电子行业2025年度中期策略：端侧AI继续升级，ASIC需求景气高企

Group 1 - The report highlights that large model technology continues to iterate, leading to an innovation wave in consumer electronics driven by AI integration [5][21][30] - Traditional consumer electronics have entered a phase of slow growth, with smartphones and PCs stabilizing in sales, while TWS devices are also experiencing low growth [15][19] - The development of model compression technology lays the foundation for deploying large models on edge devices, which offer low cost, high performance, and privacy advantages [6][30] Group 2 - The penetration rate of AI terminals is expected to rise rapidly as major companies like Google and Huawei launch AI-enabled devices, enhancing user experience [6][34] - The report predicts a compound annual growth rate of approximately 115% for high-end AI PCs and 32% for AI smartphones from 2023 to 2027 [6][52] - The introduction of system-level AI capabilities in smartphones and PCs is anticipated to set new standards in the industry, prompting other manufacturers to follow suit [36][40] Group 3 - The demand for ASICs is strong, driven by their cost-effectiveness compared to GPUs, leading many tech companies to develop their own ASICs [7][75] - The global market for ASICs is projected to grow from $6.6 billion in 2023 to $55 billion by 2028, with a compound annual growth rate of 53% [7][76] - OpenAI's recent decision to rent Google TPU for its products marks a significant milestone for ASICs, indicating their acceptance by leading AI firms [7][76] Group 4 - The report suggests focusing on companies within the edge AI and ASIC supply chains, maintaining an "overweight" rating for the electronics sector [8][9] - Specific companies to watch in the edge AI sector include Rockchip, Hengxuan Technology, and Espressif Technology, while those in the ASIC space include Chipone and Aojie Technology [8][9]