Key Takeaways
- CMSI initiates KNOWLEDGE ATLAS and MINIMAX-W at Overweight.
- Forecasts 80.7% global CAGR for large models through 2029.
- Sees unstoppable AI Agent commercialization in China's market.
Key Takeaways

China's push toward the commercialization of artificial intelligence agents is an "unstoppable" force, one that will drive the global large model market to $206.5 billion by 2029 and has prompted China Merchants Securities International to initiate coverage on two key AI pure-plays.
"The commercialization flywheel of Agents in China’s large model industry is unstoppable," CMSI said in a new report, highlighting the exponential growth in model usage as a key investment driver.
The report projects an 80.7% compound annual growth rate for the global large model market from 2024 to 2029. Based on this outlook, the broker set a HKD 1,282 price target for KNOWLEDGE ATLAS (02513.HK) and HKD 982 for MINIMAX-W (00100.HK), both with "Overweight" ratings.
The bullish ratings position these focused AI firms as prime beneficiaries of a market where architectural efficiency and multimodal capabilities are becoming more critical than raw benchmark performance, challenging established giants like Alibaba and ByteDance.
The scale of China's AI adoption is accelerating rapidly. According to the CMSI report, which cites Frost & Sullivan data, China's average daily token consumption—a measure of AI model usage—surged from 100 billion in early 2024 to 140 trillion in March. This thousandfold increase in under two years underscores the explosive demand for AI processing.
This boom is part of a larger trend that CMSI expects will see China's overall AI market reach RMB 993 billion by 2030, with the enterprise-level large model segment growing at a compound annual rate of 64%. While the U.S. currently leads in foundational algorithms, the report notes that the gap between advanced Chinese and American models is "rapidly narrowing," creating a two-way global rivalry.
CMSI's Overweight rating on KNOWLEDGE ATLAS comes with a HKD 1,282 price target, corresponding to a valuation of 68 times forward price-to-sales for fiscal 2027. The broker's thesis centers on the company's focus on pursuing the upper limits of model intelligence and strengthening its technological barriers in programming capabilities.
The firm forecasts a revenue CAGR of 172% for KNOWLEDGE ATLAS between 2025 and 2028. During this period, its adjusted net loss margin is expected to contract significantly, from an estimated -439% in 2025 to -14% by 2028, as it scales its commercial operations.
For MINIMAX-W, CMSI set a price target of HKD 982, implying a forward price-to-sales multiple of 43 for fiscal 2027. The broker is positive on the company's market position, which is built on in-house development and "extreme cost efficiency."
CMSI projects an even faster revenue CAGR of 193% for MINIMAX-W between 2025 and 2028, with the upcoming major iteration of its M3 model seen as a key catalyst. The company's adjusted net loss margin is forecast to narrow sharply from -317% in 2025 to just -12% by 2028.
The report suggests that while large, diversified tech firms like Alibaba and ByteDance are major players, the focused, neutral approach of pure-play companies like KNOWLEDGE ATLAS and MINIMAX-W offers a distinct investment thesis. Investors are now tasked with weighing the high-growth, high-multiple valuations against the execution risk as these companies race to capture a share of China's burgeoning enterprise AI market.
This article is for informational purposes only and does not constitute investment advice.