报告导读： 高密度算力中心不断发展，芯片层面的冷却技术愈发成为热管理系统的核心瓶 颈，我们认为导热界面材料（ TIM ）的需求将快速增长。 投资观点。 在面对人工智能和更新芯片设计不断增长的需求，当前的冷却技术将在短短几年内限制产业的不断发展。近日，微软宣布已成功开发出一种芯片 内的微流体（ Microfluidic ）冷却系统，可以对模拟团队会议核心服务的服务器进行有效冷却，散热效果比冷板好 3 倍。这个技术方案是将微小的通道直接 蚀刻在硅片的背面，形成凹槽，冷却液直接流到芯片上，实现更有效的散热。我们认为 AI 智算中心的热管理系统的核心瓶颈正不断聚焦于芯片层面，建议关 注 TIM1 、 TIM1.5 、 TIM2 等导热界面材料的需求增长，以及芯片内微流体技术的不断演进 。 AI 发展，热管理对完整散热系统的要求越来越高。 伴随 AI 大模型、 AI 应用的不断发展，其算力底座数据中心面临全新挑战以满足智算发展需求，数据中心 架构正加速向高密度、智能化和可持续方向演进。根据 TrendForce 表示， AI 服务器采用的 GPU 和 ASIC 芯片功耗大幅提升，以 NVIDIA GB200/GB300 ...

Core Viewpoint - The AI computing power sector is experiencing renewed attention driven by performance and policy support, with a focus on the sustainability of this trend in the long term [1][4]. Group 1: Market Performance - On August 27, the domestic computing power leader, Cambricon, saw its stock price peak at 1464.98 CNY, surpassing Kweichow Moutai, before closing at 1372.10 CNY, a rise of 3.24% [3]. - Cambricon's half-year report revealed a revenue of 2.881 billion CNY, a year-on-year increase of 4347.82%, and a net profit of 1.038 billion CNY, reversing a loss from the previous year [3]. - NewEase, another key player, experienced a stock price increase of 15% during the day, reaching a historical high of 324.42 CNY, before closing up 9.32% [3]. Group 2: Industry Trends - The AI computing power sector has seen significant growth across various sub-sectors this year, with companies like Shenghong Technology and NewEase experiencing year-to-date increases of 416.63% and 274.99%, respectively [4]. - Fund managers believe that the current market fluctuations are primarily driven by emotional responses, with profit-taking and portfolio adjustments being normal market behavior [4]. - The long-term outlook for the "selling shovel" model in the AI industry remains positive, with expectations of sustained high demand for computing power [4][5]. Group 3: Future Prospects - Fund managers anticipate that the demand for computing power will grow exponentially due to advancements in AI models and applications, with specific benefits expected for GPU and ASIC chips, optical modules, and related technologies [5][6]. - The AI industry is currently in a rapid expansion phase, with computing power as a foundational infrastructure expected to be a key beneficiary over the next three to five years [6]. - Despite challenges such as U.S.-China trade tensions, the domestic AI chip market is projected to grow, with the localization rate expected to rise from under 20% to over 40% [7]. Group 4: Policy Impact - The recent government policy document emphasizes the integration of AI across various sectors, aiming for widespread adoption of new intelligent terminals and applications by 2030 [8]. - The policy is seen as a comprehensive framework that will benefit multiple sectors, including big data, financial technology, and smart manufacturing [8][9]. - The focus on AI infrastructure, such as cloud services and data management, is expected to create a ripple effect throughout the industry, similar to the impact of mobile internet [9].

Group 1 - The AI computing power sector is experiencing renewed interest driven by performance and policy support, with companies like Cambricon and NewEase becoming market focal points [1][2] - Cambricon's stock price reached a peak of 1464.98 CNY per share, reflecting a nearly 10% increase during trading, but closed with a gain of 3.24%, valuing the company at 574 billion CNY [2] - NewEase also saw significant stock movement, peaking at 324.42 CNY per share with a 15% increase before closing up 9.32%, bringing its market cap to 306.4 billion CNY [2] Group 2 - Cambricon reported a staggering 4347.82% year-on-year increase in revenue for the first half of the year, totaling 2.881 billion CNY, and a net profit of 1.038 billion CNY, reversing a loss from the previous year [2] - NewEase attributed its revenue growth to sustained high industry demand, increased orders, and improved delivery capabilities, indicating a positive outlook for future industry conditions [2][4] - The AI computing power sector has seen substantial gains across various segments, with companies like Shenghong Technology and NewEase experiencing year-to-date increases of 416.63% and 274.99%, respectively [3] Group 3 - Fund managers express confidence in the long-term growth potential of the computing power sector, viewing it as a stable investment within the AI industry [3][4] - The upcoming release of the DeepSeek-R2 model is expected to stimulate demand for computing power, benefiting upstream sectors like chip design and wafer fabrication [5] - The recent government policy document emphasizes the integration of AI across multiple sectors, suggesting a broadening of the AI market beyond just computing power [6][7] Group 4 - The AI infrastructure, including cloud services and data solutions, is seen as a key beneficiary of policy support, with potential for significant industry transformation [7] - Fund managers recommend exploring underpriced segments within the technology growth sector, indicating that the current AI market expansion is still in its early stages [7]

Group 1 - The A-share market has been very active recently, particularly the computing power sector, which has shown strong performance since mid-August, with AI chips, AI servers, optical modules, and liquid cooling technology being standout hardware segments [1] - The strong performance of the computing power sector is driven by three main factors: first, the capital expenditure plans disclosed by major overseas cloud vendors exceeded expectations, maintaining high growth in global computing infrastructure investment; second, OpenAI's recent announcement of GPT-5 has attracted industry attention, with plans to double the scale of computing clusters in the next five months; third, feedback from the supply chain indicates robust demand for computing power, supported by new technological iterations that have spawned multiple high-growth sub-sectors [1] - Looking ahead, the demand for computing power is expected to grow exponentially due to breakthroughs in AI model performance, the evolution of multimodal technologies, and the rapid deployment of AI Agent applications, indicating a long-term high prosperity in the industry [1] Group 2 - Specific investment opportunities include GPU and ASIC chips, optical modules, optical chips, and fiber optic connectors, which are expected to continue benefiting from the growth in computing power [1] - Additionally, the rising demand for AI servers will lead to increased usage of PCBs, high-power power supply systems, and liquid cooling solutions, marking significant opportunities in these emerging fields [1]