Founded by a Stanford team and continuously funded by Sequoia and P7, this AI - enabled company raises hundreds of millions in another round of financing | Exclusive report from Yingke

Author | Huang Nan

Editor | Yuan Silai

Yingke learned that the embodied intelligence enterprise "Qiongche Intelligence" recently completed a Pre - A++ round of financing worth hundreds of millions of yuan. This round of financing was participated in by institutions such as Shengyu Investment, Tsingke Venture Capital, Jiayu Capital, Yunqi Capital, and Shanghai Science and Technology Innovation Fund. Prosperity7 and Sequoia China, as the company's Pre - A round investors, have increased their investment for three consecutive rounds. The financing funds will be used to further accelerate the company's breakthroughs in areas such as embodied intelligence basic models, data collection and evaluation, and promote the exploration of its commercial applications in scenarios such as retail fulfillment, household services, and food processing.

Since its establishment at the end of 2023, "Qiongche Intelligence" has completed four rounds of financing. This is also another round of financing completed by the enterprise after the Pre - A+ round of financing at the end of 2024.

"Qiongche Intelligence" focuses on the development and application of embodied intelligence technology. Co - founder Lu Cewu is a postdoctoral fellow at the Stanford Artificial Intelligence Laboratory. He is currently the deputy dean and professor of the School of Artificial Intelligence at Shanghai Jiao Tong University and is one of the earliest scholars in China to study the field of embodied intelligence. Another co - founder, Wang Shiquan, is a doctor from the Stanford Bionics and Dexterous Manipulation Laboratory and the Artificial Intelligence Laboratory (robotics direction). He founded FeiXi Technology in 2016, pioneering a humanoid general robot technology route based on extreme force control and hierarchical intelligence. Its adaptive robots have been widely used in real - world scenarios.

The current development of embodied intelligence faces many obstacles. The core problem lies in how to accurately and effectively describe the physical world and clarify the interaction mechanism between the physical world and robots.

From the perspective of perception, although significant progress has been made in large natural language and visual - language models, there are still obvious deficiencies when dealing with complex problems in the real world. The physical environment of the real world contains rich information, such as the physical properties of objects like weight, material, and hardness; spatial relationships such as position, distance, and angle; dynamic changes such as motion trajectory, speed, and acceleration; and core force behavior data that directly affects operations, such as contact force, friction, and torque between objects. Relying solely on visual and language modalities, it is difficult to comprehensively and accurately describe this information, and it cannot provide sufficient decision - making basis for embodied intelligence.

On the data side, training large embodied intelligence models faces difficulties in data acquisition. Different from model training in other fields, embodied intelligence requires a large amount of data reflecting real physical interactions. The acquisition of these data requires the actual operation of robots in the real environment, which is costly and inefficient. At the same time, due to the lack of unified data standards, data from different sources vary in format, annotation, and quality, making it difficult to effectively integrate and utilize them, further hindering the development of large embodied intelligence models.

"Qiongche Intelligence" has been continuously making efforts in the field of embodied intelligence. By rapidly iterating the physical world model and the force - centered behavior model, it has significantly improved the robot's ability to model, understand, predict, and interact with the physical environment. Its self - developed 3D visual imitation learning framework can further enhance the robot's generalization ability, task execution success rate, and robustness in complex environments.

Business of "Qiongche Intelligence" - Covers the entire cycle of technology pre - research and product development (Source/Enterprise)

In response to the pain points of high data collection costs and the difficulty of large - scale deployment in real scenarios at low cost, "Qiongche Intelligence" combined with its self - developed algorithm model, proposed a "production - accompanied" data collection method that does not require downtime and developed a corresponding data collection system, which is expected to break through the bottleneck of high - quality operation data collection. Since its release, the system has received nearly a hundred orders in just half a year, thanks to its efficient and convenient data collection capabilities.

Relying on advanced algorithms and data support, the core product of "Qiongche Intelligence" - the Qiongche Embodied Brain (Noematrix Brain), already has the full - closed - loop capabilities of instruction reasoning and decomposition, task planning, object classification, environmental perception, autonomous navigation, and general skill operation.

On this basis, "Qiongche Intelligence" has built a complete product matrix, including "Noematrix Brain + Training Platform + DevPlatform", "hardware body", and "CoMiner accompanied data collection system", which has the ability to be deployed across different scenarios. It focuses on the intelligent and automated needs of scenarios such as retail fulfillment, household services, and food processing and has entered the stage of commercial application.

Taking the household service scenario as an example, "Qiongche Intelligence" has reached in - depth cooperation with leading home appliance enterprises to jointly promote the research and development and application of household service robots. At the recent AWE 2025, the household robot for washing and care scenarios jointly developed by "Qiongche Intelligence" and the enterprise achieved automated operations from clothing perception, precise dosing, washing and drying to clothing retrieval without manual intervention.

AWE Washing and Care Robot of "Qiongche Intelligence" (Source/Enterprise)

In the field of food processing, "Qiongche Intelligence" has reached a cooperation intention with well - known food manufacturers. The two parties will accelerate the intelligent and automated transformation of complex food production and processing lines, and improve food production efficiency and quality in a "human - machine collaboration" mode.

After this round of financing, "Qiongche Intelligence" will focus on improving the generality of large embodied intelligence models. Relying on the Noematrix CoMiner accompanied data collection system, it will efficiently obtain high - quality operation data, break through the data bottleneck, and accelerate model iteration and performance improvement.

Floor Cleaning Robot of "Qiongche Intelligence" (Source/Enterprise)

In addition, "Qiongche Intelligence" will continue to accelerate the exploration of the application of embodied intelligence in more scenarios such as retail fulfillment, household services, and food processing. It will cooperate with industry manufacturers and scientific research institutions from all parties to jointly build high - quality and large - scale embodied intelligence data infrastructure, promote the accelerated transformation of technology from research to industry, and contribute to the large - scale application of embodied intelligence technology.