Qingyan Jingzhun says B-round series is complete
Qingyan Jingzhun announced on July 13 that it had completed two financing rounds worth several hundred million yuan during June 2026, closing its B-round series.
The company said the B2 round, also worth several hundred million yuan, was led by Xingyuan Capital, with FAW Fusheng participating. The B3 round was led by BAIC Industrial Investment, with Yulon Group joining. Guoji Industry Fund was added in the latest financing mix.
Auto-sector capital and a state-backed fund joined the round
According to the report, the financing brought in deep industrial resources, including Guoji Industry Fund, described as a central state-owned enterprise fund.
Across the full B round, six carmaker-linked investors were involved: BAIC Industrial Investment, Xingyuan Capital, FAW Fusheng, Great Wall Capital, Shaanxi Auto Capital, and Yulon Group. The report said the concentration of auto-sector capital indicates that Qingyan Jingzhun’s physical AI engineering base and testing and validation system have been integrated into the core supply chains of major domestic automakers.
In June 2026, China’s Ministry of Industry and Information Technology and the State-owned Assets Supervision and Administration Commission jointly launched a special action plan for humanoid robots and embodied intelligence in real-world training. The plan requires embodied intelligence systems to move into real factory workstations rather than staying in laboratories.
The plan also states that by the end of 2026, key products including humanoid robots should complete application validation and regular deployment in a batch of representative scenarios, form more than 100 high-value application scenarios, and help build deployment capacity on the scale of 10,000 units.
Founding and management team
Founder and CEO Dong Han is a PhD from Tsinghua University and studied under Chinese Academy of Engineering academician Li Keqiang. He formally founded Qingyan Jingzhun in June 2018 through incubation at Tsinghua University.
The company said that over eight years, its AI inspection, simulation, and testing-validation products have entered the core supply chains of nearly all domestic automakers and power battery companies. It reported cumulative shipments of more than 10,000 units and operations in more than 30 countries, serving customers in new-energy vehicles, power batteries, energy storage, core components, mining, and power.
In its embodied intelligence segment, Jingzhun Vision CEO Cao Qitong has an engineering background from Stanford University and previously conducted interdisciplinary research in life sciences and AI at the Stanford Computer Research Institute. The report said related work was published as first author in a Nature sub-journal. At Qingyan Jingzhun, Cao oversees technology transfer, iteration roadmaps, and commercial deployment in industrial scenarios.
Zhao Ran, CTO of Jingzhun Vision and chief engineer for embodied intelligence at Qingyan Jingzhun, previously served as head of embodied infrastructure at Qianxun Intelligence and Zhifang Technology, two embodied AI companies described in the report as 20 billion yuan-level leaders. Zhao, a member of academician Ding Han’s team, has worked in robotics for more than a decade and has led teams building teleoperation, data collection, low-layer data closed loops, and simulation platforms from scratch.
From EV inspection to a physical AI base layer
Qingyan Jingzhun said it has completed a strategic upgrade from a new-energy vehicle inspection company to an engineering base layer for physical AI, aimed at supporting embodied intelligence deployment in industry.
The company said it has accumulated industrial field resources over multiple years and deployed more than 2,000 industrial sensing nodes in real workstations across sectors. Those scenarios range from PACK inspection for new-energy power batteries to final vehicle assembly, and from above-ground factories to underground mines.
It said it does not build robot bodies, but provides the training ground and validation system that allow robots to work in industrial environments.
TsingLoop and Robot-in-the-Loop testing
Qingyan Jingzhun said it developed the TsingLoop multimodal data engineering pipeline, designed to convert raw signals scattered across multiple systems into standardized and reusable data asset packages through aligned time, space, and semantic layers. Historical data can also be merged automatically with new data for ongoing iteration.
Built on that pipeline, the company is developing a Robot-in-the-Loop testing framework for industrial scenarios. The report described it as a closed loop of collection, simulation, validation, evaluation, and iteration. Robots or human workers perform tasks at real workstations while the system captures multimodal data including vision, force, touch, trajectories, process parameters, equipment status, and execution outcomes.
The system then reconstructs digital twin environments from real data, replays historical operating conditions, reproduces abnormal samples, and tests different action strategies in simulation. Qingyan Jingzhun said simulation is not the endpoint, so the setup also brings real robot bodies, controllers, end effectors, sensors, and simulated environments into a closed testing loop before direct use on customer production lines.
After on-site deployment, the evaluation module generates standardized reports covering metrics such as task success rate, cycle time, abnormality rate, collision risk, energy consumption, and stable runtime. Those results are then fed back into the TsingLoop pipeline to update models and strategies.
Stated long-term target
The company described its long-term goal as “one base layer, one brain, and 100 vertical scenario applications,” using a data engineering system as the base and an industrial cognitive world model as the brain for reusable physical intelligence in sectors including power, construction machinery, new-energy manufacturing, and mining.