Group 1 - Investment Rating: Buy [3] - Latest Closing Price: HKD 20.14 [3] - Market Capitalization: HKD 269.33 billion [3] Group 2 - The company is a global leader in Autonomous Mobile Robots (AMR) for warehousing [2][9] - Expected revenue growth for 2025 is HKD 31.6 billion, with a year-on-year increase of 31.1% [9] - The company holds a 9.0% share of the global AMR warehousing market [9] Group 3 - Revenue projections for 2025-2027 are HKD 31.6 billion, HKD 41.7 billion, and HKD 55.1 billion respectively [9] - Net profit is expected to turn positive in 2026, with projections of HKD 374.08 million and HKD 674.87 million for 2026 and 2027 respectively [9] - The company’s gross margin is projected to improve from 37.51% in 2025 to 38.56% in 2027 [9] Group 4 - The global AMR market is expected to grow at a compound annual growth rate (CAGR) of 33.1% from 2024 to 2029 [9][65] - The AMR solutions market is projected to reach RMB 162.1 billion by 2029 [65] - The company’s competitive advantages include a self-developed technology platform, a comprehensive product matrix, and a strong customer ecosystem [9][76] Group 5 - The company launched the world's first general-purpose robot for warehousing, Gino 1, on February 10 [96] - Gino 1 is designed to perform multiple tasks such as picking, packing, and inspection, enhancing operational efficiency [102] - The introduction of Gino 1 marks a significant advancement in the company's strategy towards comprehensive automation solutions [102]

Core Viewpoint - The collaboration between Qianxun Intelligent and JD.com marks a significant advancement in the application of embodied intelligence in retail, transforming data collection and model training into a practical, integrated process that addresses the industry's data scarcity issue [1][5][25]. Group 1: Strategic Collaboration - Qianxun Intelligent and JD.com signed a strategic cooperation agreement from 2026 to 2029, focusing on customized consumer products, technology collaboration, and marketing co-construction to accelerate the application of embodied intelligence in retail [3]. - This partnership establishes a "technology + scene + data" model, where JD.com evolves from a retail channel to a data partner, and Qianxun Intelligent integrates deeply with JD.com to streamline the entire process from data collection to commercial implementation [5][6]. Group 2: Importance of Retail Scenarios - Retail environments are crucial for the evolution of embodied intelligence models due to their ability to generate high-value real-world interaction data, which is more valuable than traditional internet-based data [6]. - The collaboration allows for the collection of expert-level data through real-world operations, enhancing the model's training and adaptability [10][16]. Group 3: Data Collection and Model Training - Qianxun Intelligent has accumulated over 200,000 hours of real interaction data, with plans to exceed 1 million hours by 2026, significantly aided by the partnership with JD.com [13][15]. - The company emphasizes the importance of "dirty data," arguing that real-world imperfections provide better training scenarios for models, enhancing their generalization capabilities [15]. Group 4: Data Loop and Model Iteration - The data collection process is structured into three phases: initial remote operation for standardization, supervised robot task completion for complex scenarios, and full autonomous operation for a complete data loop [17][18]. - This iterative process ensures that data collection aligns with actual application needs, avoiding inefficiencies and enhancing the commercial viability of the technology [18][24]. Group 5: Cross-Industry Application - The data loop model developed by Qianxun Intelligent demonstrates strong adaptability across different sectors, including industrial and retail, proving its practical value and enabling broader applications [19][24]. - Future plans include extending robotic applications to JD.com’s pharmacy for automated sorting and precise medication dispensing, as well as other retail scenarios [24]. Group 6: Future Outlook - The year 2026 is anticipated to be pivotal for the commercialization of embodied intelligence, with the collaboration between Qianxun Intelligent and JD.com expected to drive significant advancements in the industry [25][26].

行业领先优势 在具身智能迈向 世界模型 的关键进程中，大规模、真实、自然、多样的人类行为数据是保证塑造基座模型真 正理解物理世界、复刻人类自然行为的关键。 近日， 简智机器人 正式推出行业首个面向具身世界模型的全模态数据集——Gen Ego Data。该数据集以 Ego-centric（以人为中心）为核心设计理念，聚焦下一阶段具身智能与世界模型范式。从第一视角出发、全 视野、高质量、多模态、多样记录人真实、自然的交互，其核心目标不是让模型模仿关节角度，而是真正学 会"感知—动作—结果"的因果关系与物理规律 。 01. Gen Ego Data 助力具身世界模型加速进步 1. 行业最广视野、高质量的第一视角全身覆盖： 6 RGB摄像头，水平FOV 270° 垂直150°，无遗漏、无遮 挡、无畸变，人手的细微动作（抓取、伸展、弯曲），还是人体的完整姿态（站立、行走、交互），都能被精 准捕捉。 2. 高质量的全局环境感知 ： 6视角可以有效解决遮挡不可见、不确定问题，为模型提供更多上下文，且交叉 验证，可以快速剔除异常标注结果，提升数据质量。 3. 行业最全模态覆盖 ： 第一视角、头部姿态、2D/3D手部姿态、上半 ...

Core Insights - The core issue in the embodied intelligence sector is the mismatch between talent supply and demand, with a significant shortage of qualified personnel [1][3][4] Group 1: Investment and Market Trends - In the past year, the total financing amount in China's embodied intelligence sector reached 73.543 billion RMB, with 744 investment events [1] - Companies like Songyan Power and Galaxy General Robotics secured financing of 1 billion and 2.5 billion RMB, respectively, after the Spring Festival [1] Group 2: Talent Demand and Recruitment Challenges - The demand for technical talent in the robotics industry has a supply-demand ratio of 5.2:1, indicating one qualified candidate corresponds to 5.2 vacant positions [3] - The recruitment market for the robotics industry has seen a growth rate of 31.3%, leading the entire industry, with the demand for humanoid robot positions surging over four times year-on-year [3] - There is a critical shortage of interdisciplinary talent skilled in AI algorithms, mechanical design, perception systems, and motion control, making it difficult for companies to find suitable candidates even with high salaries [4] Group 3: Recruitment Event Details - A specialized recruitment event for the embodied intelligence industry is scheduled for April 28-29 in Beijing, aimed at connecting companies with high-quality talent [8] - The event will facilitate direct interactions between companies and technical decision-makers, allowing for efficient talent acquisition and job matching [5][6][7] Group 4: Benefits of Participation - Companies participating in the event can effectively address their core talent needs and enhance their brand visibility among industry experts and high-end talent [9] - The event will also help build a talent pool for future recruitment needs, ensuring a sustainable supply of qualified personnel [9] Group 5: Broader Industry Engagement - The event will gather global experts, industry leaders, investors, and key players from the supply chain, creating a high-energy platform for collaboration and resource sharing [12] - The conference will include various formats such as forums, exhibitions, award ceremonies, and talent recruitment, aimed at fostering a robust industrial ecosystem [12]

Core Viewpoint - The founder and CEO of Unitree Robotics emphasizes that strong motion capabilities are essential for robots to perform tasks effectively, marking a critical point in the industry where robots can operate in approximately 80% of unfamiliar scenarios with minimal prior training [1][2]. Group 1: Current Challenges - The first core issue is the insufficient expressive capability of existing models, which struggle to generate and execute complex real-time motion data, leading to low utilization of scarce real-world data in robotics compared to language models [1]. - The second challenge is the lack of scale effects in reinforcement learning, where training data cannot be reused, necessitating retraining for each new task, thus failing to accumulate an exponential "snowball effect" [1]. Group 2: Future Technology Directions - The company is optimistic about the development of world models or video generation models, which can leverage vast amounts of internet video and text data, presenting a higher ceiling for advancement [2]. - Unitree Robotics has open-sourced a video generation-based world model in 2025 but acknowledges the challenge of aligning video generation with actual robotic execution [2]. Group 3: Industry Collaboration - The founder advocates for industry ecology and collaboration, highlighting that the rapid development of humanoid robots is a result of collective efforts from academia, industry, and research globally [2]. - Unitree Robotics is committed to open-source initiatives, regularly releasing algorithms, datasets, and code, and fostering a community for sharing and collaboration [2]. Group 4: Product Development and Recognition - The company has launched several representative robotic products, including the G1, AA2, H2, and AS2, each designed for different applications and capabilities [2]. - Significant advancements have been made in robotic software algorithms, including the ability to recover from falls in any posture and high success rates in full-body motion control [3]. - In 2026, the company upgraded its AI capabilities, showcasing complex motion and operational skills, receiving notable recognition, including a performance on CCTV's Spring Festival Gala [3].

扩散策略已成为机器人多模态动作建模的主流方法，但其依赖从随机噪声中进行多步迭代解噪，推理延迟 较高，在实时控制场景中构成关键瓶颈。虽然现有工作尝试优化噪声初始化，但仍未摆脱「从噪声到数 据」的长程生成路径。 这种「去噪」思维源自图像生成，但并不完全适用于机器人控制。与缺乏先验的图像合成不同，机器人拥 有持续反馈的自身状态信号，具有显著的物理一致性与运动连续性。A2A 的核心洞察在于：既然扩散模型 本质是学习分布间的映射，为何不将起点更换为信息密度更高、距离目标分布更近的历史状态先验？ 在机器人领域，扩散策略（Diffusion Policy）已经成为了标准模仿学习策略和 VLA 动作生成范式，但其 「从随机噪声中迭代解噪」的机制带来了不容忽视的推理延迟。如果机器人不再从随机高斯噪声开始「盲 猜」，是否可以基于「刚刚做了什么」来预测「下一步做什么」呢？ 新加坡南洋理工大学 MARS Lab 提出 Action-to-Action (A2A) Flow Matching 新范式：以历史机器人轨迹而 非随机噪声作为生成起点，打破了生成速度与精度的双重瓶颈，实现了更高训练效率、极速推理响应及卓 越的泛化表现。 Web ...

Core Viewpoint - The valuation of embodied intelligence is currently challenged by the lack of consensus on its business model and value assessment, with a new data-driven approach emerging as a potential solution through strategic partnerships like that of Qianxun Intelligent and JD Group [2] Group 1: Limitations of Traditional Valuation Models - Traditional hardware valuation models are inadequate for assessing the value of embodied intelligence, which is driven by dynamic growth through data and software rather than static hardware sales [4] - The core value of embodied intelligence lies in its ability to learn and evolve through interaction with the physical world, making its valuation non-linear and dependent on data accumulation [4] Group 2: Data Network Operations - The collaboration between Qianxun Intelligent and JD Group exemplifies a new value proposition, where the Moz robot performs complex tasks like coffee making, generating valuable training data through expert remote operation [5][6] - Each service provided by the robot is viewed as a data mining activity, with expert annotations enhancing the quality of the data collected [8] Group 3: Economics of Data Flywheel - The high cost and inefficiency of data collection in the physical world have been significant barriers to the development of embodied intelligence [9] - Qianxun Intelligent employs multiple data collection strategies to reduce costs, including a wearable data collection device that lowers costs to one-tenth of traditional methods, enabling large-scale data collection [10] - The company has accumulated over 200,000 hours of real interaction data, with plans to reach 1 million hours by 2026, creating a competitive advantage through rapid data accumulation [11] Group 4: Cross-Scenario Validation - Qianxun Intelligent's success in both industrial (e.g., Ningde Times) and commercial (e.g., JD Group) environments demonstrates the versatility of its data-driven model evolution [13][14] - The ability to apply the same underlying logic across different scenarios suggests the potential for a universal "intelligent generation and deployment platform" rather than isolated solutions [14] Group 5: Reevaluating Value Anchors - The value of embodied intelligence companies is shifting from tangible hardware to intangible, continuously appreciating data and models, with current valuations reflecting future data asset potential [15] - The future competition in the industry will focus on building efficient, closed-loop "data-intelligence" engines rather than merely hardware manufacturing and sales [15]

Core Viewpoint - The article discusses the emergence of OpenClaw, an AI software that enhances productivity by enabling direct computer operations, and its application in the automotive industry through the launch of the "Super Intelligent Agent IM Ultra Agent" by Zhiji Auto, which aims to revolutionize user interaction with vehicles [2][6][8]. Group 1: OpenClaw and Its Capabilities - OpenClaw is an open-source AI software that can perform tasks such as organizing files, replying to emails, and making reservations, marking a shift from simple conversational AI to a more functional digital assistant [2][5][6]. - The software's ability to operate computers directly represents a significant advancement in AI tools, transitioning from mere conversation to actionable tasks [6]. Group 2: Zhiji Auto's Super Intelligent Agent - Zhiji Auto introduced the "Super Intelligent Agent IM Ultra Agent," which aims to provide comprehensive assistance by understanding user preferences and executing tasks with minimal input [8][10]. - Unlike traditional smart cabins that only support basic dialogue, the Super Intelligent Agent can handle complex tasks like ordering coffee by understanding user habits and preferences [12][15]. Group 3: Advanced Features and User Interaction - The Super Intelligent Agent can interpret nuanced user statements, such as "the child is asleep," to adjust vehicle settings for a quieter and more comfortable environment [18][21]. - It can also respond to user emotions and tone, enhancing the driving experience by acting as a digital assistant that understands and anticipates needs [22][23]. Group 4: Technical Foundations - The system is built on a new integrated architecture that merges chassis control, intelligent driving AI, and intelligent cabin AI, allowing for a seamless interaction between these components [25][26]. - Zhiji Auto employs a fully controlled chassis system with triple redundancy to ensure safety and reliability, even without mechanical connections [31][34]. Group 5: AI Models and Collaborations - The IM AD ZETA model, developed in collaboration with Momenta, enhances driving capabilities through reinforcement learning, significantly improving performance compared to existing models [38][39]. - The integration of Alibaba's ecosystem allows for seamless data flow between various applications, enabling the Super Intelligent Agent to execute tasks efficiently [45][46]. Group 6: Future Implications and Safety Measures - The introduction of AI in vehicles signifies a pivotal shift towards smarter, more interactive automobiles that serve as intelligent partners in daily life [55][56]. - Zhiji Auto emphasizes the importance of safety and data security, implementing strict controls and a "black box" mechanism to ensure user information is protected [52][53].

Market Overview - Major indices showed positive performance with ChiNext Index up by 2.02%, STAR 50 up by 1.36%, Shenzhen Component Index up by 1.05%, North Exchange 50 up by 0.69%, and Shanghai Composite Index up by 0.32% [1] - TMT sector led the gains with SW Communication Application Value-Added Services up by 7.75%, SW Communication Network Equipment and Devices up by 6.89%, and SW Communication Engineering and Services up by 4.51% [1] - TMT sector saw a decline in SW Brand Consumer Electronics down by 0.47% [1] Domestic News - Semiconductor Investment Alliance: Peak Technology, a leading domestic motor drive control chip supplier, announced a price adjustment for its products starting April 1, 2026, with actual price changes varying by product [2] - Display industry update: Deep Tianma reported that its TM19 production line has achieved phase mass production capacity ahead of schedule, covering the entire product line and successfully delivering the first IGZO product [2] - Financing news: Light Particle Technology completed nearly 100 million yuan in Pre-B round financing, expanding from a single smart swimming goggles product to a matrix covering multiple smart sports glasses for various scenarios [2] - Solid-state battery development: Del's solid-state battery has achieved continuous trial production, showing good consistency and customization capabilities for different customer requirements [2] Overseas News - Semiconductor industry forecast: By 2028, the demand for high bandwidth memory (HBM) for AI-specific integrated circuits (ASICs) is expected to increase 35 times compared to 2024 [3] - SK Hynix indicated that the demand for HBM driven by AI is substantial, requiring four to five years to expand wafer production capacity, with shortages expected to last until 2030 [3] - Samsung Electronics plans to produce SiC power semiconductor samples by Q3 2026, focusing on high-reliability applications [3] - Samsung's co-CEO mentioned considering extending contracts from quarterly or annual to three to five years due to anticipated AI memory chip demand surge in 2026 [3] AI Insights - Gamma launched an AI image generation tool, Gamma Imagine, which integrates with tools like ChatGPT and Claude to support automated workflows [4] - XbotGo released the third-generation sports camera Falcon, featuring advanced AI capabilities for real-time tracking and dynamic zoom [4] - Alibaba Cloud announced price increases of up to 34% for AI computing and storage products due to global AI demand and supply chain price hikes [4] - Baidu Smart Cloud reported significant cost increases in core hardware and infrastructure, leading to price adjustments of 5%-30% for AI computing-related services [4] Industry Tracking - Low-altitude economy: China's low-altitude economy is developing well and is expected to transition from "pilot flights" to "scene verification" by 2026 [5] - Quantum technology: The UK announced a groundbreaking £2 billion investment plan to maintain its leadership in quantum innovation [5] - Embodied intelligence: HONOR introduced a Robot Phone concept at MWC 2026, marking the entry of smartphone manufacturers into the embodied intelligence hardware ecosystem [5] - New materials: A Chinese team developed a new flexible thermoelectric film with record-breaking performance, supporting future technologies like self-powered wearables [5]

Group 1 - The core viewpoint of the article is that New Times has clarified its positioning in the market, differentiating itself from Haier Smart Home by focusing on industrial scenarios while Haier targets home scenarios [1] - The actual controller and major shareholder of the company has issued a commitment letter to maintain the independence of the listed company post-acquisition, ensuring that business, personnel, finance, organization, and assets remain independent [1] - The company will disclose further details in its acquisition report on February 25, 2025, on the designated information disclosure platform [1]