The robotics sector suddenly experienced a wave of limit-up moves.

This afternoon, A-share robotics stocks staged a dramatic surge, with Hengshuai, Hanwei Technology, and Mailixin recording strong 20% daily limit gains, while Sanhua Intelligent Controls, Shuanghuan Driveline, and over 10 other stocks hit 10% daily limits. Analysts point out that the latest developments from the humanoid robot industry chain have ignited market optimism.

According to an official announcement from State Grid Hangzhou Power Supply Company today, on September 15, State Grid Hangzhou Power Supply Company, Unitree Robotics, Embodied Intelligence Base Company, and State Grid Zhejiang Huadian Research Institute signed a "Power + Embodied Intelligence" framework cooperation agreement at the company's integrated innovation center, launching deep collaboration in power artificial intelligence applications. Unitree Robotics also announced the open-sourcing of UnifoLM-WMA-0 - an open-source world model-action architecture across multiple robot platforms.

According to latest reports, Tesla CEO Elon Musk stated that he will hold meetings with various departments within Tesla next week, with particular focus on artificial intelligence/autonomous driving systems, Optimus robot production plans, and vehicle production and delivery. Securities institutions note that Musk has repeatedly emphasized the complexity and importance of dexterous hands in public, expecting dexterous hands to be a key focus of the Optimus V3 version upgrade, making the related supply chain worthy of continued attention.

**Limit-Up Rally**

On the afternoon of September 16, A-share robotics concept stocks collectively surged. By market close, Hengshuai, Hanwei Technology, and Mailixin strongly captured 20% daily limits, while Sanhua Intelligent Controls, Shuanghuan Driveline, Dayang Motor, Shiyun Circuit, Kersen Technology, Zhongzhong Technology, and over 10 other stocks recorded 10% daily limits. Additionally, Kangping Technology, Weike Technology, Kaite, Jiangsu Leili, Jingduan Technology, Siling, and multiple other stocks surged over 10%.

Hong Kong robotics concept stocks also continued their rally. As of 3:00 PM, Dobot surged over 5%, UBTech rose over 4%, while Chunli Medical and Capital M&E followed suit.

On the news front, on September 15, Unitree Robotics announced the open-sourcing of UnifoLM-WMA-0 - an open-source world model-action architecture across multiple robot platforms. Unitree Robotics explained that UnifoLM-WMA-0 is specifically designed for general robot learning, with its core being a world model that can understand the physical laws of "robot-environment interaction." This world model features two core functions:

First, a simulation engine that operates as an interactive simulator, providing synthetic data for robot learning;

Second, policy enhancement that can interface with an action head, further optimizing decision-making performance by predicting future interactions with the physical world.

In simple terms, the simulation engine acts like a virtual training ground that can generate large amounts of synthetic data for robot learning and training; policy enhancement can predict the robot's "next step" in real environments, providing reference for decision-making. This allows humanoid robots to become "smarter" without needing to trial-and-error in real environments every time.

Unitree Robotics completed model training on five open-source datasets, with test results showing that the model, as a simulation engine, can achieve interactive controllable generation based on "current images" and a certain number of "robot future actions."

Unitree Robotics stated that this open-sourcing will drive robot evolution from instruction execution to autonomous learning, with related code and documentation now available on the project homepage. The solution is compatible with industrial robotic arms, service robots, and other multi-scenario applications, providing a reusable general technical framework for the industry.

Unitree Robotics noted that to advance global embodied intelligence industry development, they are now completely open-sourcing this world model-action architecture with continuous updates, working together to make general robots possible as soon as possible.

Meanwhile, Tesla's Optimus robot has also made significant progress. Tesla CEO Musk stated plans to conduct a technical assessment of AI5 chip design on Saturday, with meetings scheduled for next week regarding artificial intelligence/autonomous driving systems, Optimus robots, and vehicle production.

Musk previously discussed the latest status of Optimus robots at the All-In Summit. Musk believes the Optimus robot is "the greatest product in human history," stating that the currently designed Optimus V3 will solve hand dexterity, possess an AI brain, and is expected to achieve mass production.

Musk indicated the product will enter mass production in 2026, with plans to produce one million units over the next five years. Musk also addressed costs and potential mass production challenges.

CITIC Securities pointed out that Musk has repeatedly emphasized the complexity and importance of dexterous hands in public forums, expecting dexterous hands to be a key focus of the Optimus V3 version upgrade, making the related supply chain worthy of continued attention. Tesla's upcoming third-generation new product launch after two years, following previous mass production setbacks, is gradually clarifying next year's outlook, benefiting sector beta performance.

**OpenAI Also Makes Major Moves**

Notably, amid the increasingly intense global AI competition, OpenAI is quietly accelerating its robotics sector deployment.

According to a September 15 Wired report, OpenAI is accelerating its investment in robotics technology, potentially by forming a new team focused on humanoid robots, betting that interaction with the physical world is a key step toward achieving artificial general intelligence (AGI).

OpenAI has recently hired multiple researchers with expertise in humanoid robot control algorithms. Insiders reveal that the company's recruitment targets clearly point toward humanoid robot projects. Meanwhile, public job postings show the company is seeking experts with experience in teleoperation and simulation (such as the Nvidia Isaac platform), which are key technologies for training humanoid robots.

Various signs indicate that the global tech industry is launching a new round of "humanoid robot competition," with this sector on the verge of explosion. Shanghai Securities stated that humanoid robots entering industrial scenarios has become a high-certainty application trend both domestically and internationally, with commercial deployment of humanoid robots expected. They recommend focusing on benefiting domestic component manufacturers and humanoid robot industry chain-related event catalysts.

Essence Securities believes that dexterous hands are upgrading toward highly biomimetic and refined directions, representing a strong certainty focus for Optimus V3 iteration. Approaching October, the finalization of Optimus Gen3 robots is getting closer, with subsequent order placement news providing continuous catalysts for the sector. After Optimus Gen3 model finalization, Tesla will conduct large-scale manufacturing tests, with expected shipment volumes of 30,000-50,000 units; and may simultaneously develop Optimus Gen4 models for future large-scale mass production, while making clear requirements for supplier capacity building and overseas factory construction. The robotics sector is expected to launch a major beta toward 2026 under domestic and international resonance in Q4. The Tesla chain focuses on "strong certainty," while the domestic chain grasps two directions: "supply chain positioning" and "application scenario volume expansion."

Huachuang Securities emphasized attention to three core logics for humanoid robots: 1) New technology directions for cost reduction and weight reduction, focusing on axial flux motors, MIM powder metallurgy, and rotary transformers; 2) Application scenario "robot+" targets, where scenario precedence not only determines the mass production path for domestic robots but also largely determines their market space ceiling. B-end first volume expansion trends are already quite clear, especially in high-repeat purchase, high-necessity, low-error-tolerance application areas such as intelligent logistics, textile and garment, and specialized industries; 3) Two major sectors for robots from 1 to 10: equipment, data, and vision.