ZhiXing Column · 2025-06-22

Startup Commentary”Six “Little Dragons” Compete for the First Listing in the Large Model Sector”

Read More《六小龙开抢大模型第一股》

Positive Comments: Policy Dividends and Capital Support Present Key Opportunities for Leapfrog Growth of Large Model Startups

In the large model race in 2025, the news that the “Six Little Dragons” are collectively sprinting for IPOs is undoubtedly one of the most attention – grabbing events in the industry. Behind this battle for the “first stock,” there are significant policy benefits, which also reflect the survival wisdom of large model startups under the pressure of technological iteration and commercialization. Moreover, it may inject new vitality into the capital ecosystem of the entire AI industry.

First of all, the policy relaxation has opened a “green channel” for unprofitable large model startups to go public, which is the core driving force behind this round of IPO wave. According to the information from the 2025 Lujiazui Forum mentioned in the news, the China Securities Regulatory Commission clearly stated that a new board will be established on the Science and Technology Innovation Board, restarting the application of the fifth set of standards for unprofitable enterprises to go public, with a focus on supporting artificial intelligence enterprises. This policy adjustment directly addresses the pain points of the large model industry – continuous high – investment is required for technological R & D (for example, the cost of a single pre – training reaches three to four million US dollars), but the commercial revenue cannot cover the costs in the short term (currently, the annual revenues of the Six Little Dragons are all less than one billion yuan). In the past, the profit threshold was the biggest obstacle for AI startups to go public (for example, Megvii among the “Four Little Dragons of AI” failed to go public multiple times due to profit issues). Now, the policy allows enterprises with “major technological breakthroughs and broad commercial prospects” to go public, which is equivalent to providing the Six Little Dragons with an opportunity to “exchange the future for the present.” This can not only relieve the financial pressure on enterprises but also attract more long – term investors to pay attention to the AI hard – tech field through the recognition of the capital market.

Secondly, the title of the “first large model stock” has a significant “winner – takes – all” effect and is of far – reaching significance for reshaping the industry landscape. Referring to the era of the “Four Little Dragons of AI,” SenseTime, as the first enterprise to go public, although it suffered a loss of 24.272 billion yuan in three and a half years, it obtained continuous capital infusion through the premium effect of being the “first stock” (with a market value of over HK$320 billion the day after listing) and finally remained competitive in the large model era. On the other hand, Megvii, Yuncong and other enterprises that did not go public in time were forced to shrink their businesses due to capital chain pressure. Currently, the large model race is also in the early stage of “technology – led + capital – driven.” Whoever can go public first can not only obtain a higher valuation (such as the high premium of SenseTime in the early stage of listing) but also become the pricing anchor point for the industry, squeezing the living space of subsequent listed enterprises. More importantly, the brand effect brought by going public can attract top talents, high – quality customers and ecological partners, forming a positive cycle of “technology – capital – market.” For example, if one of the Six Little Dragons enterprises rings the bell first, its B – end customers may be more inclined to choose the large model services “endorsed by a listed company,” thus accelerating the commercialization process.

Finally, the IPO wave reflects the strategic upgrade of large model startups from “burning money for R & D” to “capital operation,” which contributes to the maturity of the industry ecosystem. In the past three years, the core task of the Six Little Dragons was to catch up with the technological iteration of large models (such as parameter scale and multi – modal capabilities), and the funds mainly relied on primary market financing (for example, Zhipu raised nearly 7 billion yuan in 2024). However, since 2025, the technological layout of large enterprises (Tencent, Alibaba, ByteDance) has accelerated, and emerging players (such as DeepSeek) have continuously made technological breakthroughs. The difficulty of primary market financing has increased significantly (the single – round financing amount of Zhipu in 2025 has dropped to less than 1 billion yuan). Turning to the secondary market at this time is not only an inevitable choice to address the capital gap but also marks the transition of the industry from “wild growth” to the stage of “standardized competition.” Through going public, enterprises need to disclose financial data and strategic plans, which will force them to optimize their governance structures and improve transparency, providing reference samples for subsequent industry standard setting (such as large model technology evaluation and commercialization paths).

Negative Comments: Profitability Issues and Intense Competition Pose Multiple Hidden Risks on the IPO Road of the Six Little Dragons

Although policies and capital have sent positive signals, the IPO road of the Six Little Dragons is not smooth. From the current industry situation, problems such as unclear profit models, weak commercialization capabilities, and the squeeze from large enterprises may become key hidden dangers that put pressure on the stock price after listing or even lead to IPO failures.

The most prominent issue is the “profitability hard – hit” – the huge gap between revenue and cost is difficult to bridge in the short term. As mentioned in the news, the annual revenues of the Six Little Dragons are all less than one billion yuan. Zhipu lost about 2 billion yuan in 2024, and OpenAI’s cost in 2024 reached as high as 8.5 billion US dollars (including training, inference, and labor costs). The R & D investment in large models has a “rigid” characteristic: continuous pre – training is required for technological iteration (with a single – round cost of millions of US dollars), and the costs of computing power leasing (such as NVIDIA A100 chips) and labor (the annual salary of top algorithm engineers exceeds one million) remain high. On the revenue side, users have a low willingness to pay for C – end products (chatbots), and the B – end API services are caught in a “price war” (large enterprises have pushed the prices to rock – bottom through cloud service packages with discounts). Even some B – end customers (such as hospitals) sign contracts out of “political performance needs” rather than actual application, resulting in doubts about the quality of revenue. Even if they go public through policy relaxation, this “high – input, low – output” model may face a valuation correction due to doubts about the “sustainable operation ability” (for example, the stock price of SenseTime dropped from HK$9.7 to less than HK$2 after listing).

Secondly, the intense competition within the industry and the squeeze from large enterprises have exacerbated the commercialization difficulties of the Six Little Dragons. The current large model race has entered a dual – track competition of “giants + startups”: large enterprises (such as Alibaba and Baidu) rely on their old cloud service customers and ecological advantages to seize the B – end market through bundled sales of “cloud + large models”; emerging technology – driven players (such as DeepSeek) lower the API prices with their technological leadership. The Six Little Dragons have neither a customer base for cloud services nor the ability to comprehensively surpass in terms of technical parameters. They can only maintain their business data by “subsidizing the development of small and medium – sized customers,” and this “losing – money – making – noise” model is unsustainable. For example, although Baichuan Intelligence focuses on the medical field, after Huawei established its healthcare business group, its original hospital customers may be snatched away by large enterprises with more comprehensive technology and stronger resources; the departure of the person in charge of MiniMax’s B – end business also exposes the team’s shortcomings in commercialization. If they cannot prove their “differentiated competitiveness” before going public, even if the Six Little Dragons succeed in going public, they may face the embarrassing situation of “peaking at the time of listing.”

Finally, the risk of IPO failure and the pressure from investors may backfire on the development of enterprises. Referring to Megvii among the “Four Little Dragons of AI,” after multiple failed attempts to go public due to profit issues, not only did the difficulty of financing increase sharply, but it also had to cut its original business (such as intelligent driving), suffering a great setback. Most of the current investors of the Six Little Dragons are early – stage institutions, and “gambling clauses” (such as listing time limits) are commonly set in investment agreements. If the IPO fails, the enterprises may face the pressure of equity repurchase and even a broken capital chain. In addition, the valuation logic of secondary – market investors for AI enterprises has shifted from “technological stories” to “implementation ability” (for example, AI chip enterprises were re – valued after 2024 due to their commercialization progress). If the Six Little Dragons cannot clearly demonstrate the “revenue growth path” and “profit timetable” in their prospectuses, they may encounter problems such as insufficient subscription and the breaking of the issue price, further weakening market confidence.

Suggestions for Entrepreneurs: Focus on “Implementation Ability” and Balance the Listing Rhythm with Long – term Value

Facing the temptation and challenges of the “first large model stock,” the Six Little Dragons and other large model startups need to maintain strategic clarity and focus on the following three aspects:

  1. Focus on commercialization and build a “verifiable revenue growth story”: The core of going public is to prove the “sustainable operation ability” to the market. Therefore, enterprises need to shift from a “technology – oriented” to a “demand – oriented” approach. It is recommended to give priority to vertical tracks with “high willingness to pay + high scene matching degree” (such as medical, legal, and industrial quality inspection) to avoid direct competition with large enterprises in the general API market. At the same time, optimize the customer structure, reduce ineffective orders driven by “political performance,” and increase paying customers who can “actually reduce costs and increase efficiency” (for example, bind long – term contracts through SaaS – based services). For example, Baichuan Intelligence can deepen its cooperation with hospitals on “AI – assisted diagnosis” and shift from single – project charging to revenue sharing based on usage; Jieyue Xingchen can focus on the “intelligent cockpit agent” in the automotive industry and build a data closed – loop with automobile manufacturers to enhance customer stickiness.
  2. Optimize the cost structure and enhance the visibility of “self -造血 ability”: The high cost of large models is the core concern of the market. Enterprises need to reduce costs through technological optimization (such as model compression and improved inference efficiency) and improved operational efficiency (such as reducing inefficient marketing and focusing on core business). For example, Yuezhianmian can suspend its aggressive C – end marketing and shift its resources to B – end customer operation; MiniMax can reduce the inference cost by signing long – term agreements with computing power service providers. At the same time, disclose in detail the “cost reduction path” in the prospectus (such as a technical plan to reduce the inference cost by 30% in 2026) to enhance investors’ confidence in profitability.
  3. Reasonably plan the listing rhythm and avoid “going public for the sake of going public”: Going public is a means rather than an end. Enterprises need to choose the appropriate capital market according to their own development stage (for example, the Hong Kong stock market is more inclusive of unprofitable enterprises, and the Science and Technology Innovation Board is more friendly to “hard – tech” labels). If they cannot meet the listing conditions in the short term (such as having been established for less than three years), they can first supplement funds through strategic financing (introduce industrial capital, such as cooperate with large enterprises to develop vertical models), and at the same time, use the policy window to standardize finance and governance in advance (such as establishing an independent board of directors and improving the information disclosure system). In addition, they need to communicate with investors to avoid blindly accelerating the listing process due to “gambling pressure,” which may lead to subsequent operational risks.

In summary, the battle for the “first large model stock” is both an opportunity and a challenge. If the Six Little Dragons can take the listing as an opportunity to shift from “burning money for R & D” to a dual – wheel – driven model of “technology + business,” they can not only win survival space for themselves but also explore a feasible path of “capital empowering hard – tech” for the Chinese large model industry. On the contrary, if they only regard going public as a “blood – transfusion tool” and ignore the construction of core capabilities, they may repeat the mistakes of some enterprises among the “Four Little Dragons of AI” and be eventually eliminated by the market.

ZhiXing-AIx
Chatbot