China Securities Co., Ltd: Edge AI drives hardware upgrades, AI glasses may become a tipping point by 2025.

China Securities Co., Ltd. released a research report stating that with the continuous iterative growth of large model capabilities, the differences between models are narrowing. Giants such as Meta, ByteDance, Xiaomi, among others, have begun to vigorously layout edge-side AI, competing for the entrance of AI Agents. The potential blockbuster of edge-side AI is expected to appear in 2024, with the cost curve of AI glasses significantly decreasing. 2025 is expected to be the year of its outbreak. In addition to smartphones, PCs, glasses, and headphones, the potential base of edge-side AI is huge, with home appliances, Siasun Robot & Automation, smart cars, educational and office equipment, toys, among others, benefiting from the trend of edge-side AI. AI embedding will bring about a wide range of hardware upgrades. Focus on computing power, connectivity, storage, power, and other segments. Key points from China Securities Co., Ltd. include: Giant companies successively entering the game with edge AI on the rise. Capital investment in cloud-side AI infrastructure is expected to significantly increase in 2023-2024, with the growth momentum expected to be maintained. Capital expenditures are used to build powerful computing and data storage capabilities, enabling the continuous iterative growth of large model capabilities, while reducing the differences between models. These technologies need to be applied and implemented on terminal devices, ultimately translating the commercial value of these applications/terminals, forming a complete closed loop from investment to realization. Currently, giants such as Meta, ByteDance, Xiaomi have started to vigorously layout edge-side AI, competing for the entrance of AI Agents. In addition to smartphones, PCs, glasses, and headphones, the potential base of edge-side AI is huge, with home appliances, Siasun Robot & Automation, smart cars, educational and office equipment, toys, among others benefiting from the trend of edge-side AI. AI embedding will bring about a wide range of hardware upgrades. Focus on computing power, storage, connectivity, power, and other hardware segments. The potential blockbuster of edge-side AI: smart glasses. The cost and selling price of AI glasses have fallen into a reasonable range, making 2025 the year of the outbreak of AI glasses. The AI glasses released by Meta are priced as low as $299, which is not much different from the price of conventional glasses (in RMB). According to The Verge data, as of May 2024, Ray-Ban Meta shipments have exceeded 1 million pairs, with a forecast of over 2 million pairs per year. With more players entering the market, 2025 is expected to be the outbreak year for the AI glasses market. According to Wellsenn XR, the global AI smart glasses market is expected to grow rapidly after 2025, with global sales expected to reach 80 million pairs by 2030. After Meta, internet companies, phone manufacturers, and many XR companies have successively laid out AI glasses. In the second half of 2024, various brands began to release a dense array of AI glasses, with brands such as Thunderbird, Rokid, Baidu, and Shanj releasing multiple new AI glasses, striving to surpass Ray-Ban Meta in terms of weight, price, AI performance, and driving the industry into a period of rapid growth. Hardware upgrades are on the rise, with a focus on edge-side computing power, connectivity, storage, power, and other segments. Edge-side AI is expected to replicate the capital expenditure growth seen on the cloud side, driving hardware upgrades in computing power, storage, connectivity, and power. 1) Computing power + connectivity: The core upgrade point for edge-side processors is computing power and connectivity, with different product applications and feasible model deployment methods emphasizing different upgrades. In the future, after the sales volume of a certain form reaches the level of millions, there may be customized and optimized processor product series introduced by manufacturers. Focus on SoC, MCU, ISP. 2) Storage: With added features, to store more firmware and code programs, as well as local inference computation of edge-side models, there is a trend of capacity expansion for DRAM, NAND, and NOR Flash in small AIoT devices. Focus on NOR Flash, storage-computing integration, among others. 3) Battery: Increasing the number of battery cells and energy density, as well as the popularization of fast charging, will optimize battery life and recharging experiences; Heat dissipation: With the increase in edge-side AI computing power, heat dissipation materials will see a definitive upgrade opportunity. In addition, attention should also be paid to branding/contract manufacturing, optics, sensors, and other segments. Risk Warning: Risk of macroeconomic environmental fluctuations; AI technology implementation falling short of expectations; intense industry competition; escalating trade disputes between China and the US.