In September 2026, the A-share market exhibited a structural trend in AI concept stocks amid volatility. Although the overall index performance was mediocre, the AI sector was internally dynamic: the divergence between the two main tracks—computing power and applications—intensified, leading to subtle shifts in institutional capital allocation, with hot sectors switching from traditional computing power to AI applications. This divergence not only reflects different market expectations for the development of the AI industry but also reveals that the process of AI commercialization is accelerating.
I. Overview of AI Concept Stock Performance in September: Intensified Divergence, Led by Dual Tracks
According to Wind data, the AI concept stock index (884123.SH) fell by 1.2% in September, but the performance of individual stocks within the sector showed significant differences. The computing power sector (such as Gexun Technology, Innolight) declined by 3.5%, while the AI application sector (such as iFlytek, Kingsoft Office) rose by 2.8%. This divergence is not accidental but reflects the market's differing valuations of various segments of the AI industry.
1. Computing Power Sector: Short-term Adjustment, Unchanged Long-term Logic
Computing power, as the "infrastructure" of the AI industry, has always been a focus of market attention. The main reason for the adjustment in the computing power sector in September was the previous excessive gains and concerns over short-term supply and demand balance. However, in the long term, the demand for computing power continues to grow rapidly. According to IDC forecasts, the global AI computing power market size will reach $450 billion in 2026, with a compound annual growth rate of 35%. Domestically, the "New Blueprint for the AI Industry" released by the Ministry of Industry and Information Technology proposes that the core AI industry scale will reach 800 billion yuan by 2028, with computing power infrastructure being a vital support. Therefore, the adjustment in the computing power sector is more likely a short-term fluctuation and still holds long-term investment value.
2. AI Application Sector: Accelerated Commercialization, Becoming the New Main Track
In contrast to the computing power sector, the AI application sector performed brilliantly in September. Taking iFlytek as an example, its stock price rose by 5.2% in September, mainly due to the progress of AI applications in fields such as education and healthcare. Data shows that iFlytek's "AI + Education" solutions released in September have covered over 2,000 schools nationwide, and its "AI + Healthcare" image diagnosis system has received approval from the National Medical Products Administration. Furthermore, Kingsoft Office's AI office assistant "WPS AI" has surpassed 50 million users, gradually demonstrating its commercial monetization capabilities. These cases indicate that AI applications are moving from concepts to reality, becoming a new direction for institutional capital allocation.
II. Institutional Capital Movements: Shifting from Computing Power to Applications, Deploying New Logic
Fund holdings data shows a significant change in the holding structure of AI concept stocks by public funds in September. According to Tianxiang Investment Consulting statistics, the proportion of computing power sector holdings by public funds decreased from 12.5% in August to 10.8% in September, while the proportion of AI application sector holdings increased from 8.2% in August to 9.5% in September. This change reflects the deepening understanding of the AI industry by institutions: computing power is the foundation, but applications are the ultimate source of commercial value.
1. Reasons for Institutional Reallocation: Commercialization Realization is Key The main reason institutional capital is shifting to AI applications is the acceleration of commercialization. In the past, the market focused more on AI technological breakthroughs, but nowadays, companies are more concerned with whether AI can generate actual returns. For example, the large model company MiniMax announced in September that its AI assistant "Xiao M" surpassed 100 million users and achieved 200 million in commercial revenue, prompting institutions to include it in their core holdings. Furthermore, the penetration of AI applications in B-side (enterprise services) and C-side (consumer) is increasing, with growing demand in areas such as AI customer service and AI content creation, providing growth momentum to the application sector.
2. Hotspot Shift: From Hardware to Software, From Technology to Scenarios
In September, the hot topics among AI concept stocks shifted from hardware (computing power chips, optical modules) to software (AI applications, large models). For instance, on September 25th, large model concept stocks experienced a surge against the trend, with MiniMax rising by 20% in a single day, while the optical module leader OptiXun Technology fell by 3.2%. This shift indicates that the market is moving from "technology-driven" to "scenario-driven," focusing on whether AI can solve real problems rather than just focusing on the technology itself.
III. Industry Dynamics: Policy Reinforcement and Commercialization Inflection Point Support Market Divergence
In September, the AI industry received multiple policy benefits, and the commercialization inflection point emerged, supporting market divergence.
1. Policy Reinforcement: Open Collaborative Ecosystem Becomes a Keyword
The "New Blueprint for the AI Industry" released by the Ministry of Industry and Information Technology proposed building an "open collaborative" AI ecosystem, encouraging corporate cooperation and promoting the application of AI technology across different industries. This policy direction has driven the market to pay attention to the synergistic effects of AI applications, such as the integration of AI+Education, AI+Healthcare, and AI+Finance. For example, the "AI+Education" platform launched in collaboration between the Ministry of Education and iFlytek integrates resources from multiple educational institutions, enhancing the effectiveness of AI applications.
2. Commercialization Inflection Point: Decreasing Landing Costs of Large Models
In September, breakthroughs in domestic AI inference chips reduced the landing costs of large models. For example, the T20 inference chip released by Siwei Technology improved performance by 30% compared to the previous generation while reducing costs by 20%. This made AI large models affordable for small and medium-sized enterprises, promoting the popularization of AI among SMEs. Furthermore, the upgrade of AI application tools (such as Cursor and Copilot) allows users with no background to use AI programming, further expanding AI application scenarios.
IV. Investor Recommendations: Focus on Application Leaders and Invest in Long-Term Value
For investors, the divergence of AI concept stocks in September provides opportunities for positioning. It is recommended to focus on the following areas:
AI Application Leaders: Such as iFlytek, Kingsoft Office, and Yonyou Network, these companies have advantages in their respective fields, fast progress in AI application implementation, and strong commercialization capabilities.
AI+Scenario Integration: Companies in fields such as AI+Education, AI+Healthcare, and AI+Finance, these fields have strong policy support and clear demand growth.
Large Model Commercialization: Such as MiniMax and Zhipu AI, these companies have fast-growing large model user bases and increasing commercial revenue, possessing long-term investment value. It should be noted that AI concept stocks are highly volatile, and investors should focus on the fundamentals of the companies and avoid blindly chasing high prices. Furthermore, in the long run, the AI industry is still in its early stages, and technological breakthroughs and commercialization will be the core drivers of the future.
V. Conclusion: Differentiation is Opportunity, and Layout Requires Rationality
In September 2026, the differentiation among AI concept stocks reflects the market's maturity. The divergence between the computing power and application tracks reflects different stages of industrial development, while the changing logic of institutional capital deployment reveals the process of AI commercialization. For investors, focusing on leading companies in the application sector and positioning for long-term value will be key to seizing opportunities in the AI era. Supported by policy reinforcement and the commercialization inflection point, AI concept stocks still have significant investment potential, but short-term fluctuations must be viewed rationally, focusing on the fundamentals and long-term development potential of the companies.