Tsinghua Precision raised several hundred million yuan in a financing round backed by a national team player in the equipment manufacturing sector.

According to I-News, today (July 13), Qingyan Precision announced that it has rapidly completed two rounds of financing totaling hundreds of millions of RMB within June, marking the official conclusion of its Series B financing.

A lineup of "state-backed capital + nearly the entire automotive industry" has emerged: the Series B2 round of hundreds of millions of RMB was led by Xingyuan Capital, with FAW Fusheng participating in the investment; the subsequent Series B3 round was led by BAIC Capital, with Yulon Group joining in. This round also saw new participation from SINOMACH Industrial Fund.

Back in June 2026, the Ministry of Industry and Information Technology and the State-owned Assets Supervision and Administration Commission jointly launched the "Special Action for Real-Scene Practical Training of Humanoid Robots and Embodied Intelligence", requiring that embodied intelligence cannot only run in laboratories, but must enter real factory workstations and activate the "operation mode".

Long before that, Qingyan Precision had already secured its position in the engineering foundation of Physical AI. With 8 years of accumulation in industrial sites, it has enabled embodied robots to "learn to work" in real, complex, and demanding industrial scenarios, achieving genuine practical implementation.

Rare Investment from Central SOE Capital

Looking at the bigger picture, the industrial resources behind Qingyan Precision's this round of financing are remarkably abundant.

Among the investors is a central SOE fund — SINOMACH Industrial Fund.

What is even rarer is the formation of an uncommon automotive industry capital matrix: the entire Series B round has gathered 6 automotive enterprises: BAIC Capital, Xingyuan Capital, FAW Fusheng, Great Wall Capital, Shaanxi Automobile Capital, and Yulon Group. The concentrated capital injection from automotive enterprises means that Qingyan Precision's Physical AI engineering foundation and testing verification system have been embedded into the core supply chain of mainstream domestic automotive manufacturers. This is recognition from all links of the automotive industry chain.

The highly specialized investment lineup with strong industrial attributes proves that the investment logic in the second half of the embodied intelligence industry has shifted — capital no longer blindly chases after demo videos of humanoid robots, but heavily invests in Physical AI infrastructure enterprises that master real industrial scenarios, have high-quality data closed loops, and possess engineering implementation capabilities.

For Physical AI to be truly implemented, it must go through stages including product development, supply chain management, on-site delivery, customer service, and continuous operation and maintenance. In other words, it needs real-world testing to ensure usability on production lines.

Only deep integration between capital and business can guarantee continuous and stable access to real industrial scenarios, thus forming a virtuous cycle.

As mentioned in the "Special Action for Real-Scene Practical Training", by the end of 2026, key products such as humanoid robots will complete application verification and regular deployment in a number of representative scenarios, activating the operation mode; more than 100 high-value application scenarios will be refined and formed, further enriching the embodied intelligence application ecosystem and driving the formation of implementation capabilities at a scale of 10,000 units.

Qingyan Precision has precisely positioned itself, and both of these two financing rounds are accompanied by key strategic shifts: starting from completing the closed loop of new energy physical intelligence, it is gradually expanding into broader industrial scenarios, committed to building an engineering foundation for industrial Physical AI, and deeply deploying in the embodied intelligence field.

From this perspective, its breakthrough does not only rely on a single technology, but comes from a composite barrier formed by real scenario access, data production capabilities, testing and evaluation systems, engineering delivery capabilities, and world model capabilities. More importantly, it has completed the full-chain layout in advance before relevant policies were introduced.

Strong Alliance of Tsinghua University, Stanford University, and Industry Veterans in Robotics

Dong Han, Founder and CEO of Qingyan Precision, pursued his doctoral degree at Tsinghua University, under the supervision of Academician Li Keqiang of the Chinese Academy of Engineering. He officially founded Qingyan Precision in June 2018, incubated by Tsinghua University.

Over the 8 years since its establishment, Qingyan Precision has integrated its AI inspection, simulation, and testing verification products into the core supply chains of almost all domestic vehicle manufacturers and power battery enterprises. It has shipped more than 10,000 units, implemented its solutions in over 30 countries, and its industrial customers cover core sectors such as new energy vehicles, power batteries, energy storage, core components, mining, and electric power.

(From left to right)

Cao Yitong, CEO of Precision Vision, the embodied intelligence division of Qingyan Precision, holds an academic background in engineering from Stanford University. She once conducted cross-disciplinary research on life sciences and AI at the Stanford Computer Science Institute, and her relevant research results were published as the first author in a sub-journal of *Nature*. At Qingyan Precision, Cao Yitong mainly coordinates the company's technology migration and iteration roadmap as well as commercial scenario implementation, highlighting the enterprise's core advantages in overcoming the last-mile challenge of industrial embodied intelligence implementation.

Her core research areas involve the evolution laws of system states derived from high-dimensional, multi-modal, dynamic data. When migrated to industrial scenarios, the essential problem is similar: robots perceive not just a workpiece, but a dynamic physical system composed of vision, force perception, tactile perception, process parameters, and environmental variables. This perfectly aligns with the industrial physical world model built by Qingyan Precision.

Zhao Ran, Chief Engineer of Embodied Intelligence at Qingyan Precision and CTO of Precision Vision, previously served as the head of Embodied Infrastructure at two leading embodied enterprises valued at 20 billion RMB, Qianxun Intelligence and Zhi Square Technology. The joining of Dr. Zhao Ran provides a solid guarantee for Qingyan Precision to build embodied infrastructure and advance engineering implementation. As a member of the team of Academician Ding Han, a towering figure in the robotics field, Dr. Zhao Ran has been deeply engaged in the robotics industry for more than 10 years, with both solid academic foundation and industrial implementation experience.

He once led the team to build a teleoperation, data collection, underlying data closed loop, and simulation platform from scratch. With more than 10 years of accumulated robotics technology, he can systematically connect key links such as robot bodies, data, simulation, and models, forming the core capabilities required for embodied intelligence infrastructure construction. His platform-based and engineering experience, combined with the team's profound R&D accumulation, further promotes the deep integration of top-tier academic genes and down-to-earth industrial engineering capabilities.

Since then, the team, which integrates world-class forward-looking vision, industrial engineering heritage, and tens of billions of commercial verification experience, has stood at the forefront of China's embodied intelligence industrialization, and has become the industry-recognized "technical anchor" and "implementation navigator".

Engineering Foundation of Physical AI

Based on all these foundations, Qingyan Precision has successfully completed its strategic upgrading and capability expansion — transforming from a new energy vehicle inspection enterprise into a Physical AI engineering foundation, aiming to act as the Physical AI foundation for embodied intelligence implementation in the industrial field.

In response to the "Special Action for Real-Scene Practical Training", the industrial sites accumulated by Qingyan Precision over the years are already fully prepared. In different industrial sectors, more than 2,000 industrial perception nodes it has accumulated are deployed on real workstations, ranging from PACK inspection of new energy power batteries to vehicle final assembly, from ground factories to underground mines, transforming key workstations into data fields and training grounds for embodied intelligence. These scenarios with sufficient data, dedicated workstations, and real operation tasks are the most effective in verifying practical value.

The embodied model is the "brain", and Qingyan Precision provides a practical training base and teaching materials that enable the brain to learn "body coordination" and verify its capabilities; it does not manufacture robot bodies, but it empowers robots with the ability to work in industrial sites.

In addition, the "Special Action for Real-Scene Practical Training" mentions that we must adhere to application-driven development, continuously optimize embodied intelligence model algorithms through real-scene training, and accumulate high-quality real-device data.

Nowadays, Qingyan Precision has essentially become a Physical AI data foundation provider.

Qingyan Precision has independently developed the TsingLoop multi-modal data engineering pipeline — it converts raw signals scattered across multiple systems into standardized, reusable data asset packages through unified time-space-semantic alignment. Data collected once can be upgraded into industrial "data assets" after being processed by the pipeline; historical data can be automatically integrated with new data and continuously iterated, forming a perpetually growing data flywheel.

Furthermore, based on the TsingLoop multi-modal data engineering pipeline, Qingyan Precision is building a Robot-in-the-Loop testing system oriented to industrial scenarios.

This system can be understood as the industrial embodied intelligence version of the "collection-simulation-verification-evaluation-iteration" closed loop: when robots or workers perform tasks on real workstations, TsingLoop synchronously collects multi-modal data including vision, force perception, tactile perception, motion trajectories, process parameters, equipment status, and execution results; subsequently, the system reconstructs a digital twin scenario based on real data, replays historical working conditions in the simulation environment, reproduces abnormal samples, and conducts low-cost, high-frequency hypothetical deduction for different action strategies.

However, simulation is not the end point. For industrial robots to eventually enter real workshops, they must overcome the reality gap between the virtual and the physical world. Therefore, Qingyan Precision will further introduce Robot-in-the-Loop testing: enabling the real robot body, controller, end effector, sensors, and simulation scenario to form a closed-loop interaction, and verifying action strategies, force control boundaries, safety envelopes, and abnormal takeover mechanisms in advance without directly occupying the customer's production line.

After deployment on site, the evaluation module will continuously output standardized evaluation reports, covering indicators such as task success rate, cycle time, anomaly rate, collision risk, energy consumption, and stable operation duration. These evaluation results are not only the basis for acceptance, but also fed back into the TsingLoop data pipeline to drive continuous model optimization and strategy updating.

This systematically addresses three more critical questions: whether the system can stably complete tasks under real working conditions, whether it can pass customer acceptance, and whether it can be reused on the next production line. In this way, a complete data foundation is established.

Having come so far, Qingyan Precision has outlined its ultimate vision: "One Foundation, One Brain, Hundreds of Vertical Scenario Applications", taking the data engineering system as the foundation and the industrial cognitive world model as the brain, to precipitate reusable physical intelligence in hundreds of clearly defined industrial tasks such as electric power, engineering machinery, new energy manufacturing, and mining.

At the critical juncture where Physical AI is transitioning from concept to industrial implementation, industrial capital has heavily invested in Qingyan Precision, precisely because of its irreplaceable scenario implementation capabilities.

While the industry is still debating algorithmic approaches, Qingyan Precision, which has taken root in industrial sites and quietly forged its Physical AI engineering foundation, has quietly become the most critical "shovel seller" in the embodied intelligence era.

In the second half of the industry's development, this significance is self-evident.