策略专题：千家公司年报，看AI在A股的“渗透率”

Group 1 - The penetration rate of AI in A-shares is still low, with TMT and large-cap companies leading the way [77][93][84] - The overall penetration rate of companies mentioning AI is less than 20%, with the highest rates in the TMT sector [93][124] - The number of companies mentioning AI has increased significantly from 172 in 2020 to over 1200 in 2023, but the growth rate remains slow [94][93] Group 2 - Most industries are in the application stage, with significant representation in beauty care, retail, and computing, while sectors like non-ferrous metals and social services are still in early layout and R&D stages [3][42][43] - Companies in the application layer focus on improving quality and efficiency, while those in the basic layer emphasize demand generation [46][49] - The expected effects of AI vary by industry, with cost reduction being a priority for sectors like textiles and light industry, while quality improvement is emphasized in sectors like machinery and home appliances [46][49] Group 3 - The majority of companies adopt a self-research approach to AI, with nearly 70% of companies engaging in self-research, while about 21% collaborate [23][24] - Over 75% of companies have released or promoted AI products, while nearly 25% are still in the R&D or testing phases [23][24] - The output effects of AI are primarily focused on quality improvement and efficiency enhancement, with only about 8% of companies reporting significant revenue growth [23][24] Group 4 - Future development directions for companies include expanding application scenarios, conducting technical research, and product development, with about 25% of companies focusing on these areas [23][35] - Companies in the basic layer are more inclined towards self-research, while those in the technical and application layers show a higher tendency for collaboration [24][37] - The capital expenditure of domestic AI leaders is significantly lower compared to their counterparts in the US and Hong Kong [28][40]