In the Nanshan Zhigu Industrial Park in Shenzhen, a bipedal robot is facing repeated kicks from staff, maintaining its balance and escaping while standing firm. This is a daily scene at Zhujidongli Laboratory and has become the unique “traffic password” for the company at major exhibitions.
In early 2025, when leaders from Guangdong Province visited several humanoid robot companies in Shenzhen, Zhujidongli showcased its self-developed full-size humanoid robot performing coherent actions such as squats, waist twists, and 360-degree hip joint rotations from a prone to standing position. This scene amazed the attendees, and it is just a glimpse of Zhujidongli’s technical strength.
Founded in January 2022, Zhujidongli has quickly become a dark horse in the field of embodied intelligence in China within just three years. It has not only attracted investments from giants like Alibaba and China Merchants Venture Capital but has also completed a total of 500 million yuan in Series A financing within six months.
Under the leadership of founder Zhang Wei, Zhujidongli aims for a grand goal—to become the “NVIDIA of the embodied intelligence field.”
01
The story of Zhujidongli is inseparable from its founder Zhang Wei’s academic background and industry insights. Zhang Wei’s academic resume can be described as a “textbook in the field of machine learning theory and algorithms.”
In 1999, Zhang Wei enrolled in the Department of Automation at the University of Science and Technology of China. After graduating with his bachelor’s degree, he went abroad, embarking on a 16-year journey of studying and teaching overseas—earning a master’s degree in Electrical and Computer Engineering from the University of Kentucky, followed by a Ph.D. from Purdue University.
During his postdoctoral period, he joined the Hybrid Systems Laboratory at the University of California, Berkeley, under the mentorship of National Academy of Engineering members Claire Tomlin and Shanker Sastry.
Since 2011, Zhang Wei has been teaching in the Department of Electrical and Computer Engineering at Ohio State University, where he was promoted to tenured associate professor in June 2017.
While conducting robotics research abroad, Zhang Wei often purchased components from China and discovered that many hardware products were shipped from Shenzhen’s Yuehai Street.
“If I found a part unsuitable or needed a replacement, I would have to wait another month or two for a new order, which slowed down the iteration speed of the entire robot hardware.” This inefficiency in hardware iteration made him acutely aware of the differences between China and the U.S. in terms of robot industrialization.
In May 2019, Zhang Wei returned to China, joined Southern University of Science and Technology, and founded the Robotics Control and Learning Laboratory, focusing on research in robot control theory and learning algorithms.
Along with him from the U.S. was his doctoral student Chen Hua, who later became a co-founder of Zhujidongli.
At SUSTech, Zhang Wei’s team independently designed and developed a quadruped robot platform—Nankai Xiaotian—and conducted multiple studies on legged robots, with their papers being accepted at IROS multiple times.
02
In January 2022, Zhang Wei founded Zhujidongli. At that time, the domestic robotics sector already had several competitors, including UBTECH, Yushutech, Fourier, and Leju, among others, who had been in the market for years.
However, Zhang Wei chose a differentiated positioning—not to be a direct competitor in robot applications but to be an enabler.
Zhujidongli’s financing capabilities are impressive:
In October 2023, the company completed angel and Pre-A round financing, totaling nearly 200 million yuan.
In March 2025, Zhujidongli announced the completion of A+ round financing, cumulatively achieving 500 million yuan in Series A financing within six months.
Strategic industrial investors include Alibaba Group, China Merchants Venture Capital, Shangqi Capital, NIO Capital, and Lenovo Capital.
Notably, this is Alibaba’s first investment in a humanoid robot company, and they continued to increase their stake in the A+ round. According to Tianyancha information, Zhujidongli currently has 18 shareholders, with Alibaba’s investment platform Hangzhou Haoyue Enterprise Management Co., Ltd. becoming the second-largest shareholder after the founding team.
In July 2025, Zhujidongli also announced a new round of financing led by JD.com, with both parties exploring collaboration in retail, logistics, and services. This indicates that Zhujidongli has gained the favor of two major e-commerce giants in China within just one year.
03
On the product technology front, Zhang Wei chose the differentiated track of legged robots.
He once analyzed that the key core of legged robots lies in the “legs,” which need to develop all-terrain mobility capabilities, thus covering the shortcomings of wheeled robots while avoiding the already saturated market of wheeled robots.
The unique “small brain AI” technology route is key to Zhujidongli’s emergence.
In Zhang Wei’s view, the control of humanoid robots and legged robots requires small brain AI. The controllers for robotic arms and wheeled dual arms are not new; basic kinematics and dynamics are sufficient to solve them, and they do not require AI to be mature.
However, “the AI small brain is key to creating humanoid robots.”
Zhang Wei further elaborated on the essence of the AI small brain: “It is about generating a lot of data using models in a simulator and training controllers through this data, which is its most fundamental change.”
This method significantly shortens the development cycle of robots and improves the performance and adaptability of motion control.
In early 2024, Zhujidongli’s bipedal robot PI appeared in the Tanglang Mountain of Shenzhen’s suburban park, completing walking tests dynamically across various complex terrains such as wild stone paths, gravel slopes, grassy slopes, and ditches.
During this time, despite external disturbances such as pushing, kicking, and hitting with sticks from engineers, the robot still demonstrated strong control and stability.
Subsequently, Zhujidongli iterated this bipedal robot into the world’s first multi-form bipedal robot TRON 1, equipped with three foot designs: dual-point foot, bipedal, and dual-wheeled foot, which not only possesses the standing and walking capabilities of humanoid robots but can also achieve the speed and efficiency of wheeled robots, thus accommodating various terrains.
04
In terms of business strategy, Zhujidongli has a clear positioning—Zhang Wei hopes to become the “NVIDIA of the embodied intelligence field,” a company providing “robot bodies and AI software tools” for embodied intelligence.
Zhang Wei greatly admires NVIDIA’s business model: “Jensen Huang has reduced computing costs by a million times in the past decade, and NVIDIA has achieved this, which has enabled various AI applications to land. He is essentially serving innovators, and we want to do something similar.”
Based on this philosophy, Zhujidongli does not directly enter terminal scenarios but instead provides bodies, small brains, and model development toolchains to help downstream customers create their brains.
Zhang Wei candidly stated, “We do not have the know-how to enter specific scenarios.”
Zhujidongli abstracts its target customers into the “IDS system”: I (Innovator), including research institutes and technology companies; D (Developer), developers who create new functions based on existing technologies; S (System Integrator), system integrators who integrate technologies and functions into practical solutions.
Through these three types of users, Zhujidongli serves the innovation process of robots, including technological innovation, development innovation, and solution innovation.
In July 2025, Zhujidongli launched the full-size humanoid robot LimX Oli, with a starting price of 158,000 yuan, featuring a fully open modular architecture.
This product is positioned as a general-purpose humanoid robot platform, aimed at AI research, motion control algorithm development, and system integrators, providing modular hardware and open software ecosystem support.
This product strategy further clarifies Zhujidongli’s positioning as a platform company.
05
Zhujidongli’s ability to stand out among numerous robotics companies is closely related to founder Zhang Wei’s unique way of thinking.
Zhang Wei excels at “learning mathematics from an engineering perspective,” “explaining theory from a user perspective,” and is particularly adept at reconstructing problems in chaos. He believes that the essence of technology is to serve real problems, not to prove superiority of identity.
In terms of the landing of robots, Zhang Wei has proposed many counter-consensus insights.
He believes that “the fixed investment in hardware and software for robot landing only accounts for 10%-20% of the entire value chain, while subsequent deployment and maintenance are the largest costs.” Replacing humans with robots is a task that can easily create illusions.
He further pointed out, “When robots land, people often focus on CAPEX (capital expenditure), which is the fixed investment, forgetting that its OPEX (operational expenditure) is the largest and most challenging.”
In Zhang Wei’s view, “the last 10% of robot landing is enough to kill the preceding 90%.”
Regarding the current norm of robot industry landing, Zhang Wei’s description is sharper: “To replace a worker with a humanoid robot often requires ‘a few very talented algorithm PhDs,’ and the more projects you take on, the more the company loses.”
This clear awareness makes Zhujidongli’s path to commercialization more stable.
06
Looking to the future, Zhang Wei has laid out a clear roadmap for Zhujidongli: first, make the bipedal humanoid robot user-friendly, with rich functions and smooth movements; then make it easy to program, allowing actions to be called using natural language; and finally, grow an ecosystem, forming an Agentic OS for embodied intelligence, akin to the Windows of humanoid robots.
The ultimate goal is “to make it so that there are no difficult robots to land in the world.”
Zhang Wei believes that the current humanoid robot operating system is in a transitional period from “DOS to Windows,” laying the foundation for the developer ecosystem.
He hopes that Zhujidongli can become a builder of this ecosystem.
On the product level, Zhujidongli plans to focus on three core embodied intelligence technologies: design and manufacturing of body hardware, full-body motion control based on reinforcement learning, and training strategies for embodied brain models, gradually providing the market with humanoid robot body hardware and software systems and embodied agent development toolchains.
On the technical level, Zhujidongli is tackling general mobile operation capabilities and gradually integrating multi-modal large model technology into humanoid robots.
The company uses a large amount of human motion data to pre-train humanoid robots, helping them better complete tasks and adapt to human society.
This process is divided into three steps: data screening, pre-training, and reinforcement learning, utilizing Real2Sim2Real closed-loop, allowing real-world data to continuously learn and train in a simulated environment before entering the real world to validate learning outcomes.
07
Zhujidongli’s rapid development is inseparable from the overall rise of China’s robotics industry.
According to a report released by GaoGong Industry Research Institute in May 2023, the global market size for humanoid robots is expected to exceed 200 billion yuan by 2030.
Zhang Wei’s return to China to start a business coincided with a critical moment for the Chinese robotics industry, driven by policies such as the Ministry of Industry and Information Technology’s “14th Five-Year Plan for Robotics Industry Development” and technological breakthroughs in humanoid robots propelled by multi-sensor fusion and reinforcement learning.
At the same time, China’s advantages in the robotics hardware supply chain also provide strong support for startups like Zhujidongli.
Zhang Li, co-founder and COO of Zhujidongli, stated: “Shenzhen has a clear advantage in the robotics hardware supply chain; some hardware can be ordered in the morning and delivered in the afternoon, greatly enhancing the iteration speed of hardware products for robotics companies.”
This speed advantage in hardware iteration is something Zhang Wei found unimaginable while conducting research abroad and has become a unique competitive edge for Chinese robotics companies.
In August 2025, at the World Robot Conference in Beijing, Zhujidongli’s LimX Oli completed its first public demonstration.
Facing the crowd, the robot walked steadily, turned, and squatted, as if announcing the arrival of the era of embodied intelligence in China to the world.
Zhang Wei and his team are already contemplating their next steps—how to build the “Windows ecosystem” for humanoid robots, allowing global developers to create infinite possibilities for embodied intelligence based on Zhujidongli’s platform.
In Zhang Wei’s view, humanoid robots are “AI with a hardware carrier,” where AI determines the robot’s generalization ability, and manufacturing determines the product’s reliability.
Zhujidongli stands at the rare intersection of the AI world and the physical world.