通信行业深度：黎明已经到来—从技术演进看国产算力投资机会

Investment Rating - The report maintains an "Accumulate" rating for the industry [1] Core Insights - The demand for computing power in China is at the dawn of a new era, driven by the rapid development of AI and the need for large-scale deployment of AI services by domestic internet giants [2] - GPGPU is expected to dominate the domestic computing power landscape in the coming years due to its usability and ease of deployment, compared to ASIC chips which require more complex software ecosystems [16][21] - The construction of autonomous and controllable AI communication capabilities in China is urgent, with a focus on both "Scale-Out" and "Scale-Up" networks [33][44] - China's infrastructure and manufacturing capabilities are seen as critical strengths in the development of domestic computing power, particularly in the context of energy supply and electronic manufacturing [74][106] Summary by Sections 1. Investment Requirements - The understanding of the development path for domestic computing chips is shallow among investors, with GPGPU being more suitable for the current market compared to ASIC [16][21] - The complexity of developing ASIC chips, particularly in terms of compiler and software ecosystem integration, poses significant challenges [16][21] 2. Domestic Computing Power - Usability and Ease of Use - GPGPU architecture is highlighted as the more appropriate path for China's computing power needs, allowing for rapid deployment and user acceptance [21][58] - The report emphasizes the importance of optimizing GPGPU architecture to enhance AI precision and performance [17][58] 3. The Path of AI Communication in China - The report outlines the need for a dual approach in AI communication, focusing on both "Scale-Out" networks led by Ethernet and "Scale-Up" networks that require a collaborative effort among chip manufacturers and telecom giants [33][44] - The establishment of a domestic version of "NV-Link" is deemed essential for enhancing AI communication capabilities [93][96] 4. Infrastructure and Manufacturing: The Backbone of China's Computing Power - China's electronic manufacturing capabilities and advanced power grid infrastructure are identified as key advantages in supporting AI computing power [74][106] - The report suggests that the domestic IDC sector will play a crucial role in the rapid deployment of computing power due to China's robust energy infrastructure [76][106] 5. Investment Recommendations - The report recommends focusing on core companies in four key areas: computing power (e.g., Cambrian-U), communication capabilities (e.g., ZTE, Shengke Communication-U), manufacturing capabilities (e.g., Xinyi Sheng, Zhongji Xuchuang), and infrastructure capabilities (e.g., Runze Technology, Guanghuan New Network) [35][106][107]