Recently, BYD announced an investment in a humanoid robot company. The company is Zhiyuan Robotics, and this represents BYD's first disclosed investment in the humanoid robot sector.
New Products and Funding
Zhiyuan Robotics recently unveiled its first-generation general-purpose bipedal humanoid robot, Yuanzheng A1. The robot is 175 cm tall and weighs 53 kg; it can participate in visual inspection processes and assemble chassis components on busy automobile assembly lines. Shanghai Zhiyuan Xinchuang Technology Co., Ltd. was established in February of this year, with business scope that includes the research and development of intelligent robots, artificial intelligence theory and algorithm software development, AI hardware sales, and electronic product sales.
In addition to BYD, investors in the recent funding round for Zhiyuan Robotics include Lanchi Ventures, Wofu Ventures, and Qiyu Investment. Founded six months ago, Zhiyuan Robotics has completed multiple financing rounds, reflecting strong market interest in humanoid robots.
In the humanoid robot field, products such as Xiaomi's CyberOne, Dreame's general-purpose humanoid robot, and Cloud Ginger from CloudMinds have already emerged. Goldman Sachs projects the humanoid robot market could reach $154 billion by 2035, potentially becoming another major AI application scenario following autonomous electric vehicles.
What Makes Development Difficult?
At Tesla's 2023 shareholders meeting on May 16, Tesla released a video of the Optimus humanoid robot, showing progress in natural walking, motor torque control, force control, and environment recognition and memory. Elon Musk emphasized that now is an excellent opportunity to advance humanoid robot development because many software and hardware achievements from autonomous driving can be shared; Optimus is planned to use the same full self-driving system as the company's electric vehicles. He suggested that a natural person might even need more than one humanoid robot and proposed that global humanoid robot numbers could reach 10 to 20 billion.
Andrew Yao, an academician of the Chinese Academy of Sciences and director of the Institute for Interdisciplinary Information at Tsinghua University, stated that an embodied form of general artificial intelligence (AGI) would ideally take the shape of a humanoid robot.
Humanoid robot design and manufacturing aim to interact with human tools and environments, thereby assisting or replacing humans in production and daily life. Traditional robots often focus on specific applications—for example, industrial robots emphasize motion control, while vacuum or delivery robots emphasize navigation.
Compared with other service robots, humanoid robots require higher capabilities in perception, motion control, and interaction. Their motion capability represents a system integration challenge involving autonomous driving, visual navigation, sensors, servo systems, gearboxes, and control systems among other software and hardware technologies. Integration complexity is higher, and development is an extension and evolution of existing applications.
Because of this high complexity and the need for substantial R&D investment, humanoid robot development has been one of the more difficult areas in robotics for decades. Early humanoid robots such as Honda's ASIMO and SoftBank's Pepper have been retired.
Industry reports decompose humanoid robots into core components such as control systems, the physical body, and sensing systems. The integration of large models with speech, vision, decision-making, and control can help robots form perception-decision-control loops and acquire common-sense knowledge. However, hardware components needed by humanoid robots—servomotors that act like human joints, sensors, and ball screws for transmission—still require continued development across the supply chain.
Another report notes the humanoid robot industry chain is still in an early stage, concentrated upstream on core components such as reducers, sensors, controllers, and chips, while downstream application areas are broad, including logistics, security, elderly care, and disability assistance. For large-scale commercialization and household adoption, resolving high cost remains the key challenge.
Upstream in the industry chain are raw materials and components; the midstream covers robot bodies and system integration; downstream are application scenarios. For the three core industrial robot components—servo motors, reducers, and controllers—research indicates that these parts account for over 70% of an industrial robot's cost, with reducers, servo systems, and controllers approximately 36%, 24%, and 12% respectively. Given the significantly higher degrees of freedom in humanoid robots compared with traditional industrial robots, the share of cost for reducers and motors is expected to be even higher.
Why Is the Sector Hot Now?
Several companies have recently launched new humanoid robots, and many university labs and research institutions are commercializing their projects. One reason for this trend is the development of large AI models, which provide technical support for deploying humanoid robots. Industry leaders such as Nvidia’s CEO Jensen Huang believe embodied intelligence will be the next wave in AI.
Hardware for robot bodies is increasingly constrained by the physical limits of electric drive forms, so competition has shifted toward whether robots are smarter—that is, toward the AI capabilities provided by large models. Baidu's founder Robin Li has said large models compress human understanding of the world and suggest a path toward general AI, marking a new era centered on large models.
A June report from the Ministry of Science and Technology showed that the distribution of released large models is heavily concentrated in China and the United States, together accounting for more than 80% of the global total, with the United States leading in the number of large models.
How can general-purpose robots be manufactured at low cost and deliver real application value? Peng Zhihui, co-founder of Zhiyuan Robotics, said hardware is only one prerequisite; the more important factor is the robot's "brain." Zhiyuan's embodied intelligence architecture, EI-Brain, splits the robot's thinking system into cloud-level superbrain, edge-level brain and cerebellum, and brainstem layers corresponding to task-level, skill-level, instruction-level, and servo-level capabilities.
Andrew Yao noted that large AI models will create new industrial value and help build general-purpose robots capable of wide application. Such robots need to combine the three key parts of an embodied intelligent agent—brain, cerebellum, and body—and adapt to human environments. In this view, the "brain" handles higher-level reasoning, decision-making, planning, and natural language communication; the "cerebellum" coordinates vision, touch, and other perception to control the body for complex tasks; and the "body" requires adequate hardware such as sensors and actuators.
Many companies have begun deploying robot-specific large models. For example, DeepMind released the RT-2 model to improve robots' natural language understanding, and domestic firm CloudMinds released RobotGPT, a multi-modal large model for robots. These developments mean that applying general AI to humanoid robots will strengthen capabilities in simulation and data generation, task decomposition and planning, and autonomous operation control, with significant application potential.
Some analysts predict that after Tesla released its humanoid robot product, commercialization will accelerate. Under Tesla's influence, global humanoid robot commercialization could advance in 2025 with sales of 20,000–30,000 units, new demand of 1 million units by 2030, and potential sales exceeding 10 million units by 2035.
Approaching a Tipping Point
Robotics appears to be approaching a transformative tipping point. A 2023 report on robotics technology and industry development notes rapid progress in common and frontier robotics technologies. Disciplines including electronics, mechanical engineering, biology, and materials science are increasingly converging, with innovation, high-end development, intelligence, and ecosystem building highlighted as future priorities for the robotics sector in China.
The three core technologies for humanoid robots are human-machine interaction, scene perception, and motion control. Scene perception has advanced rapidly. Motion control typically uses hydraulic or all-electric drives, but motion control algorithms still have room for significant improvement. Human-machine interaction has seen breakthroughs with large AI models, but substantial gaps remain before robots achieve autonomous decision-making.
Yao identified four main challenges for embodied intelligent robots: first, there is no single embodied large model that achieves everything in one step as large language models do; second, computational power constraints; third, how to fuse multimodal sensory inputs; and fourth, the need for large-scale data collection.
Existing humanoid robots are constrained by technology and cost, limiting current application scenarios, but the potential remains vast. Policy support is also being introduced. On August 22, China's Ministry of Industry and Information Technology and three other departments issued the "New Industry Standardization Pilot Project Implementation Plan (2023–2035)," focusing on standardization work for emerging and future industries and identifying priority areas that include humanoid robots and generative AI. The national robotics standardization technical committee has recently established a humanoid robot standardization working group with several research and industry institutions designated as deputy leaders to promote standardization efforts in the Chinese market. At the local level, municipal measures in Beijing propose organizing a foundation-building project for the robotics industry and publishing a list of key technological challenges, as well as setting up a robotics industry fund with a planned scale of 10 billion yuan, with the first phase no less than 2 billion yuan.
At the 2023 World Robot Conference, the vice minister of the Ministry of Industry and Information Technology stated that new technologies and products represented by humanoid robots and general AI are developing vigorously and are becoming focal points in global technological innovation and new industry tracks. The momentum in China's robotics industry provides a foundation for humanoid robot development.
Robotics applications are accelerating across new-energy vehicles, medical technology, logistics and warehousing, outdoor delivery, and commercial services. Advances in AI, biotechnology, new materials, and brain-computer interfaces are rapidly progressing, and robots are becoming platforms that integrate these frontier technologies, affecting robot form, function, and application scenarios.
BYD's investment in Zhiyuan Robotics demonstrates how humanoid robots can be applied in tasks such as visual inspection and chassis assembly; at a deeper level, industry chain deployment may also promote the development and application of technologies like autonomous driving. Regardless of BYD's future plans in humanoid robots, the sector is likely to experience rapid development.
Whether humanoid robots will eventually enter millions of homes remains an open question. Zhou Jian, founder and CEO of UBTech, believes household companion scenarios for humanoid robots are a matter of time, but he also notes that task planning for humanoid robots requires integrating technologies, common sense, and knowledge across multiple domains, which remains a major challenge.
Major automakers and technology companies alike—from Tesla, Honda, and Toyota to Huawei, Xiaomi, Samsung, Google, and Amazon—face the same core challenges: reducing costs and building out the industrial supply chain. Overcoming these challenges remains essential for the future development of humanoid intelligent robots.
