Summary of Key Points
Chinese AI companies ZhipuAI and MiniMax have seen their stock prices soar by 16 times and 4 times respectively within 5 months of listing on the Hong Kong Stock Exchange, becoming what are known as “AI miracle stocks.” They have now both initiated A-share IPOs (Zhipu targeting the STAR Market, while MiniMax is aiming for the main board), with plans to meet each other on the A-share market. This news article analyzes the reasons behind their rapid stock price increases, their different business strategies, their motivations for returning to the A-share market, and the possibility of replicating the Hong Kong stock market’s success on the A-share platform.
I. The Wild Rise of “Miracle Stocks” on the Hong Kong Stock Exchange: Driven by Limited Circulation and Business Growth
Just how wild has been the rise of these two companies on the Hong Kong Stock Exchange? Zhipu’s stock price increased from HK$116 to a peak of HK$1993 (a 16-fold increase), with its market value briefly surpassing that of Xiaomi; MiniMax’s stock price rose from HK$165 to HK$840 (a 4-fold increase).
Why such sharp increases?
1. Limited circulation: Only a very small portion of the company’s shares were issued upon listing (for example, Zhipu may have only sold around 5%). With more buyers than sellers, the imbalance in supply and demand naturally drove up prices—just like when there are only 10 buns available and 100 people competing for them, the price will inevitably rise.
2. Actual business growth: This is not just pure speculation; both companies have seized the opportunity presented by AI agents (intelligent entities). For instance, Zhipu’s GLM-5 model, which is used for enhanced tool calls, has seen its API usage increase by 4 times, accompanied by a 83% rise in price (a simultaneous increase in volume and price). MiniMax has also surpassed 200 million overseas users, and its annual recurring revenue (ARR) doubled in just 60 days.
Why the sudden fluctuation on May 29th for Zhipu?
- Investors who had gained a 16-fold return began to lock in their profits and sell their shares.
- The stocks held by cornerstone investors were set to be released from lock-up, and they might sell them once the restriction is lifted (as these investors purchased at lower prices).
II. Two Different Business Approaches
Although both companies are AI large-model providers, their strategies differ significantly:
1. ZhipuAI: The Chinese version of Anthropic, focusing on B2B services
- Background: Leveraging Tsinghua University’s expertise, the company follows a “general large model + enterprise services” approach (B2B).
- Seizing the opportunity: As AI agents became popular, Zhipu launched the AutoClaw tool, allowing ordinary users to deploy intelligent entities. As a result, API revenue increased from 10% to 26%, and gross profit margin soared from 3% to 18%.
- Revenue structure: Enterprise services serve as a stabilizing factor; by 2025, enterprise-level large-model revenue is expected to reach HK$366 million (a 70% increase), while intelligent entity-related revenue will be HK$166 million (a 248% increase).
2. MiniMax: The king of Chinese AI companies going global for C-end markets
- Strengths: Specializes in audio/video models, targeting the global consumer market.
- impressive figures: By 2025, overseas revenue will account for 73% (US$57 million), serving 236 million users globally (a benchmark few Chinese AI companies can match).
- Rapid progress: Launched three flagship models this year, two of which are open-source. The upgraded Agent product, Mavis, enables multiple intelligent entities to work together, and it currently serves over a million enterprises/developers.
III. Returning to the A-share Market: To Fuel the AI Race
Why do these companies want to list on the A-share market after their initial success on the Hong Kong Stock Exchange? The main reason is a lack of funding—AI is an industry that requires significant investment. Training large models demands powerful computers, hiring top talent costs a lot, and building an ecosystem requires substantial capital. Listing on both platforms allows them to raise more money. For example, Zhipu has already raised funds on the Hong Kong Stock Exchange; by raising additional capital on the A-share market, it can gain a stronger position in the global AI competition.
Additionally, the current high interest in AI among A-share investors makes this a favorable time for them to list.
IV. Can the Hong Kong Success Be Replicated on the A-share Market?
Opportunities:
1. Solid fundamentals: Both companies have seen double-digit revenue growth (Zhipu by 131%, MiniMax by 159%), and their commercialization efforts are accelerating (with simultaneous increases in volume and price for Zhipu, and significant progress with MiniMax’s globalization strategy).
2. Continuing AI momentum: Global investment in AI is on the rise, which could support these companies’ stock prices.
Risks:
1. High valuations: Stock prices on the Hong Kong Stock Exchange have already reached very high levels (Zhipu’s market value once exceeded HK$88 billion). If A-share prices are set similarly to those on the Hong Kong market, investors may face the risk of buying into overvalued stocks; if the Hong Kong market declines, it could affect A-share prices as well.
2. Still in the red: Neither Zhipu nor MiniMax has turned a profit (MiniMax reported a loss of US$1.87 billion in 2025, with a net loss of HK$250 million after adjusting for special items) due to substantial R&D investments (US$250 million spent on research and development).
3. The sustainability of the AI boom: If interest in AI fades in the future, stock prices could decline.
In Conclusion
ZhipuAI and MiniMax are returning to the A-share market to secure more funding for their AI initiatives. Whether they can replicate the success seen on the Hong Kong Stock Exchange depends on the effectiveness of their models and their ability to commercialize their technologies. After all, money is just an entry ticket; it’s the technology and business strategies that will ultimately determine their success in this AI race, which is still in its early stages.