Investment Rating - The report maintains a "Recommended" investment rating for the humanoid robot industry [2][32]. Core Insights - The humanoid robot sector is experiencing significant product launches and events, such as the World Artificial Intelligence Conference (WAIC), which is expected to catalyze growth in the industry [32]. - The report emphasizes the importance of tracking the production progress of Tesla's Optimus and developments from domestic companies like Huawei and Yushu Chain to capitalize on the industry's expansion in 2025 [32]. Summary by Sections Market Review - From July 21 to July 25, 2025, the humanoid robot index increased by 2.13%, outperforming the Shanghai Composite Index, which rose by 1.67% [5][12]. - Notable individual stock performances included North Rare Earth (+16.05%) and Changsheng Bearing (+15.33%) [14]. Industry Dynamics - Yushu Technology launched its bipedal humanoid robot R1, priced starting at 39,900 yuan, featuring 26 joints and advanced motion capabilities [5][18]. - The WAIC showcased over 150 humanoid robots, marking the highest participation in its history, with various companies demonstrating applications in retail, industrial, logistics, and urban services [21][22]. - Tesla's Optimus robot began service in a restaurant, generating $47,000 in revenue within six hours, indicating strong market interest [24][25]. Company Developments - ByteDance introduced the GR-3 VLA model and the ByteMini robot, designed for household tasks with advanced capabilities [25]. - UBTECH launched the Walker S2, an industrial humanoid robot, featuring innovative technologies for autonomous operations [27][30]. Investment Recommendations - The report suggests focusing on companies with high certainty and incremental technology, particularly those linked to Tesla, Yushu, and Huawei [32]. - Key stocks to watch include those in the Tesla chain (e.g., Sanhua Intelligent Control, Top Group) and Yushu chain (e.g., Changsheng Bearing, Mannesmann) [32][33].

Investment Rating - The report assigns an "Overweight" rating for the robotics industry [4]. Core Insights - The robotics industry is expected to continue its growth trajectory, driven by new product launches and events like the WAIC conference [2]. - Key investment opportunities are identified in both complete robotics manufacturers and core component suppliers, including motors, sensors, and actuators [25]. Summary by Sections Industry News and Company Dynamics - China Railway Design has developed an intelligent quadruped robot capable of autonomous inspection in data centers, showcasing advancements in IoT and AI technologies [6]. - Yuejiang's collaborative welding robot has achieved breakthroughs in high-precision applications, enhancing efficiency in various manufacturing sectors [8]. - ByteDance's GR-3 humanoid robot demonstrates significant advancements in multi-modal interaction technology, improving task execution capabilities [9]. Investment Recommendations - Focus on robotics manufacturers and core component suppliers, with specific recommendations for: 1. Motors: Mingzhi Electric 2. Rotational joints: Zhongchen Technology, Shuanghuan Transmission, Landai Technology 3. Linear joints: Hengli Hydraulic, Zhejiang Rongtai, Demais [25]. - The report highlights the importance of domestic suppliers benefiting from the growing demand in the robotics sector [25]. Financing Dynamics - UBTECH Technology secured a record order for humanoid robots, accelerating its commercialization efforts [22]. - Tom Cat collaborates with partners to develop AI companion robots, leveraging their respective strengths in IP and technology [22]. - Yushun Technology is progressing towards an IPO, aiming to become the first humanoid robot company listed on the Sci-Tech Innovation Board [22].

Core Insights - ByteDance's Seed team launched the GR-3 general robot model, which shows superior operational performance in new environments and object handling compared to the GR-2 model set to release in October 2024 [1][2] - The GR-3 model demonstrates significant improvements in generalization and complex task execution success rates over the industry-leading embodied model π0 [1][4] Model Architecture and Training - The GR-3 model utilizes a MoT+DiT network structure, integrating the "vision-language module" and "action generation module" into a 4 billion parameter end-to-end model, enhancing dynamic instruction following through RMSNorm [2] - The training methodology for GR-3 includes a three-in-one data training approach, utilizing high-quality remote operation data, low-cost human VR trajectory data, and publicly available image-text data to improve generalization capabilities [2] Hardware Development - To maximize the potential of the GR-3 model, ByteDance introduced the ByteMini, a dual-arm mobile robot designed specifically for GR-3, featuring 22 degrees of freedom and a unique wrist ball joint design for enhanced flexibility [3] - The ByteMini includes a multi-camera coordination system for comprehensive situational awareness and a whole-body motion control system to ensure smooth trajectory generation and adaptive force adjustment during tasks [3] Performance Comparison - In comparative testing, GR-3 outperformed π0 in four categories: basic environment, new environment, complex instructions, and new objects, achieving a 17.8% higher success rate in new object handling [4] - GR-3 can elevate the success rate of new object operations from 60% to over 80% with just 10 human trajectory data points, showcasing its high generalization and complex task execution capabilities [4]

Core Insights - ByteDance's Seed team launched a new Vision-Language-Action Model (VLA) named GR-3, which boasts strong generalization capabilities, understanding of abstract concepts, and the ability to manipulate flexible objects [2][3] Model Features - GR-3's key advantage lies in its exceptional generalization ability and understanding of abstract concepts, allowing for efficient fine-tuning with minimal human data [3] - The model utilizes a Mixture-of-Transformers (MoT) architecture, integrating visual-language and action generation modules into an end-to-end model with 4 billion parameters [3] - GR-3 can perform a series of actions based on verbal commands, such as "clean the table," executing tasks like packing leftovers and disposing of trash [3] Training Methodology - GR-3 employs a three-in-one data training method, combining teleoperated robot data, human VR trajectory data, and publicly available image-text data to enhance model performance [4] - The inclusion of teleoperated robot data ensures stability and accuracy in basic tasks, while human VR trajectory data allows for rapid learning of new tasks at nearly double the efficiency of traditional methods [4] Application and Performance - In practical applications, GR-3 demonstrates outstanding performance in general pick-and-place tasks, maintaining high command adherence and success rates even in unfamiliar environments [6] - For long-range table cleaning tasks, GR-3 achieves an average completion rate exceeding 95% based solely on the command "clean the table" [6] - The model exhibits remarkable flexibility and robustness in delicate operations, successfully completing tasks like hanging clothes regardless of the garment type [6] Future Developments - The Seed team plans to expand the model's scale and training data to further enhance GR-3's generalization capabilities for unknown objects [7] - Future enhancements will include the introduction of reinforcement learning (RL) methods to allow the robot to learn from trial and error during actual operations [7] - The release of GR-3 is seen as a significant step towards developing a general-purpose robotic "brain," with aspirations for robots to assist in daily human tasks [7]