Two AI Unicorns Set to List in HK: Same Stage, Divergent Paths
According to reports on December 22, Zhipu AI and MiniMax have recently passed the listing hearing of the Hong Kong Stock Exchange and plan to be listed on the Main Board of the Stock Exchange of Hong Kong Limited as “specialist technology companies”.
In 2025, when the capital market’s valuation of large model enterprises is becoming increasingly rational, the simultaneous debut of these two Chinese AI unicorns has attracted extraordinary attention.
Although both belong to the large model track, boasting top-tier R&D teams and considerable market valuations, their strategic positioning, technical routes, business models and development paths show distinct differences. This capital drama of “sharing the same stage but taking different paths” reflects the new pattern of diversified development in China’s AI industry.
Zhipu AI: Open-Source Ecosystem & AGI-Oriented Vision
In its prospectus, Zhipu AI clearly positions itself as “a leading artificial intelligence company in China, dedicated to developing advanced general-purpose large models”. Since its establishment in 2019, the company has taken “pursuing general artificial intelligence (AGI) innovation in China” as its vision, focusing on building an AI model system that can deeply think, perceive the world and use tools like humans.
Meanwhile, Zhipu AI has established a distinct open-source strategic positioning since its inception.
The GLM series of large models it launched, ranging from the 13-billion-parameter GLM-130B to the more large-scale GLM-4, has always adhered to the concept of openness and sharing. As of June 30, 2025, Zhipu AI has open-sourced more than 50 models, with cumulative downloads of open-source large models exceeding 45 million times. This choice of technical route has enabled Zhipu AI to quickly build an extensive developer ecosystem, forming a technology radiation network centered on open-source models.
MiniMax: Closed-Loop Multimodal Strategy & AI-Native Products
In contrast, MiniMax describes itself as a “global AI large model company”, whose vision focuses on “developing advanced large models” and “building AI-native products” to improve social productivity and enrich personal lives.
MiniMax has chosen a more focused approach of parallel development of “closed-source commercial models + open-source basic models”, concentrating resources on building exclusive models with ultimate user experience. Covering text, voice and video, MiniMax is committed to constructing a complete technology stack for multimodal AI, and its launched abab series models have achieved outstanding performance in multiple Chinese-language evaluations. This closed-source strategy allows MiniMax to more strictly control the quality of technology output, establish high technical barriers, and provide customized solutions for high-end customers.
Business Models Compared: B-End Focus vs. Hybrid Ecosystem
Under such route choices, the business models and revenue sources of the two companies also show significantly different focuses.
This means that Zhipu AI’s customers are mainly B-end clients such as large enterprises and government agencies, and its business model is more inclined to the traditional sales of enterprise-level software and services, featuring high unit prices and deep customization. In its revenue structure, B-end solutions account for over 80%. Through the dual-track strategy of “open-source model for user acquisition + commercial version for monetization”, Zhipu provides customized large model solutions for customers in fields such as finance, education, scientific research, and industrial manufacturing—including intelligent empowerment solutions for universal assembly systems. As of June 30, 2025, its models have served more than 8,000 institutional customers.
This means that Zhipu AI’s customers are mainly B-end clients such as large enterprises and government agencies, and its business model is more inclined to the traditional sales of enterprise-level software and services, featuring high unit prices and deep customization. In its revenue structure, B-end solutions account for over 80%. Through the dual-track strategy of “open-source model for user acquisition + commercial version for monetization”, Zhipu provides customized large model solutions for customers in fields such as finance, education and scientific research. As of June 30, 2025, its models have served more than 8,000 institutional customers.
Although this model brings a high gross profit margin of 50%, it requires heavy human input and is difficult to scale up. To break through the bottleneck, Zhipu is accelerating its transformation to MaaS (Model-as-a-Service), attracting developers through API interfaces.
MiniMax’s revenue comes from two major segments: AI-native products and open platform and other enterprise services. Its AI-native product matrix includes Talkie/Xingye, Hailuo AI, MiniMax and MiniMax Voice, mainly targeting global individual users, with monetization methods including subscriptions, in-app purchases and online marketing services. Its open platform provides API calling services for enterprises and developers.
According to the data for the first nine months of 2025, MiniMax’s AI-native products contributed 71.1% of the revenue, while the open platform contributed 28.9%. MiniMax has successfully built a hybrid ecosystem covering over 200 million individual users and more than 100,000 enterprises/developers.
Star-Studded Shareholder Lineups: A Side-by-Side Look
Prospectus shows that since its establishment, Zhipu AI has completed 8 rounds of financing, with a total financing scale exceeding RMB 8.3 billion, gaining recognition from various well-known investment institutions including state-owned capital, industrial capital, and VC/PE firms. Participants include industrial capital such as Meituan, Ant Group, Alibaba, Tencent, Xiaomi, BOSS Zhipin and TAL Education Group, first-tier institutions such as Legend Capital, Sequoia Capital, Hillhouse Capital, Qiming Venture Partners and Shunwei Capital, as well as state-owned capital support from local governments.
Before the IPO, the control of Zhipu AI is collectively held by the controlling shareholder group composed of the Tsinghua-originated founding team led by Professor Tang Jie, which jointly controls approximately 33.03% of the company’s voting rights prior to listing. Meituan holds 4.27% of the shares, Ant Group holds 3.99%, Legend Capital holds approximately 6.73% in total, and Capital Today (Xu Xin) holds 2.82%.
MiniMax’s capital camp is equally star-studded. MiniMax has previously gone through 7 rounds of financing, with investors including Cloud Alpha Capital, IDG Capital, Hillhouse Ventures, Ming Capital and miHoYo. Tencent Investment and Alibaba participated in its Series A and Series B financing. The nearly USD 300 million Series C financing completed by MiniMax in July 2025 pushed its valuation to over USD 4 billion. Despite being established for only 3 years, it has accumulated financing of over USD 1.5 billion, with cash on hand reaching USD 1.1 billion, making it the AI company with the shortest time from establishment to IPO.
Before the IPO, Alibaba, miHoYo, IDG Capital and Tencent are the controlling shareholders of MiniMax. Alibaba holds 13.66% of the shares, miHoYo holds 6.4%, and Tencent holds 2.58%.
The simultaneous listing applications of Zhipu AI and MiniMax on the Hong Kong Stock Exchange demonstrate two successful paradigms and future prospects of China’s large model entrepreneurship wave, and will also provide valuable references for the commercialization path of the large model industry. In the future, whether the “depth” of deepening industrial integration will win, or the “breadth” of conquering users will reign supreme, the market will wait and see.
Automated assembly mechanical connection equipment
Artificial intelligence automated assembly mechanical connection robot
