Investor.ai learned that on July 13th, Qingyan Jingzhun announced it had swiftly completed two funding rounds totaling hundreds of millions of RMB within June, officially closing its Series B financing.
A lineup of 'State-Owned Giants + Half the Auto Industry' emerges: the hundreds of millions of RMB B2 round was led by Xingyuan Capital, with Yifu Sheng participating. The subsequent B3 round was led by BAIC Capital, with Yulon Group participating. Notably, this round also saw the addition of Sinomach Industry Fund.
In June 2026, the Ministry of Industry and Information Technology and the State-owned Assets Supervision and Administration Commission jointly launched the 'Humanoid Robotics and Embodied AI Real-World Training Special Initiative,' mandating that embodied intelligence must not just operate in labs but enter real factory workstations to start 'operation mode.'
Long before this, Qingyan Jingzhun had strategically positioned itself in the Physical AI engineering foundation. Through eight years of accumulation in industrial settings, it has enabled embodied robots to 'learn to work' in real, complex, and harsh industrial scenarios, achieving true implementation.
Rare Involvement of Central Enterprise Capital
Looking at Qingyan Jingzhun's latest funding round, the industrial resources are exceptionally rich.
Among them is a central enterprise fund—Sinomach Industry Fund.
Even more notable is the formation of a relatively uncommon automotive capital matrix—the entire Series B round attracted six automakers: BAIC Capital, Xingyuan Capital, Yifu Sheng, Great Wall Capital, Shaanxi Auto Capital, Yulon Group. The concentrated investment from automakers signifies that Qingyan Jingzhun's Physical AI engineering foundation and testing/validation system have been embedded into the core supply chains of mainstream domestic automakers. This is recognition from upstream and downstream partners in the automotive industry.
This highly vertical and strongly industry-specific investor profile demonstrates that the investment logic in the embodied intelligence space has shifted—capital is no longer blindly chasing after demo videos of humanoid robots. Instead, it is heavily betting on Physical AI infrastructure companies that master real industrial scenarios, possess high-quality data closed loops, and have engineering implementation capabilities.
For Physical AI to truly land, it must cross hurdles like product development, supply chain, on-site delivery, customer service, and ongoing operations. In other words, it requires real-world trials and must be usable on production lines.
Deep integration of capital and business is essential to ensure continuous, stable access to real industrial scenarios, thus forming a virtuous cycle.
As mentioned in the 'Real-World Training Special Initiative,' by the end of 2026, key products like humanoid robots are expected to complete application validation and routine deployment in a batch of representative scenarios, initiating operation mode; condense and form over a hundred high-value application scenarios, further enriching the embodied intelligence application spectrum, and driving the formation of a ten-thousand-unit scale implementation capacity.
Qingyan Jingzhun can be said to have perfectly positioned itself. These two funding rounds coincide with a critical strategic pivot: starting from establishing a closed loop for new energy Physical Intelligence, gradually moving towards broader industrial scenarios, dedicated to building an engineering foundation for Industrial Physical AI, and deeply investing in the embodied intelligence field.
From this perspective, its breakthrough is not just a single technology point but a composite barrier formed by real-scenario access, data production capability, testing/evaluation systems, engineering delivery capability, and world model capability. It's about completing the full-chain layout ahead of policy directives.
A Strong Alliance of Tsinghua, Stanford, and Robotics Industry Veterans
Dong Han, founder and CEO of Qingyan Jingzhun, holds a PhD from Tsinghua University, where he studied under Professor Li Keqiang, a member of the Chinese Academy of Engineering. He officially founded Qingyan Jingzhun in June 2018, incubated by Tsinghua University.
Over eight years, Qingyan Jingzhun has introduced its AI inspection, simulation, and test validation products into the core supply chains of nearly all domestic automakers and power battery manufacturers, shipping over ten thousand units, deploying in over 30 countries. Industrial clients cover core sectors like new energy vehicles, power batteries, energy storage, key components, mining, and power.
(From left to right
Cao Qitong, CEO of Qingyan Jingzhun's Embodied Intelligence division—Jingzhun Shijie—brings an academic background from Stanford University's engineering school. She previously conducted cross-disciplinary research on life sciences and AI at the Stanford Computer Research Institute, with related results published as first author in a *Nature* sub-journal. At Qingyan Jingzhun, Cao Qitong primarily oversees the company's technology migration, iteration roadmap, and commercial scenario implementation, highlighting the company's core advantage in tackling the final mile of industrial embodied intelligence deployment.
Her core research involves deducing system state evolution from high-dimensional, multimodal, dynamic data. Transferred to industrial scenarios, the fundamental problem is similar: a robot sees not just a workpiece but a dynamic physical system composed of visual, force, tactile, process parameters, and environmental variables. This aligns perfectly with the industrial physical world model Qingyan Jingzhun is building.
Zhao Ran, Chief Engineer of Embodied Intelligence and CTO of Jingzhun Shijie, previously served as Head of Embodied Infrastructure at two leading embodied intelligence companies valued at over 20 billion RMB, Qianxun Intelligence and Zhipingfang Technology. Dr. Zhao Ran's addition provides solid assurance for Qingyan Jingzhun's efforts in building embodied infrastructure and engineering. As a team member of robotics authority Professor Ding Han, Dr. Zhao Ran has over a decade of deep experience in robotics, combining solid academic foundation with industrial implementation experience.
He led teams to build teleoperation, data acquisition, underlying data closed loops, and simulation platforms from scratch. His over-a-decade of robotics expertise enables a more systematic integration of key links like robotics hardware, data, simulation, and models, forming the core capabilities required for embodied intelligent infrastructure construction. His platformization and engineering experience, combined with the team's deep R&D accumulation, further drives the integration of cutting-edge academic genes and pragmatic industrial engineering capabilities.
Thus, the team now combines world-class foresight, industrial engineering heritage, and billion-scale commercial validation, positioning itself at the forefront of China's embodied intelligence industrialization, recognized as the industry's 'Technical Anchor' and 'Deployment Navigator.'
Physical AI Engineering Foundation
Building upon this foundation, Qingyan Jingzhun has successfully completed a strategic upgrade and capability expansion—evolving from a new energy vehicle testing company into a Physical AI Engineering Foundation, aiming to serve as the Physical AI foundation for embodied intelligence deployment in the industrial sector.
Corresponding to the 'Real-World Training Special Initiative,' the industrial sites Qingyan Jingzhun has accumulated over the years are already prepared. Across different industrial fields, over 2,000 industrial perception nodes they've deployed are positioned on real workstations, from new energy power battery pack inspection to vehicle final assembly, from ground factories to underground mines, transforming key workstations into data fields and training grounds for embodied intelligence. These scenarios have data, workstations, and real tasks, offering the best validation of value.
The embodied model is the 'brain,' while Qingyan Jingzhun provides the training base and 'textbook' that enables the brain to learn 'body coordination' and validate its capabilities. They don't build robot bodies, but they create the ability for robots to work effectively in industrial settings.
Furthermore, the 'Real-World Training Special Initiative' emphasizes application traction, using real-world scenario training to continuously optimize embodied AI model algorithms and accumulate high-quality real-robot data.
Today's Qingyan Jingzhun effectively operates as a Physical AI Data Foundation provider.
Qingyan Jingzhun independently developed the TsingLoop multimodal data engineering pipeline. It transforms raw signals scattered across multiple systems, through unified time-space-semantic alignment, into standardized, reusable data asset packages. Data collected once, processed through this pipeline, is upgraded from raw data to industrial 'data assets'; historical data can automatically integrate with new data, continuously iterating, forming a growing data flywheel.
Moreover, based on the TsingLoop multimodal data engineering pipeline, Qingyan Jingzhun is constructing a Robot-in-the-Loop testing system for industrial scenarios.
This system can be understood as an industrial embodied intelligence version of the 'Acquisition-Simulation-Validation-Evaluation-Iteration' closed loop. A robot or human worker executes tasks at a real workstation while TsingLoop simultaneously collects multimodal data like vision, force, touch, trajectory, process parameters, equipment status, and execution results. Subsequently, the system reconstructs a digital twin scene based on real data, replays historical operating conditions, reproduces anomaly samples, and conducts low-cost, high-frequency hypothetical reasoning on different action strategies in the simulation environment.
However, simulation is not the end goal. Industrial robots must ultimately enter real workshops, requiring crossing the simulation-to-reality gap. Therefore, Qingyan Jingzhun introduces robot-in-the-loop testing: enabling closed-loop interaction between the real robot body, controller, end-effector, sensors, and the simulation scene, validating action strategies, force control boundaries, safety envelopes, and anomaly handling mechanisms in advance without directly occupying the customer's production line.
After deployment on-site, the evaluation module continuously outputs standardized evaluation reports, including metrics like task success rate, cycle time, anomaly rate, collision risk, energy consumption, and stable operation duration. These evaluation results are not just acceptance criteria but also feed back into the TsingLoop data pipeline, driving continuous model optimization and strategy updates.
It systematically answers three more critical questions: Can it stably complete tasks under real operating conditions? Can it pass customer acceptance? Can it be reused on the next production line? In this way, a data foundation is achieved.
To date, Qingyan Jingzhun's ultimate vision has taken shape: 'One foundation, one brain, a hundred vertical scenario applications.' Using a data engineering system as the foundation and an industrial cognitive world model as the brain, it aims to accumulate reusable physical intelligence across hundreds of well-defined industrial tasks in power, construction machinery, new energy manufacturing, mining, and other fields.
At this critical juncture where Physical AI moves from concept to industrial implementation, industrial capital is betting on Qingyan Jingzhun precisely for its irreplaceable scenario implementation capability.
While the industry is still debating algorithm approaches, Qingyan Jingzhun, deeply rooted in industrial sites and quietly forging a Physical AI engineering foundation, has quietly become the most essential 'shovel seller' in the embodied intelligence era.
In the second half of the game, its importance is already self-evident.
