Recently, with companies like Zhipu Huazhang, MiniMax, TianShu ZhiXin, and Biren Technology intensively listing on the Hong Kong Stock Exchange and the STAR Market, China's AI industry has officially entered a new phase of commercial validation and large-scale application. However, returning to industrial reality, the "bottleneck" dilemma faced by domestic chips still exists under the high ecological walls built by Nvidia. The significant stock price corrections of some already-listed GPU companies after substantial gains also reflect, to some extent, the market's scrutiny of their commercialization pathways and long-term growth logic. Since domestic chips find it difficult to match Nvidia's absolute computing power in the short term, the focus shifts to seeking superiority in system efficiency and scenario alignment. Recent announcements from chip companies and large model firms have emphasized "domestic adaptation," aiming to enhance computational efficiency through joint optimization and accelerate the application deployment of large models across various industry scenarios.
Industry consensus suggests that breakthroughs in single-point technologies are insufficient to win this competition; ecological synergy, particularly the "mutual advancement" between models and chips, is becoming critical for the genuine autonomy of domestic AI. CICC points out that reflecting on 2025, global large model technological capabilities advanced, gradually overcoming productivity scenarios and making significant progress in reasoning, programming, agentic capabilities, and multimodality, yet shortcomings remain in the stability and hallucination rates of general model capabilities. Looking ahead to 2026, CICC believes large models will achieve more breakthroughs in areas like reinforcement learning, model memory, and context engineering, progressing from short-context generation to long chain-of-thought tasks, from text interaction to native multimodality, moving closer to the long-term goal of achieving AGI.
GF Securities previously opined that as the commercial value of AI-assisted programming tools in enhancing enterprise R&D efficiency and optimizing business operations becomes increasingly recognized, corporate users' willingness to pay for these tools in scenarios like software development, data analysis, and business process automation is expected to strengthen. Domestic AI large models are anticipated to benefit from this trend, achieving successful commercial implementation. Hong Kong-listed stocks related to AI large models include: MINIMAX-WP(00100), KNOWLEDGE ATLAS(02513), KUAISHOU-W(01024), among others.
Comments