced cooling technologies despite high costs [18]. - Alibaba is also exploring cooling plate technology to find more economical solutions [19]. Challenges and Innovations - Dual-phase cooling technology is gaining traction, especially with NVIDIA's upcoming Ruby architecture, but faces challenges such as medium boiling point limits and pressure handling [7][24]. - The use of refrigerants in dual-phase systems is expected to grow, with potential for mainstream adoption by mid-2026 [24]. Additional Important Content - The volatility and maintenance requirements of different cooling media vary, with silicone oil and fluorinated liquids requiring less frequent additions compared to triol [14][16]. - The efficiency of cooling media is measured by specific heat capacity, with water being the most efficient but limited in application due to its non-insulating properties [22]. - The domestic market for fluorinated liquids includes major manufacturers like Juhua, New Zhiyuan, and Tianjin Changlu, indicating a competitive landscape [21]. This summary encapsulates the critical insights from the conference call, highlighting the evolving landscape of liquid cooling technologies and their implications for the AI and data center industries.