智谱带来 AI 市场估值新叙事。 文 丨 江思远 243.77 亿元，17 日通过港交所聆讯的智谱在网上公开的招股书中确认了它自己的最新估值，这也是第 一次，人们确切知道中国大模型公司的估值金额。 这是一个恰切的时间点，ChatGPT 在业界投下的震撼弹已经过去三年，经历了百模大战的喧嚣，留在 牌桌上的中国大模型选手，在证明了自己创新与技术能力不逊色于任何人之后，开始向资本市场发起 冲刺。 在 "少年走向成年" 的关键转型期，市场期待大模型公司如何向所有人论证，"新奇类、炫技型" 的模 型技术向 "实际类、适配型" 的大模型端到端落地的过程。 作为百团大战中第一家成功 "上岸" 的公司，智谱给出的回答不那么让人满意。招股书披露的亏损规模 远大于收入增长，而研发费用还在大幅增加，似乎没有要停止 "烧钱" 的趋势。 如果是一个成熟公司，这样的资产负债表实在让人难以满意，但大模型是特殊的。 智谱是中国成立最早的独立大模型公司之一，但它也只有不到 6 年历史，AI 行业让无数人兴奋的原因 是它在潜在未来的革命性，而这很难从过往的历史中直接推导出来，因为变革往往在某个时刻突然以 指数级形式发生。 这是智谱的赌注，或者可 ...

Core Viewpoint - The semiconductor chip industry is experiencing a strong rebound, highlighted by the performance of the first Hong Kong ETF focused on the chip sector, which saw a daily increase of 1.58% and a trading volume of 52.89 million yuan [1][8]. Group 1: ETF Performance - The Hong Kong Information Technology ETF (159131) opened high and closed with a gain of 1.58%, recovering its 5-day moving average [1][8]. - The ETF's underlying index has a price-to-earnings ratio of 33.23, which is at the 37.9% percentile over the past three years, indicating significant room for growth compared to its peak earlier in the year [3][10]. Group 2: Key Stocks Performance - Over 70% of the ETF's constituent stocks showed positive performance, with SenseTime-W leading with an increase of over 8%, followed by SMIC and Kinsus Interconnect Technology, both rising by 6% [4][11]. - The ETF is heavily weighted towards semiconductor and electronic companies, with SMIC holding a weight of 20.48% and Xiaomi Group-W at 9.53% [6][13]. Group 3: Technological Advancements - Researchers from Shanghai Jiao Tong University have made a breakthrough in optical computing chips, which could alleviate issues in the AI industry related to high costs, slow training, and energy consumption [2][9]. - The development of optical computing chips is expected to provide disruptive pathways for real-time AI, edge AI, and green AI applications, potentially transforming business and operational models in the AI sector [2][9]. Group 4: Market Trends and Future Outlook - The IPO of Zhizhu AI, regarded as the "Chinese version of OpenAI," signals a new phase in China's large model industry, focusing on technological strength and sustainable business models [3][10]. - The investment attractiveness of the Hong Kong chip sector is highlighted by its favorable valuation compared to other major tech indices, suggesting a strong potential for future growth [3][10].

小虎要长成猛虎了。 12月19日，中国大模型六小虎之一智谱，通过港交所聆讯，正式向IPO发起冲击。 两天后，六小虎中的另一虎MiniMax（稀宇科技）也同样来到这一步，争夺全球大模型第一股。 不论这两家中国公司谁能先成功上市，都意味着在资本化程度上将领先OpenAI等美国一众大模型巨头。 国产大模型有1%的概率成为OpenAI 恰好沐曦和摩尔线程市值就在3000亿元人民币左右。 都说国产GPU有1%概率成为英伟达，因而估值就是1%的英伟达，对应市值是440亿美元（3097亿元人民币）。 搬到大模型领域，这个逻辑依然行得通。 12月18日，外媒消息，OpenAI正在洽谈千亿美元融资，估值可能达到8300亿美元，其1%是83亿美元（约584亿元人民币）。 今年7月MiniMax完成近3亿美元融资，估值达到约300亿元人民币；智谱则累计已完成数十轮融资，仅今年就斩获超30亿元，估值约在300-400亿元人民币 区间。 甚至还不到1%，为什么差距能这么大？ 2022年，ChatGPT横空出世，2个月内用户数破亿，成为互联网史上增长最快的应用。 从聊天助手ChatGPT到视频生成Sora，再到推理模型o3、智能体工具A ...

Group 1: Financing Trends - The technology and manufacturing sectors, along with healthcare, continue to dominate the financing market with a high density of cases and significant funding amounts [1] - In the past week, over 20 financing cases were reported in the technology and manufacturing sectors, including multiple transactions exceeding 500 million RMB [1] - The healthcare sector is also showing steady growth, with large transactions and innovative sub-sectors receiving attention [1] Group 2: Regional Financing Distribution - The majority of financing activities were concentrated in Jiangsu Province, Beijing, and Guangdong Province, with 7, 6, and 5 cases respectively [3] Group 3: Active Investment Institutions - Yizhuang Guotou and Guoke Investment were notably active, each disclosing 2 investments primarily in the technology and manufacturing, as well as healthcare sectors [5] Group 4: Notable Financing Cases in Healthcare - Beijing Bohui Biotechnology completed nearly 100 million RMB in A+ round financing, led by Yizhuang Guotou, to accelerate clinical progress and expand its pipeline [7] - Suzhou Jishi Medical Technology completed A round financing, focusing on eye disease treatment and prevention solutions [8] - Suzhou Baifu Laser Technology completed over 100 million RMB in C round financing to enhance product development and market expansion [9] - Kangyuan Bochuang completed 250 million RMB in B round financing to advance research in tumor immunity and next-generation kinase inhibitors [10] Group 5: Notable Financing Cases in Technology and Manufacturing - Chengdu Xinglian Xintong completed over 100 million RMB in B round financing to strengthen its position in the satellite communication sector [17] - Suzhou Xinshi Nuo Semiconductor Equipment completed over 500 million RMB in A+ round financing to enhance its capabilities in semiconductor automation [18] - Shanghai Fangqing Technology completed over 500 million RMB in Pre-A round financing to advance its distributed AI computing architecture [20] Group 6: Additional Notable Financing Cases - Shenzhen Haodian Technology completed over 100 million RMB in B round financing, focusing on lithium battery adhesive development [21] - Suzhou Junji Gene Technology completed nearly 100 million RMB in A round financing to enhance its automated gene technology applications [22] - Jiangxi Panmeng Semiconductor completed over 100 million RMB in A+ round financing for capacity expansion and technology development [39]

12月19日，港交所官网显示，北京智谱华章科技股份有限公司（以下简称"智谱"）已通过聆讯并公布了 IPO（首次公开募股）招股书等资料，不仅成为业界所称的"大模型六小龙"中率先步入IPO流程的企 业，而且也抢在OpenAI、Anthropic等海外大模型明星企业前面，有望成为"全球大模型第一股"。随着 488页的招股书披露，公众终于得以一窥智谱大模型生意的"成色"。 中国城市专家智库委员会常务副秘书长、浙大城市学院副教授林先平认为，从财务数据来看，智谱当前 收入规模难以支撑长期高研发投入，需尽快提升收入或控制成本。依赖企业本地化部署，市场空间或受 限，且面临定制化成本高、毛利率压力等问题。智谱的商业化路径可持续性面临挑战，不过仍有调整空 间，高研发投入若转化为产品差异化，可能提升竞争力，这取决于未来一两年内能否实现收入规模化， 并优化研发效率。 需要指出的是，就在12月21日一早，"大模型六小龙"中另一家MiniMax也在港交所披露了聆讯资料集。 《中国经营报》记者注意到，不同于此前多数企业在递表时会先公示一个"申请版本"的招股书，智谱与 近日通过聆讯的壁仞科技、天数智芯等，均是直接披露聆讯后资料集，这是港交所 ...

Core Insights - The competition for the title of "the first global large model stock" is intensifying, with Minimax releasing its IPO prospectus shortly after Zhipu [1] Group 1: Minimax's Business Developments - Minimax has made significant progress in the AI video generation sector, despite challenges in monetizing user subscriptions for large language models [2] - The company has developed a core suite of self-researched large models, including MiniMax M2, Hailuo-02, and Speech-02, leading to applications like MiniMax and Hailuo AI [2] - Hailuo, launched in August 2024, has already become a key revenue source, generating $0.17 billion (1.2 billion RMB) in the first three quarters of 2025, accounting for 32.6% of total revenue [2] Group 2: Market Performance and User Engagement - Hailuo's paid user base reached 310,000, with an average revenue contribution of $56 per user [2] - However, Hailuo's revenue still lags behind Kuaishou's AI video generation app "Ke Ling," which achieved over $0.25 billion in revenue in the second quarter of this year [2] - Hailuo's pricing strategy includes "Basic" and "Premium" packages priced at $9.99/month and $199.99/month, respectively, targeting overseas markets where user willingness to pay is higher [2] Group 3: User Retention Challenges - The AI video generation sector faces significant uncertainty regarding user retention, with early data showing low retention rates for similar applications like Sora [3] - Hailuo's user retention rates in Singapore are also concerning, with 1-day, 7-day, 30-day, and 60-day retention rates at 22.57%, 4.62%, 0.8%, and 0.66%, respectively [4] Group 4: Financial Performance and Strategic Adjustments - Minimax reported net losses of $0.465 billion in 2024 and $0.512 billion in the first three quarters of 2025 [6] - To mitigate losses, Minimax has reduced its promotional spending, with sales expenses in the first three quarters of 2025 at $0.039 billion, a decrease of over 25% year-on-year [6] - Despite these efforts, the company still struggles to cover its computing costs, which totaled $0.18 billion in sales and R&D expenses for the first three quarters of 2025 [6]

Core Insights - The first large model IPO is set to take place in China, with the AI company Zhipu being the first among the "six small tigers" in the large model sector to initiate the IPO process [5][6] - Zhipu's listing will provide a quantifiable valuation reference for the long-lacking public market in the large model industry, marking a shift from a "technology race" to "capital validation" for China's AI sector [5][6] - Since its establishment in 2019, Zhipu has raised over 8 rounds of funding, accumulating over 8.3 billion RMB, with a pre-IPO valuation of 24.38 billion RMB [5][6] Company Overview - Zhipu is a leading AI company in China focused on developing advanced general large models, aiming for innovation in artificial general intelligence (AGI) [8] - The company provides general large model services to institutional clients and individual users, supporting over 8,000 institutional clients and approximately 80 million devices as of June 30, 2025 [7][8] - According to Frost & Sullivan, Zhipu ranks first among independent general large model developers in China, with a market share of 6.6% [7] Financial Performance - Zhipu's revenue has shown significant growth, with figures of 57.4 million RMB in 2022, 125 million RMB in 2023, and 312 million RMB in 2024, reflecting a compound annual growth rate of over 130% [7][8] - Despite revenue growth, the company reported substantial losses of 144 million RMB, 788 million RMB, 2.958 billion RMB, and 1.236 billion RMB for the years 2022, 2023, 2024, and the first half of 2025, respectively [7][8] Research and Development - Zhipu has made significant investments in R&D, with expenditures rising from 84.4 million RMB in 2022 to 529 million RMB in 2023, and further to 2.195 billion RMB in 2024 [8] - The company’s R&D spending for the first half of 2025 is projected to be 1.595 billion RMB, indicating an 85.6% increase from the previous period [8] Product Offerings - Zhipu's MaaS platform offers four types of models: language models, multimodal models, agent models, and code models, along with integrated tools for model fine-tuning and deployment [9][10] - The platform is designed to provide comprehensive model capabilities, extensive application scenarios, and adaptable computing infrastructure [9][10] Intellectual Property - As of the last feasible date, Zhipu holds 86 registered patents in China, with 84 being invention patents, and has filed 234 patent applications [10]

智谱成立于2019年，由清华大学技术成果转化而来，是国内启动大模型研究的开拓者，也是国产大模型 技术领军企业。其原创提出了基于自回归填空的通用预训练范式GLM，并研发出中国首个预训练大模 型框架。是国内罕有在原创技术路线上与全球顶尖水平保持同步的厂商。值得注意的是，智谱已于日前 通过港交所上市聆讯，作为"大模型六小龙"中首家启动IPO的企业，智谱赴港冲刺"全球大模型第一 股"。 据悉，智谱2022年-2024年收入分别为5740万、1.245亿、3.124亿元，连续三年翻倍。研发人员占比 74%，2022年、2023年、2024年公司研发费用分别为8440万、5.289亿、21.954亿，2025年上半年，研发 投入为15.947亿，累计研发44亿左右。研发投入支撑其技术快速迭代，GLM系列模型每3-6个月完成一 次基座迭代。其最新模型在Code Arena竞技场上代码能力与全球顶级模型并列第一。截止今年9月30 日，智谱模型赋能全球12000家企业客户、逾8000万台终端用户设备及超4500万名开发者，是中国赋能 终端设备最多的独立通用大模型厂商。 格隆汇12月22日｜格隆汇今日线上举办"科技赋能·资本破局 ...

日前，同为"AI六小虎"的智谱AI与MiniMax(稀宇科技)先后通过港交所上市聆讯，消息一经传出便引发 关注。截至目前，两家公司均未对此公开置评。 如何让营收增长跑赢烧钱速度？ 原生大模型公司业绩首度披露 12月19日晚间和12月21日晚间，港交所网站分别更新了智谱和MiniMax的招股书，资本市场得以第一次 窥见原生大模型公司的完整业绩。 成立于2019年的智谱由清华大学技术成果转化而来，是中国最早研发大模型的企业。从招股书披露的数 据来看，自2022年以来，智谱的收入连续三年增长，2022年为5740万元，2023年增长至1.245亿元， 2024年进一步扩大至3.124亿元，三年收入复合年增长率达到130%。2025年上半年收入为1.9亿元。同 时，毛利率持续保持在50%以上。以2024年的收入计，智谱在中国独立通用大模型开发商中位列第一， 在所有通用大模型开发商中位列第二，市场份额为6.6%。 作为一家通用大模型公司，智谱拥有657名研发人员，占公司总员工数的74%，与此同时智谱的研发投 入呈现指数级增长。2022年为8440万元，2023年攀升至5.289亿元，2024年进一步扩大至21.954亿 ...

Core Viewpoint - MiniMax Group Inc. is preparing for an IPO in Hong Kong, aiming to be one of the first companies in the "big model" AI sector to go public, alongside another company, Zhiyu [2] Company Overview - MiniMax was established in June 2021 and began operations in January 2022, focusing on research and development [2] - The company has a young workforce, with an average employee age of 29 and an average board age of 32 [2] - MiniMax has raised a total of $390.4 million in its latest funding round, achieving a post-money valuation of approximately $4.24 billion (about 29.84 billion yuan) [2] Funding History - MiniMax has completed seven funding rounds since its inception, including a $31 million angel round in December 2021 and a $50 million Pre-A round in April 2022 [2] - The company has consistently raised funds in 2023 and 2024, with two rounds of financing each year [2] Share Structure - MiniMax employs a dual-class share structure, where Class B shares carry ten votes per share, while Class A shares carry one vote [3] - Key shareholders include Dr. Yan Junjie, who holds approximately 28.25% of shares and 74.91% of voting rights, and Ms. Ren Yeyi, who holds 2.50% of shares and 6.65% of voting rights [3] Product Focus - MiniMax specializes in developing advanced large models and AI-native products, targeting both consumer and enterprise markets [4] - The company has adopted a mixture of expert (MoE) architecture and mixed attention mechanisms to enhance performance while reducing computational resource consumption [4] Revenue Generation - MiniMax's revenue primarily comes from AI-native products, an open platform, and other AI-based enterprise services [7] - The company has reported significant user engagement, with over 200 million individual users and more than 100,000 enterprise clients globally [7] Financial Performance - Revenue from AI-native products has shown substantial growth, with figures of approximately $0, $758,000, $2.18 million, and $3.80 million for the years 2022 to 2024 [8] - The company reported a total revenue of approximately $3.46 million in 2022, $30.52 million in 2023, and projected $53.44 million in 2024 [10] Future Plans - MiniMax plans to use the proceeds from its IPO to fund research and development over the next five years, focusing on large model development and scaling AI-native products globally [12]