What are the beneficial stocks of Edge AI/AI-RAN? In the past two years, AI infrastructure investment has mainly focused on GPUs, HBMs, data centers, optical modules, and power. These correspond to the large model training phase. But as the demand for inference continues to grow, AI is expanding from centralized cloud training to edge inference. Huang Renxun has consistently emphasized in recent GTC, earnings reports, and AI-RAN related events that AI will gradually enter scenarios such as operator networks, edge nodes, industrial sites, robots, and autonomous driving. The future AI architecture will gradually form a collaborative model of "cloud+edge+end". The cloud is responsible for training and centralizing computing power, the edge is responsible for low latency inference and real-time response, and the terminal is responsible for local perception and execution. AI-RAN is still in its early stages, but operators, equipment manufacturers, and chip companies have already begun laying out related ecosystems. Several directions worth paying attention to NOK Nokia is one of the participants in the AI-RAN Alliance and has collaborated with Nvidia in the AI-RAN and Cloud RAN directions. The company has long been deeply involved in the construction of wireless networks for global operators, and has an industrial foundation in areas such as Open RAN, Cloud RAN, and edge networks. At present, the market still mainly regards it as a traditional communication equipment company. If AI-RAN gradually enters the operator CapEx cycle in the future, Nokia's role in related infrastructure may be re priced. MRVL The relationship between Marvel and AI-RAN is more reflected in network silicon, data flow custom ASIC、 Bottom level capabilities such as edge network interconnection. The company is not an AI-RAN equipment provider, but it specializes in Cloud RAN, high-speed interconnection packet processing、 Operators have laid out infrastructure and other directions. The market generally views MRVL more as an AI data flow layer related company, and AI-RAN can be seen as one of its potential incremental application scenarios. ERIC Ericsson is one of the world's major RAN equipment manufacturers and a member of the AI-RAN Alliance. Continuously invest in Cloud RAN, Open RAN and other directions. If AI-RAN commercialization is promoted in the future, Ericsson theoretically has strong industry participation capabilities. However, the overall industry business model and CapEx rhythm are not yet fully clear and still need to be observed. ADI ADI belongs to an analog chip company. Edge AI, wireless communication, industrial equipment and other scenarios require RF, ADC/DAC, power management, and signal chain. With the increasing complexity of edge inference and wireless networks, the importance of high-performance analog devices may continue to rise. The relationship between ADI and AI-RAN is mainly reflected in the underlying simulation and signal processing stages. QCOM Qualcomm has a long-term accumulation in the fields of wireless communication, low-power SoC, edge AI, and terminal side inference. We have launched the X100 accelerator for Open RAN, while continuously laying out AI PC, edge AI, and 6G. In the past, the market focused more on its mobile business, but in the future, as the demand for edge inference grows, its capabilities in low-power AI and wireless connectivity may receive more attention. BB The core asset of BlackBerry currently is QNX, a real-time operating system that has been widely used in automotive, industrial equipment, embedded systems, and ADAS scenarios for a long time. With the development of Edge AI and Physical AI, the requirements for stability, low latency, and real-time response may increase. The market is paying attention to the potential application space of QNX in future edge smart devices, but the commercial scale and pace still need to be continuously observed. Let's briefly divide the risk return ratio High long-term industry relevance: NOK MRVL Relatively stable industrial status: ERIC ADI Relatively large difference in market expectations: QCOM BB It should be noted that AI-RAN and Edge AI are still in the early stages of development. There are many participants in the industrial chain, and the business model and industry pattern are still evolving. The future benefits of related companies depend on the operator CapEx, the speed of edge AI penetration, the growth of inference demand, and the progress of AI-RAN commercialization.