Summary of Key Points
Zhipu, which was listed on the Hong Kong Stock Exchange just over 5 months ago, has now initiated a plan to list on the STAR Market, aiming to raise 15 billion yuan to fund the development of large-scale models and service platforms. This move is not only aimed at diversifying its funding sources to cope with the high investment demands of the large-model industry but also reflects that the competition among domestic large-model companies has shifted from a focus on model capabilities alone to a comprehensive assessment of capital strength, commercialization capabilities, and long-term investment potential. At the same time, there is a clear divergence among the once prominent “AI Six Tigers” in the industry, with leading firms vying for dominance in the next round.
Detailed Analysis
#### 1. Why Is Zhipu Rushing to the STAR Market So Soon After Its Hong Kong Listing? — The Need for More Funds
The large-model industry is a classic example of a “money-burning” sector: In 2025, Zhipu’s research and development expenses amounted to 3.18 billion yuan, resulting in a net loss of 4.7 billion yuan. The 4.9 billion yuan raised from the Hong Kong Stock Market (with another 2.8 billion yuan remaining unused) is far from sufficient to support its long-term R&D efforts. The A-share market, however, offers higher valuations and larger financing amounts for technology companies. For instance, the STAR Market provides a more favorable environment for AI firms. With some previous barriers to listing in the A-share market having been removed, Zhipu intends to use a “A+H” dual-listing strategy to secure additional funding to ensure it has enough capital to sustain its operations—after all, in the large-model race, it’s about who can endure the longest.
#### 2. Where Will the Raised 15 Billion Yuan Be Used?
Of the 15 billion yuan raised, 80% (12 billion yuan) will be invested in “general-purpose foundation models,” which serve as the basis for all AI applications (such as ChatGPT’s GPT-4). The remaining 2 billion yuan will be used to develop a “MaaS” (Model as a Service) platform, allowing companies to access pre-made models or customize their own models effortlessly, just like using utilities. The final 1 billion yuan will be used for working capital to cover daily operations. This indicates that Zhipu still places a strong emphasis on core technologies while also seeking to generate revenue through its services.
#### 3. The Competition in the Large-Model Field Has Changed: From “Who Has the Strongest Model” to “Who Can Continuously Burn Money and Make Profit”
Two years ago, the focus was on which model had the largest number of parameters and higher evaluation scores. Now, it’s different: without sufficient capital (given the high costs of computing power and talent), companies cannot survive. For example, Zhipu’s MaaS platform generates an annual stable revenue (ARR) of 1.7 billion yuan, a 60-fold increase in just 12 months, with a gross margin of 41%. This shows that Zhipu not only has the ability to invest heavily but also generates profits. The industry’s focus has shifted to “comprehensive strength”—whether a company can afford to invest, convert technology into revenue, and sustain long-term investment.
#### 4. Clear Divergence Among the “AI Six Tigers”
The once prominent six companies (Zhipu, MiniMax, Moon’s Dark Side, etc.) are now taking different paths:
- Listing-Focused Companies: Zhipu (listed on both the Hong Kong and STAR Markets) and MiniMax (listed on the Hong Kong Stock Exchange) are seeking capital from the stock market.
- Funding-Focused Companies: Moon’s Dark Side (raised 500 million US dollars in its Series C round) and Jieyue Xingchen (raised nearly 2.5 billion US dollars) are still preparing for listings.
- Switching to New Tracks: Baichuan Intelligence (focusing on medical AI) and Lingyi Wanshu (focusing on AI applications) have moved away from developing general-purpose large models.
In the future, smaller companies may be acquired as capital flows towards the leading players, leaving them with insufficient resources to sustain their operations.
#### 5. Market Focus Has Cycled: From “Model Performance” to “Commercialization,” and Now Back to “Model Quality”
In 2023, people were concerned about whether models could chat or answer questions accurately; in 2024, the focus shifted to profitability; after 2025, it will return to model capabilities. Investors have realized that the model itself is a product, and only by continuously improving model quality can companies maintain their competitiveness. This means that leading firms must balance R&D with commercialization to emerge victorious in the battle for dominance.
In One Sentence
The large-model industry has entered a phase of comprehensive competition involving capital, technology, and commercialization. Zhipu’s dual-listing strategy exemplifies this trend, and the industry’s divergence will become even more pronounced. Ultimately, only a few top players will emerge as winners.