--- title: "First Take Analysis of Zhipu AI's Prospectus: Annual Revenue of 300 Million with a Growth Rate of 130%, \"Chinese Version of OpenAI\" Aiming to Be the First Global Large Model Stock" type: "News" locale: "zh-HK" url: "https://longbridge.com/zh-HK/news/270375005.md" description: "Zhipu AI is regarded as the \"Chinese version of OpenAI\" and is racing towards an IPO on the Hong Kong Stock Exchange. The company was established 6 years ago, has completed 8 rounds of financing, raised over 8.3 billion yuan, and has a valuation of 24.38 billion yuan. Zhipu AI focuses on AGI research and development, has developed the \"GLM\" series of models, and has built a complete system from foundational models to application products. The company plans to release the AGI L1-L5 roadmap in 2024, covering pre-training, alignment and reasoning, autonomous learning, self-awareness, and conscious intelligence stages" datetime: "2025-12-20T07:45:43.000Z" locales: - [zh-CN](https://longbridge.com/zh-CN/news/270375005.md) - [en](https://longbridge.com/en/news/270375005.md) - [zh-HK](https://longbridge.com/zh-HK/news/270375005.md) --- > 支持的語言: [简体中文](https://longbridge.com/zh-CN/news/270375005.md) | [English](https://longbridge.com/en/news/270375005.md) # First Take Analysis of Zhipu AI's Prospectus: Annual Revenue of 300 Million with a Growth Rate of 130%, "Chinese Version of OpenAI" Aiming to Be the First Global Large Model Stock **"The world's first stock of large models"** has initiated its sprint on the Hong Kong Stock Exchange. Regarded as the "Chinese version of OpenAI," **ZhiPu AI** has just announced its prospectus and is sprinting towards an IPO. This technology company, incubated by a top team from Tsinghua University, obtained the entry ticket for IPO counseling registration in April this year. Since then, the buzz around ZhiPu's IPO has never ceased, but unexpectedly, the progress surged at the end of this year— ZhiPu AI has just **passed the hearing** on the Hong Kong Stock Exchange in a secret filing and is about to ring the bell. In its 6 years of establishment, ZhiPu AI has completed 8 rounds of financing, raising over 8.3 billion yuan in total, with a star-studded lineup of capital behind it, and its latest valuation has reached 24.38 billion yuan. ## **What kind of company is ZhiPu AI?** **ZhiPu (Z.AI)** is actually quite familiar to everyone. It is one of the earliest companies in China to start LLM research and is regarded as the "Chinese version of OpenAI," focusing on the development of AGI with the aim of "making machines think like humans." The core technology of the company revolves around its self-developed **"GLM" series models**, which have now built a complete system from foundational models to application products. In **mid-2024**, ZhiPu AI will release its AGI **L1-L5 roadmap** concurrently with OpenAI: - Pre-training stage: ZhiPu aims to teach machines to understand, write, and communicate using human language. - Alignment and reasoning stage: Align machine intelligence with human intentions and teach machines to reason and plan. This includes enhancing safety, reducing hallucinations, and achieving alignment of language intelligence with images, videos, speech, and actions. - Autonomous learning stage: Teach machines to learn from their own thoughts and actions through self-criticism, self-reflection, and contemplation. - Self-awareness stage: Without human supervision, machines gradually form their own attitudes and emotions by observing their own behavior and self-interpretation. - Conscious intelligence stage: Machines begin to perceive their internal states and external environments, exhibiting characteristics similar to human consciousness—this concept remains complex even for humans. Currently, ZhiPu has developed large models and intelligent agents in three stages, such as China's first hundred billion scale large model GLM-130B (pre-training stage), and China's first open-source large chat model ChatGLM (alignment and reasoning stage), among others. Specifically regarding the product matrix: First, let's look at the **base model**, which is the foundation for developing various specialized models. ZhiPu's "GLM" series models adopt a **self-regressive fill-in-the-blank architecture**, possessing strong general understanding and generation capabilities, supporting multi-modal input and output such as text, images, speech, and video, with outstanding performance in robustness, controllability, and hallucination In the second half of this year, Zhipu launched the new generation flagship models GLM-4.5/4.6, achieving the native integration of reasoning, coding, and agent capabilities in a single model for the first time. According to Frost & Sullivan data: Once released, GLM-4.5 quickly rose to the top of the global Baobao Face hot list within 48 hours; in July's 12 industry standard benchmark tests, GLM-4.5 ranked third globally, first in China, and first among global open-source models. The GLM-4.6, launched in September, primarily enhanced coding capabilities and was released with compatibility for Cambrian and Moore Threads chips. In the large model arena Code Arena, GLM ranks alongside models from Anthropic and OpenAI as the global leader in coding, surpassing overseas closed-source models Google Gemini and xai's Grok. To address the pain points of "low reasoning efficiency and high costs" in large models, Zhipu AI developed a dedicated **reasoning model**, optimizing reasoning speed and reducing deployment costs through techniques such as model compression and quantization. For example, GLM-4-9B has accumulated over 150,000 GitHub stars since its open-source release, with global downloads exceeding 30 million. Next is Zhipu AI's **multimodal models**, such as CogVLM, CogView, and CogVideoX, which achieve cross-modal understanding and generation of text, images, videos, and speech. Finally, **Agent** is Zhipu's core strategic direction this year, as AI begins to shift from "dialogue" to "action." Among them, AutoGLM is the world's first intelligent agent capable of autonomously operating a mobile phone, allowing users to complete cross-application tasks such as ordering food, booking tickets, checking rooms, making appointments, writing documents, and generating videos through voice commands without manual operation. The updated AutoGLM 2.0 in August also supports integration with AI glasses, watches, home appliances, and other hardware, enabling hardware to possess "mobile-level operational capabilities," such as ordering coffee with Rokid smart glasses. In terms of commercialization, Zhipu started early, beginning to lay out the **MaaS** (i.e., large model API calling) business model in 2021. Similar to the monetization methods of OpenAI and Anthropic, Zhipu is currently the only startup that has achieved substantial revenue from MaaS. The advantage of this model is its high flexibility and scalability, delivering "general intelligence" rather than "custom projects," allowing for faster and lower-cost deployment across more scenarios Currently, the Zhipu MaaS open platform has **2.7 million +** enterprises and application developers, making it one of the most active large model API platforms in China. The customer base primarily targets the technology internet and enterprise service market, with **9 out of the top 10** internet companies in China using Zhipu's GLM large model. In the global large model marketplace OpenRouter, the call volume for GLM-4.5/4.6 ranks in the top 10 worldwide, and its paid traffic revenue has surpassed the total of all domestic models combined. Moreover, GLM-4.5/4.6 is not only generating high revenue domestically but is also profitable overseas. Its model business aimed at global developers (GLM coding plan) has already generated over 100 million RMB in revenue, attracting more than 150,000 paid developer users in just three months. As of September 30 this year, Zhipu's models have reached 12,000 enterprise customers globally, over 80 million terminal user devices, and more than 45 million developers. In November this year, Zhipu's average daily cloud token call volume exceeded 4.2 trillion times, making it the independent general large model vendor with the most empowered terminal devices in China. It is this business model and scale that has led to Zhipu's rapid growth in financial metrics. ## **Key Financial Metrics of Zhipu AI** Looking at the revenue side, in 2022, 2023, and 2024, Zhipu's operating revenue was 57.4 million RMB, 124.5 million RMB, and 312.4 million RMB, respectively, with a compound annual growth rate of 130%. In the first half of 2025, the company's revenue reached 190 million RMB, surpassing the total for the entire year of 2023. Among these, the AI large model is Zhipu's core product, thus the company's main revenue comes from **large model income**. According to a Frost & Sullivan report, based on 2024 revenue, Zhipu is the largest independent large model vendor in China. Looking at the profit side, from 2022 to 2024, Zhipu's gross profit margins were 54.6%, 64.6%, and 56.3%. In the first half of this year, the company's gross profit margin was 50%. Over the past three years, Zhipu has consistently maintained a high gross profit margin of over 50%. In terms of operating expenses, Zhipu's focus has primarily been on **R&D**. From 2022 to 2024 and in the first half of 2025, the company's R&D expenses were 84.4 million RMB, 528.9 million RMB, 2.1954 billion RMB, and 1.5947 billion RMB, respectively As of June 30 this year, the company's R&D team consists of 657 people, accounting for 74% of the total number of employees. Its core research team and academic advisory team have published 500 top high-impact papers, with a total citation count exceeding 58,000. Finally, in terms of cash, as of the end of June this year, Zhipu's cash and cash equivalents amounted to 2.52 billion yuan. ## **Led by Tsinghua Team, 6 Years and 8 Rounds of Financing Over 8.3 Billion** Zhipu AI was established in 2019, derived from the technology transfer of **Tsinghua University**, with its founding team originating from the **Knowledge Engineering Laboratory (KEG) of the Department of Computer Science at Tsinghua University**. This laboratory was established in 1996 and is one of the earliest laboratories in China to conduct research on natural language processing and knowledge graphs. CEO **Zhang Peng** is a 2018 doctoral graduate in Innovation Leadership Engineering from Tsinghua University, graduating from the Department of Computer Science and Technology at Tsinghua University. His career began at the Knowledge Engineering Laboratory (KEG) at Tsinghua University, where he joined the lab after graduating with a master's degree in 2002 and led the development of the technology intelligence platform AMiner starting in 2006—this system is the technical prototype of Zhipu AI. At Zhipu, Zhang Peng is responsible for the overall operation and strategic planning of the company, leading the R&D and commercialization of core products such as the GLM series large models and the ChatGLM dialogue model. Zhipu's chief scientist **Tang Jie** is also a core figure in the KEG laboratory. His research directions include social network analysis, data mining, machine learning, and knowledge graphs. Chairman **Liu Debing**, a PhD from the Institute of Computing Technology, Chinese Academy of Sciences, studied under Gao Wen, an academician of the Chinese Academy of Engineering and an authoritative expert in the field of computer science in China. He has served as the deputy director of the Technology Big Data Research Center at the Tsinghua Institute of Data Science, a senior researcher at Thomson Beijing Research Institute, and a senior engineer in the Department of Computer Science at Tsinghua University. Zhipu AI President **Wang Shaolan**, a doctoral graduate in Innovation Leadership Engineering from Tsinghua University, previously served as the deputy director of the Technology Big Data Research Center at the Tsinghua Institute of Data Science, facilitating Zhipu AI's transition from "laboratory" to "market." The core team members of Zhipu mainly come from the Knowledge Engineering Laboratory of the Computer Science Department at Tsinghua University, and have long been engaged in research in areas such as large-scale pre-training models, semantic big data analysis, and intelligent Q&A. With such a top-notch technical team backing it, Zhipu has had a smooth journey in the capital market. In its founding year, the company secured 40 million yuan in angel round financing, with investors from Tsinghua University's asset management and Zhongke Chuangxing. In the following six years, Zhipu has raised over **8 rounds** of financing, accumulating over **8.3 billion yuan**. The investors behind it are a gathering of star capital, including industry capitals such as **Meituan, Ant Group, Alibaba, Tencent, Xiaomi, Kingsoft, Boss Zhipin, and TAL Education**, as well as first-tier institutions like **Junlian, Sequoia, Hillhouse, Qiming Venture Partners, and Shunwei**, along with support from various **local government state-owned assets**. As of the IPO, Zhipu's latest **post-investment valuation** has reached 24.38 billion yuan. The reason this company is so sought after in the capital market is not only due to its strong foundation and technical capabilities but also because the entire industry is becoming unprecedentedly hot. According to Frost & Sullivan data, in terms of revenue, the market size of China's large language model (LLM) is expected to reach 5.3 billion yuan in 2024; of which, institutional clients will contribute 4.7 billion yuan, and individual clients will contribute 600 million yuan. With the continuous advancement of LLM technology and the growing demand from institutions and individuals, it is expected that by 2030, the market size of China's LLM will increase to 101.1 billion yuan, with a compound annual growth rate of 63.5% from 2024 to 2030, driven primarily by institutional clients. Against this backdrop, the commercialization path of the enterprise-level LLM market is becoming increasingly clear and moving towards a diversified pattern. The industry prospects are promising, and now is the time to recruit talent. At this juncture, OpenAI has already made its IPO plans clear, while the "Chinese version of OpenAI" is now taking the lead in entering the public market. _Prospectus link:_ _https://www1.hkexnews.hk/app/sehk/2025/107977/documents/sehk25121901699\_c.pdf_ Risk warning and disclaimer The market has risks, and investment requires caution. This article does not constitute personal investment advice and does not take into account the specific investment objectives, financial situation, or needs of individual users. Users should consider whether any opinions, views, or conclusions in this article are suitable for their specific circumstances. Investment based on this is at one's own risk ### 相關股票 - [OpenAI (OpenAI.NA)](https://longbridge.com/zh-HK/quote/OpenAI.NA.md) ## 相關資訊與研究 - [OpenAI Launches Bug Bounty Program To Cripple AI Weaponization](https://longbridge.com/zh-HK/news/280830008.md) - [The OpenAI/TBPN deal is a shocker that makes sense](https://longbridge.com/zh-HK/news/281574542.md) - [OpenAI's ad push explained in one chart](https://longbridge.com/zh-HK/news/281235106.md) - [OpenAI is now bringing in $2 billion a month - and 3 more highlights from its latest update](https://longbridge.com/zh-HK/news/281258023.md) - [OpenAI says Codex now offers pay-as-you-go pricing for teams](https://longbridge.com/zh-HK/news/281564073.md)