On November 5, Guanglun Intelligent (Beijing) Technology Co., Ltd. (hereinafter referred to as Guanglun Intelligent) announced a strategic partnership with Yushu Technology Co.,Ltd. (688836.SH), with the two parties planning to collaborate on robot model training, capability validation, and real-world scenario applications.
Guanglun Intelligent will provide training data, simulation evaluation, and deployment feedback platforms, while Yushu Technology Co.,Ltd. will supply the robot hardware, motion control, and perception systems. The two parties have not yet disclosed the cooperation amount, procurement scale, or execution timeline.
According to the cooperation arrangement disclosed by Guanglun Intelligent, in the task training phase, Guanglun Intelligent combines the degrees of freedom, load capacity, and end-effector characteristics of Unitree's robots to supply human video data and simulation-synthesized data. In the evaluation phase, relying on Guanglun Intelligent's self-developed simulation foundation SimFoundry and evaluation platform RoboFinals, the robot task performance undergoes reproducible and comparable validation. In the deployment phase, through the RoboStack platform, failure samples from real-machine operations are collected and fed back into the next round of training. The above arrangements did not disclose specific projects or delivery progress.
Guanglun Intelligent was founded in 2023, and its founder Xie Chen previously served as the head of autonomous driving simulation at NVIDIA and NIO. Since 2026, the company has disclosed funding progress multiple times. Based on publicly announced amounts, the two rounds in March and June totaled 2 billion yuan, while a round led by Ant Group in May did not disclose the amount.
Regarding operational data, Guanglun Intelligent previously disclosed that in the first quarter of 2026, new orders reached 550 million yuan, with cumulative delivery of over 1.5 million hours of human video data. Yang Haibo, co-founder and president of Guanglun Intelligent, previously stated that some high-quality scenario data can achieve over 10 times resale. The company has not disclosed the statistical method for this metric.
Order amount does not equal operating revenue, delivery duration does not equal database inventory scale, and resale multiple also differs from customer repurchase rate. The three metrics cannot be directly inferred from one another.
Yushu Technology Co.,Ltd. listed on the STAR Market on August 19 this year. On October 28, Yushu Technology Co.,Ltd. disclosed its third-quarter report, showing revenue of 1.817 billion yuan for the first three quarters, a year-on-year increase of 55.71%; R&D investment was 263 million yuan, a year-on-year increase of 191.48%, mainly directed toward embodied large model research, computing power, and new product development. The training, evaluation, and deployment aspects involved in this cooperation between Guanglun Intelligent and Yushu Technology Co.,Ltd. are business-related to the R&D direction that Unitree is currently advancing.
Guanglun Intelligent's previously disclosed partners include model and computing power companies such as NVIDIA and Google, as well as hardware manufacturers such as AgiBot and Galbot, both of which are simultaneously building their own data systems.
The cooperation between Guanglun Intelligent and Yushu Technology Co.,Ltd. expands the publicly disclosed scope of hardware manufacturer partnerships. Currently, the two parties have not yet disclosed actual procurement volumes, data delivery scale, or project revenue.
For independent embodied data service providers, the key to the business model lies in whether training data and simulation platforms can be reused across customers and across robot models, thereby amortizing delivery costs.
Some leading hardware manufacturers have already built their own data systems, and the space for third-party data services needs to be verified through actual procurement scale, reuse effects, and delivery costs.
Going forward, attention should still be paid to the launch and delivery of specific cooperation projects between the two parties, the reproducible evaluation results of Unitree's robot models, and the customer procurement scale and revenue contribution of Guanglun Intelligent's data services.
Comments