以下文章来源于睿兽Pro ，作者Bestla 睿兽Pro . 创业邦旗下横跨一二级市场的科创数据平台。实时投资数据、追踪产业创新。找数据、做分析、链资 源，就上睿兽分析。 行业定义 光模块（ Optical Transceiver Module，简称Optical Module）是一种利用光信号实现高速数据传 输的光电器件。它通过电信号与光信号的相互转换，结合光纤传输介质，为通信设备提供稳定、高速 和远距离的数据交互能力。 光模块并不等同于整个光通信系统，两者在功能定位上存在显著差异。光模块的核心是完成电信号与 光信号的转换，是光通信网络的关键接口；而光通信系统则涵盖传输、交换、处理等更为复杂的整体 架构。可以认为光模块是光通信的基础。 从使用场景来看，光模块已经渗透到多类网络设备中。在数据中心，光模块承担服务器与交换机之间 的大规模高速互联；在电信网络中，它为基站、核心网与骨干网提供稳定传输；在企业园区和超算中 心，光模块则保证低时延与大容量通信需求。 从技术构成来看，光模块主要由光芯片、电芯片、封装和散热系统组成。光芯片实现电 —光/光—电 转换，电芯片负责信号调制、驱动和放大，封装工艺决定其传输速率与功 ...

格隆汇9月26日丨木林森(002745.SZ)在投资者互动平台表示，公司对前瞻性技术布局始终持积极态度， Micro-LED与LED产业的强关联性使其天然具备成为行业头部的潜力。在光互联这一交叉领域，公司不 排除未来与光通信领域的领先企业展开合作，通过优势互补——例如公司提供高精度封装能力，合作伙 伴贡献光模块经验——共同推动技术落地。这种开放合作模式既能加速产业化进程，又能确保公司在新 兴赛道中占据关键位置，为长期增长注入新动能。 ...

Summary of Conference Call Notes Industry or Company Involved - The discussion primarily revolves around the optical interconnect industry, focusing on technologies such as CPU, OCS (Optical Circuit Switching), and OIO (Optical Interconnect). Core Points and Arguments 1. **Market Sentiment on Quantum X800 CPU**: Following the release of the Quantum X800 CPU switch by Nvidia, market enthusiasm has waned due to slower-than-expected product rollout, currently in the customer validation phase without large-scale commercialization [1][3][4]. 2. **Challenges Facing CPU Solutions**: The CPU solutions face maintenance issues and production yield challenges, requiring power shutdowns for maintenance, which increases costs and complexity. Additionally, competition from Google’s OCS poses a threat, with some cloud service providers showing insufficient recognition of CPU solutions [1][6]. 3. **Power Consumption Reduction**: The CPU solution has achieved only about 30% power reduction in data center network systems, which is below initial expectations and has not gained widespread customer acceptance. In contrast, OCS solutions can achieve nearly 90% power savings [1][7][8]. 4. **Cost Comparison**: OCS solutions demonstrate clearer cost advantages over CPU solutions, which are hindered by maintenance and yield issues. Despite potential advantages, CPU commercialization faces significant barriers [1][9]. 5. **Future of OIO Technology**: OIO technology is seen as a crucial method for GPU interconnects, but its necessity is debated within the industry, as it primarily addresses performance rather than core power consumption and cost issues. The commercial viability of OIO is expected to lag behind that of CPU [1][10][12]. 6. **Current Interconnect Methods**: The existing interconnect methods rely on PCB copper wire connections, which have limitations. The future direction is shifting towards OIO to enhance interconnect scalability and overcome current bottlenecks [1][11]. 7. **Market Outlook for O2O Technology**: O2O technology is viewed as a promising new market with greater bandwidth potential, significantly expanding the existing market size despite potential value loss during the transition from optical modules to CPU and OIO [2][15]. 8. **Role of Domestic Optical Module Companies**: Domestic optical module companies are evolving beyond simple assembly to become semiconductor-oriented firms with capabilities in high-precision coupling and optical system design [2][16][17]. 9. **Survival of Passive Component Manufacturers**: Passive component manufacturers are expected to have a higher survival rate amid technological changes, as products like fiber arrays and connectors will continue to be relevant [2][18]. 10. **Performance of Leading Companies**: Leading companies in the industry exhibit higher survival rates and market shares compared to second-tier firms. Their strong adaptability and semiconductor positioning are expected to enhance their valuations, particularly in the current AI-driven market cycle [2][19]. Other Important but Possibly Overlooked Content - The transition to OIO technology is anticipated to occur after the successful commercialization of CPU technology, with OIO potentially entering the market around 2029-2030 [1][12]. - The value distribution in the optical module production process indicates that semiconductor components account for 50% of the total value, while optical components and light source modules account for 30% and 20%, respectively [1][14].

Summary of Conference Call Records Industry Overview - The cloud service industry is experiencing significant capital expenditure (CAPEX) growth, with major players like Meta and Google expected to spend $72 billion and $85 billion respectively in 2023, requiring over 10% quarter-on-quarter growth in the second half of the year [1][2] - The optical module industry saw a quarter-on-quarter growth of 30% in Q2, with leading companies achieving up to 50% growth, indicating a strong demand driven by cloud service investments [1][3] Key Points and Arguments - **CAPEX Growth**: North American cloud providers' total CAPEX reached approximately $95 billion, a year-on-year increase of 66.59% and a quarter-on-quarter increase of over 30% [2] - **Microsoft's Performance**: Microsoft reported a Q2 CAPEX of $24.2 billion, a year-on-year increase of over 50%, with expectations to reach $30 billion in Q3, indicating a robust growth trajectory in cloud services [6] - **Future Industry Outlook**: The industry is expected to maintain a healthy growth rate of around 30% over the next few quarters, with potential for further upward revisions in performance forecasts [7] - **Optical Interconnect Market**: The optical interconnect market is projected to achieve a compound annual growth rate (CAGR) of approximately 20% from 2027 to 2030, driven by the expansion of single-node computing scenarios [8] - **AI Investments**: Google is heavily investing in AI applications, achieving a monthly growth rate of nearly 40% in AI token processing, indicating a closed-loop formation in AI applications that stimulates demand for training and inference computing power [1][13] Additional Important Insights - **Domestic IDC Valuation**: The domestic IDC sector shows reasonable valuations, with companies like Runze and David having price-to-earnings ratios around 25 times, while Data Harbor and Guanghuan Xinwang are around 15-16 times, suggesting opportunities for investment in undervalued companies [3][16] - **Technological Trends**: Key technological directions include GEL up, OIO, and CPU developments, with significant increases in fiber optic connections anticipated in future applications [10][12] - **Investment Opportunities**: The current IDC market presents stable investment opportunities, particularly in companies like Guanghuan and Data Harbor, which have a high safety margin [17] - **Concerns Over PCB Technology**: There are market concerns regarding the impact of orthogonal backplane PCB technology on the copper cable sector, but advancements in high-speed copper AEC are expected to maintain demand in data centers [19] This summary encapsulates the critical insights from the conference call, highlighting the growth trends, investment opportunities, and technological advancements within the cloud and optical interconnect industries.

Core Insights - The article highlights the significant advancements in China's AI capabilities, particularly through Huawei's development of the Ascend 384 super node, which boasts a bandwidth of 1229 TB/S, marking a shift away from reliance on U.S. technology [1][6] - The U.S. has implemented new tariffs and restrictions on semiconductor exports to China, aiming to isolate China from the global tech supply chain [3][5] - China's AI industry is experiencing a collective breakthrough, with Alibaba's international AI service usage skyrocketing to over 1 billion calls per day, a 1000-fold increase from 2023 [10] Group 1: U.S. Tariffs and Restrictions - The U.S. Treasury announced new sanctions, including a 50% tariff on TSMC's chip exports to mainland China and restrictions on NVIDIA's H20 chip sales [1][3] - The tariffs are expected to increase annual household spending in the U.S. by $2,100 to $3,800, disproportionately affecting low-income groups [3] Group 2: Huawei's Technological Advancements - Huawei's Ascend 384 super node utilizes a light interconnect network with 3168 optical fibers, achieving a 107% performance improvement over NVIDIA's H100 and reducing energy consumption for large model inference by 18% [6][8] - The project took 720 days to develop, resulting in 128 patent applications, with 73 already certified by the international PCT [8] Group 3: China's AI Industry Growth - The AI industry in China is implementing a "three-line breakthrough" strategy, focusing on technology, application, and industry lines [10] - The upcoming World Robot Conference in Beijing will showcase advancements from 50 humanoid robot companies, indicating a growing interest in robotics [10] Group 4: Emerging Economic Trends - The low-altitude economy is emerging as a new growth point, exemplified by a $1.75 billion order for 500 eVTOL aircraft signed between Volant and a Thai company [12] - The mass production of the Ascend 384 is expected to reduce AI model training costs by 40%, with application prices potentially halving in the next three years [13] Group 5: Job Creation and Investment Opportunities - The rise of eVTOL technology is projected to create new job opportunities, including drone operators and air traffic managers, with demand expected to exceed 1 million positions by 2030 [14] - The acceleration of domestic substitution in sectors like semiconductors, AI, and robotics presents potential investment opportunities for companies with core technologies [16]