Beijing Zhongguancun Embodied AI Industry "Metamorphosis": One Building Covers Upstream and Downstream Industrial Chain, Intelligent Data Becomes New Wind

Our newspaper (chinatimes.net.cn) reporter Liu Shimeng, Beijing report

Over the past two years, robots have become regulars on the Spring Festival Gala, from wearing floral padded jackets and handkerchiefs in the “YangBOT” in 2025, to performing somersaults and spinning kicks with steady landings in the “WuBOT” in 2026. In just one year, what has caused the embodied intelligent robots’ movement capabilities to evolve at “double speed”? The answer may involve many factors, but besides hardware iterations, data is another very important reason.

With the rapid development of large AI models, breakthroughs have been made in algorithms and computing power among the “Three Elements of AI,” but high-quality data remains a true scarcity. Recently, a reporter from Huaxia Times visited the Zhongguancun (Haidian) Embodied Intelligence Innovation Industrial Park. Li Yao, Vice President and Partner of Nuoyiteng Robot Technology (Beijing) Co., Ltd. (hereafter “Nuoyiteng Robots”), told the reporter that since 2023, the demand for motion capture in the humanoid robot industry has rapidly surged. “At first, we thought this might be a fleeting trend, but by 2024, it became a huge wave. We also started to seriously reconsider, reorganize resources, and build teams to focus on serving this track’s needs.”

The reporter learned that, as the first industrial park in the country named after “embodied intelligence,” the Zhongguancun (Haidian) Embodied Intelligence Innovation Industrial Park, established in February 2025, has gathered 14 embodied intelligence companies including Yikong Zhijia, Meixin Sheng, Rosenbot, Wujie Power, and Nuoyiteng Robots, as well as 7 AI companies. It has rapidly built an “upstream-downstream” embodied intelligence industrial ecosystem, enabling companies along the supply chain to achieve “zero-distance” technological collaboration and innovation.

Data for AI Becomes a Hot Topic

What is “motion capture”? At the scene of the interview, a male demonstrator was wearing a full-body inertial motion capture system, controlling a robot that “copies” his movements. He extended both hands forward, and the robot did the same; he turned slightly, and the robot followed; he performed a lunge, and the robot mimicked it.

Live demonstration of remote operation via motion capture

In fact, the demonstrator was wearing a full-body inertial motion capture system, with inertial sensors at key points on his body, including his hands, to collect acceleration, angular velocity, and other data. These data are used to real-time reconstruct human motion through human body algorithms. After being “translated” by a general motion pre-trained large model, they become full-body coordinated control commands that the robot can execute. Once receiving the commands, the robot imitates and follows human movements.

Nuoyiteng Technology was founded in Beijing in 2012 and has become one of the leading domestic companies in motion capture and human-computer interaction over the past decade. Before the rapid advancement of AI technology, motion capture was mainly used in film and game development, sports health assessment, and medical diagnosis. In 2023, as the embodied intelligence industry “began to show its sharp edges,” all this changed. Driven by exploding demand, Nuoyiteng Robots spun off independently in 2025, completed Pre-A+ funding led by Qiming Venture Partners, with investments from FiveYuan Capital, Junlian Capital, and others, including additional investments from Jingwei Venture Capital and InnoAngel Fund, raising over several hundred million RMB.

“In the past two years, the embodied intelligence track has seen rapid growth in both business demand and actual orders. I believe it will continue to grow quickly in the foreseeable future,” Li Yao explained to Huaxia Times. Essentially, embodied intelligence differs from film special effects, animation, and gaming data in terms of data transformation, data types, and data quality. For example, motion capture used for film effects doesn’t require generalized technology; it’s best to capture a professional stunt actor’s seamless movements. But embodied intelligence cannot do that. If only data from one person’s movement habits is fed into the model, or if limited scene data is used, the model will overfit and struggle to adapt to complex, changing real environments. Therefore, it’s necessary to build diverse motion datasets covering different people, environments, and tasks to improve the model’s generalization and stability.

Up and Down the Supply Chain

The rapid growth in data demand reflects the fast development cycle of the embodied intelligence industry chain. As embodied intelligence was included as a future industry in the government work report in 2025, many cities including Beijing, Shanghai, Shenzhen, Guangzhou, Suzhou, and Hangzhou have issued special plans.

Notably, at the beginning of 2026, Beijing held a conference on advancing artificial intelligence innovation hubs, aiming to significantly enhance the originality of AI foundational theories and core technologies within about two years, fostering more pioneering achievements, and positioning the development of scientific intelligence and embodied intelligence at a global leading level.

The layout of the Zhongguancun (Haidian) Embodied Intelligence Innovation Industrial Park also reflects Beijing’s confidence in developing the embodied intelligence industry. Centered on embodied intelligence, the park also includes data elements and life sciences, covering the entire “data-brain-small brain-ontology” industrial chain. Upstream companies include Kunlun Chip in the chip sector, Lansi Technology and Meixin Sheng in sensors and motors, Nuoyiteng Robots and BitFlyer in data collection, midstream companies like Wujie Power and Cell Wall in large models and ontology, and vertical application companies such as Rosenbot in healthcare and Yikong Zhijia in autonomous driving. This “up and down the building” ecosystem promotes technological synergy and innovation collaboration among enterprises, with a clear cluster effect emerging.

Wang Yu, Deputy Director of the Park Department of the Beijing Municipal Science and Technology Commission and Zhongguancun Management Committee, told media including Huaxia Times that Zhongguancun has formed a trillion-yuan-level industry cluster in next-generation information technology, along with nine other billion-yuan-level clusters in pharmaceuticals, healthcare, and integrated circuits. The AI industry’s overall strength ranks among the top globally, with nearly 30% of national large model filings, and leading the country in approvals for innovative drugs and medical devices. The park’s forward-looking layout includes future industries like embodied intelligence, brain-computer interfaces, and commercial space, with the world’s first embodied intelligence robot 4S store, the first embodied intelligence innovation industrial park nationwide, and the release of the country’s first brain-computer interface innovation development plan, accelerating the formation of a future industry ecosystem.

Editor: Xu Yunqian Chief Editor: Gong Peijia