| 交易日期 | 证券简称 | 证券代码 | 成交价(元) 成交金额(万元) 成交量( * ) 买入营业部 | | | 卖出营业部 | | --- | --- | --- | --- | --- | --- | --- | | 2026-01-26 | 合合信息 | 688612 | 293.99 6.686 10 | 机构专用 | 类大证券股份有限 公司上海徐汇区渠 | | 1月26日，合合信息大宗交易成交10万股，成交额2939.9万元，占当日总成交额的2.23%，成交价293.99元，较市场收盘价296.06元折价0.7%。 ...

随着人工智能产业迈入"落地为王"的新阶段，技术与多元场景的深度融合成为行业核心竞争力。近期， 上海合合信息科技股份有限公司集中发布系列基于多模态大模型的创新产品，覆盖AI教育、AI健康管 理、AI基础设施及AI智能体应用等领域，以文本智能技术与垂直场景的深度绑定，为AI商业化落地提 供了可落地的实践路径。 （扫描全能王"CS-AI一站式智能化文档解决方案"功能一览） 在通用文档处理领域，合合信息旗下扫描全能王推出"CS-AI一站式智能化文档解决方案"，实现从影像 数字化向文档全周期智能服务的跨越。该方案覆盖扫描、阅读、编辑、学习等核心场景，可自动修复图 像质量问题，完成文档智能重排与排版优化。依托在文档解析、版面还原上的技术积淀，其在跨境电 商、出境游、专业文档翻译等出海场景中展现出强劲潜力。 （"蜜蜂试卷"举一反三功能演示） 基于多模态大模型文本智能技术，合合信息将AI能力延伸至民生垂直场景，打造千人千面的智能体 验。教育领域，面向国内外市场推出"蜜蜂试卷""QuizAI"两款AI错题学习管理工具，可智能识别手写体 试卷，提供批改及"举一反三"互动学习功能，助力个性化因材施教。健康领域，AI饮食健康助手 A ...

Core Viewpoint - The company is experiencing stable cash flow and financial growth, primarily driven by its "Scan All-in-One" product, which constitutes over 70% of its C-end revenue. The C-end business is projected to generate revenue of 1.075 billion yuan in 2024, with a compound annual growth rate (CAGR) of 27% from 2020 to 2024. The revenue from "Scan All-in-One" is expected to reach 982 million yuan in 2024, with a CAGR of 31% during the same period. The company maintains stable financial metrics, with gross margins around 80% and net margins around 20% in recent years [1][2]. Group 1 - The C-end revenue from the "Scan All-in-One" product is stable at over 70% of total revenue [1] - The projected revenue for the C-end business in 2024 is 1.075 billion yuan, with a CAGR of 27% from 2020 to 2024 [1] - The revenue from "Scan All-in-One" is expected to be 982 million yuan in 2024, with a CAGR of 31% from 2020 to 2024 [1] Group 2 - The company's vertical AI overseas business is anticipated to become a second growth curve characterized by "high win rates and high odds" [2] - The company has developed a vertical AI matrix, launching a series of innovative products based on multimodal large models across various fields, including AI education, AI health management, AI infrastructure, and AI agent applications [1] - The AGI technology platform and text intelligence technology platform provide a robust technical foundation for rich product functionalities and excellent performance for both C-end and B-end customers [1] Group 3 - The AI programming technology is expected to significantly reduce trial and error costs, enhancing programming efficiency by over 80%, which accelerates product iteration and upgrades [2] - The company is positioned as a leading domestic SaaS provider, rapidly expanding its vertical AI overseas business [2] - The expected earnings per share (EPS) for 2025, 2026, and 2027 are projected to be 3.19 yuan, 4.48 yuan, and 5.81 yuan, respectively, with corresponding price-to-earnings ratios of 96, 68, and 53 times based on the closing price on January 23, 2026 [2]

Investment Rating - The report initiates coverage with a "Buy" rating for the company [2]. Core Insights - The company generates stable cash flow from its flagship product, "Scan All-in-One," which contributes over 70% of its revenue. The revenue from the C-end business is projected to reach 1.075 billion yuan in 2024, with a compound annual growth rate (CAGR) of 27% from 2020 to 2024. The B-end business is also expected to grow steadily, reaching 360 million yuan in 2024, with a CAGR of 24% during the same period [7][27]. - The company is positioned as a leader in the domestic SaaS market and is accelerating its expansion into vertical AI applications, which are anticipated to become a second growth curve characterized by high probability and high returns [7][27]. Summary by Sections 1. Building a Mature C+B Product Matrix with Steady Financial Growth - The company has established a robust product matrix based on its proprietary intelligent text recognition and commercial big data technologies, serving both C-end users and B-end clients [10][12]. - The C-end revenue is primarily driven by the "Scan All-in-One" and "Business Card All-in-One" apps, while the B-end services include intelligent text recognition and commercial big data solutions [10][14]. - The company maintains a high gross margin of around 80% and a net margin of over 20%, with revenue growth outpacing expense growth, leading to sustained profitability since 2020 [18][27]. 2. Vertical AI Expansion: High Probability + High Returns as a Second Growth Curve - The company is developing a vertical AI matrix that includes innovative products across various fields such as AI education, health management, and AI infrastructure [35]. - The AGI technology platform and text intelligence technology platform provide a strong technical foundation for both C-end and B-end products, enhancing product performance and functionality [44]. - AI programming is expected to significantly reduce development costs and improve efficiency, providing a favorable environment for the company's vertical AI development [48]. 3. Profit Forecast and Investment Recommendations - The company is expected to achieve earnings per share (EPS) of 3.19 yuan, 4.48 yuan, and 5.81 yuan for the years 2025, 2026, and 2027, respectively, with corresponding price-to-earnings (P/E) ratios of 96, 68, and 53 [7][52]. - The report suggests that the company's C-end product matrix will continue to grow, with revenue growth rates projected at 25%, 28%, and 30% for 2025, 2026, and 2027, respectively [52].

Investment Rating - The report indicates a positive outlook for the computer industry, suggesting that it has fundamental support for investment and potential for future growth [1][2]. Core Insights - The computer sector is expected to experience a rebound in performance due to the recovery of the macro economy and advancements in AI technology, which are anticipated to enhance profitability [1][10]. - The report highlights a significant increase in demand for CPUs, driven by the rise of AI applications and the need for high-performance computing [16][19]. - The introduction of sandbox environments for AI agents is seen as a critical development, enhancing system security and performance in high-concurrency scenarios [22][29]. Summary by Sections Investment Allocation - In Q4 2025, the public fund allocation for the computer industry was 1.6%, a decrease of 0.8 percentage points from the previous quarter, ranking it 14th among 30 sectors [2][3]. - The top ten holdings in the computer sector included Inspur Information, Kingsoft Office, and Sangfor Technologies, with notable changes in the composition of these holdings [4][5]. Valuation Metrics - As of January 23, 2026, the PE (TTM) for the computer industry was 94.6x, placing it in the 94.80th percentile historically, while the PS (TTM) was 4.0x, in the 71.40th percentile [10][11]. - The report suggests that despite high PE levels, there is still room for growth in PS and PCF valuations as the market sentiment improves [10][11]. CPU Price Expectations - Intel and AMD are expected to raise server CPU prices by 10% to 15% in Q1 2026 due to supply constraints and increased demand from AI applications [16][21]. - The report emphasizes the growing importance of CPUs in AI workloads, particularly in managing complex tasks and ensuring system stability [19][40]. Key Investment Themes - The report identifies several key investment themes, including leadership in the digital economy, AIGC applications, and the importance of data elements and new industrialization [41][42]. - Specific companies recommended for investment include Inspur Information, Kingsoft Office, and various firms involved in AI and data solutions [41][42].

Investment Rating - The report maintains a positive outlook on the computer industry, indicating a "Buy" rating for the sector [2]. Core Insights - The computer sector is supported by strong fundamentals, with expectations of increased allocations to high-performing stocks. The sector's public fund allocation is currently at 1.6%, down 0.8 percentage points from the previous quarter, marking a low point since 2010 [2][3]. - Valuation metrics such as PE, PS, and PCF suggest that there is still room for growth in the computer sector, despite high current levels. The PE (TTM) is at 94.6x, in the 94.80% historical percentile, while PS (TTM) is at 4.0x and PCF (TTM) is at 51.6x, indicating a divergence that could be resolved as market sentiment improves [2][11]. - Anticipated price increases for server CPUs due to supply constraints and rising demand driven by AI applications are highlighted. Major players like Intel and AMD may raise CPU prices by 10% to 15% in Q1 2026 [2][18]. - The report emphasizes the importance of sandbox environments for agents, which enhance system performance and security by isolating tasks, thus improving user experience in high-concurrency scenarios [24][32]. Summary by Sections 1. Fund Holdings Analysis - The computer sector's public fund allocation decreased to 1.6% in Q4 2025, ranking 14th among 30 sectors, reflecting a decline in investor confidence [3][4]. - The top ten holdings in the computer sector include Inspur Information, Kingsoft Office, and Sangfor Technologies, with notable changes in positions among these stocks [5][6]. 2. Valuation Metrics - The report notes that the current PE (TTM) of 94.6x exceeds previous peaks in 2020 and 2023, indicating market optimism about future earnings potential [11][12]. - The divergence between PE and PS/PCF suggests that as the macroeconomic environment stabilizes, the sector's valuations may realign, leading to potential market cap growth [11][12]. 3. CPU Price Expectations - The report anticipates a supply-demand mismatch for CPUs, with significant price increases expected due to locked-in capacities and rising AI-driven demand [18][23]. - The role of CPUs is evolving, with increased responsibilities in managing complex workflows driven by AI agents, necessitating higher performance specifications [21][42]. 4. Sandbox Technology - The report discusses the growing importance of sandbox environments for AI agents, which enhance performance and security by isolating tasks and managing resources effectively [24][32]. - The sandbox model is seen as crucial for handling high concurrency and ensuring system stability, particularly in AI applications [32][40]. 5. Key Investment Themes - The report identifies several key investment themes, including digital economy leadership, AIGC applications, and new industrialization, with specific companies highlighted for each theme [45].

1月23日，合合信息大宗交易成交10万股，成交额3042.9万元，占当日总成交额的2.09%，成交价304.29 元，较市场收盘价306.44元折价0.7%。 | 交易日期 证券简称 | | 证券代码 | 成交价(元) 成交金额(万元) 成交量(*) 买入营业部 | | | 卖出营业部 | 是否为专场 | | --- | --- | --- | --- | --- | --- | --- | --- | | 2026-01-23 | 含合信息 | 688615 | 304.29 10 | 3042.9 | 机构专用 | 签合适者就是想要 | Ka | ...

Investment Rating - The report maintains a positive outlook on the industry, specifically highlighting the potential for effective control of AI model hallucinations by 2026 [2]. Core Insights - The report emphasizes that while hallucinations in AI models are inevitable, advancements in algorithms, data quality, and engineering practices can significantly reduce their occurrence. The top 25 global models have achieved a hallucination rate below 8% [5][6]. - The report identifies three key areas for investment: mature AI applications, marketing AI that is less sensitive to hallucinations, and data plus AI infrastructure [6]. Summary by Sections 1. Hallucinations - The Lower Bound of Model Capability - The report defines hallucinations as overconfident errors produced by language models, which can include fabrications, factual inaccuracies, contextual misunderstandings, and logical fallacies. For instance, GPT-3.5 had a hallucination rate of approximately 40%, while GPT-4's rate was 28.6% [14][15]. 2. Sources of Hallucinations - Hallucinations arise from several factors, including model architecture, toxic data, lack of accuracy in reward objectives, and context window limitations. Addressing these factors is crucial for controlling hallucinations [7][8]. 3. Reducing Hallucinations: From Models, Data, Engineering, and Agents - The report discusses various strategies to mitigate hallucinations, such as using larger training datasets, extending context windows, and incorporating human feedback through reinforcement learning (RLHF) [25][26]. - Engineering practices like Retrieval-Augmented Generation (RAG) are becoming standard, with Gartner predicting a 68% adoption rate by 2025 [56][57]. 4. 2B Application Penetration and Evolution - The report notes that the control of hallucinations in mainstream models has made significant progress, with the top 25 models in the Vectara HHEM ranking achieving hallucination rates below 8%. For example, the Finix model developed by Ant Group has a hallucination rate of only 1.8% [72].

Investment Rating - The report maintains a "Buy" rating for the following companies: Weike Technology (301196.SZ), Nengke Technology (603859.SH), Hehe Information (688615.SH), and Maixinlin (688685.SH) [6][50]. Core Insights - The AI application landscape is evolving, with the launch of the new "Task Assistant" feature in the Qianwen app, which integrates over 400 services from Alibaba's ecosystem, marking a significant shift from information processing to task execution [3][27]. - DeepSeek has released an open-source Engram module that enhances memory retrieval and reasoning efficiency in large models, addressing traditional architecture challenges [2][20]. - SkildAI has completed a $1.4 billion Series C funding round, indicating strong market interest in general AI models for robotics, with a valuation exceeding $14 billion [36][38]. Summary by Sections Computing Power Dynamics - The rental prices for computing power remain stable, with specific configurations like Tencent Cloud's A100-40G priced at 28.64 CNY/hour and Alibaba Cloud's A100-40G at 31.58 CNY/hour [17][19]. - DeepSeek's Engram module introduces a "lookup-computation separation" mechanism, significantly improving model efficiency in memory retrieval and reasoning tasks [2][20]. AI Application Dynamics - QuillBot's weekly traffic increased by 13.20%, indicating growing user engagement in AI tools [25][26]. - The Qianwen app's upgrade allows users to complete complex tasks such as ordering food and booking travel through natural language commands, showcasing the practical application of AI in daily life [3][28]. AI Financing Trends - SkildAI's recent funding round attracted major investors, including SoftBank and NVIDIA, highlighting the increasing capital flow into AI robotics [36][39]. - The company's innovative "hardware-agnostic" architecture aims to address the scarcity of training data in robotics, positioning it as a leader in the emerging market for general AI models [38][39]. Investment Recommendations - The report suggests focusing on companies like Maixinlin (688685.SH), Weike Technology (301196.SZ), Hehe Information (688615.SH), and Nengke Technology (603859.SH) for their growth potential in AI applications and computing power [48][50].

Investment Rating - The report maintains a "Buy" rating for the following companies: Weike Technology (301196.SZ), Nengke Technology (603859.SH), Hehe Information (688615.SH), and Maixinlin (688685.SH) [6][50]. Core Insights - The AI application landscape is evolving, with the launch of the new "Task Assistant" feature in the Qianwen app, which integrates over 400 services from Alibaba's ecosystem, marking a significant shift from information processing to task execution [3][27]. - DeepSeek has released an open-source Engram module that enhances memory retrieval and reasoning efficiency in large models, addressing traditional architecture challenges [2][20]. - SkildAI has completed a $1.4 billion Series C funding round, indicating strong market potential for general AI models in robotics, with a valuation exceeding $14 billion [36][38]. Summary by Sections Computing Power Dynamics - The rental prices for computing power remain stable, with specific configurations priced at 28.64 CNY/hour for Tencent Cloud and 31.58 CNY/hour for Alibaba Cloud [17][19]. - DeepSeek's Engram module introduces a "lookup-computation separation" mechanism, significantly improving model efficiency in knowledge retrieval and reasoning tasks [2][20]. AI Application Dynamics - QuillBot's weekly traffic increased by 13.20%, indicating growing user engagement in AI applications [25][26]. - The Qianwen app's upgrade allows users to complete complex tasks like ordering food and booking travel through natural language commands, showcasing the practical application of AI in daily life [3][28]. AI Financing Trends - SkildAI's recent funding round attracted major investors, including SoftBank and Bezos Expeditions, highlighting the increasing interest in AI robotics and its potential across various industries [36][39]. Investment Recommendations - The report suggests focusing on companies like Maixinlin (688685.SH), Weike Technology (301196.SZ), Hehe Information (688615.SH), and Nengke Technology (603859.SH) for their growth potential in AI applications and computing power [48].