Kinetix AI has completed more than 500 million yuan in Angel+ financing, LatePost reported on September 11. The Chinese startup builds humanoid robots and the data-collection tools that help developers teach them physical tasks. Its approach begins with a person doing something ordinary, such as folding clothes, and tries to turn that movement into training material a robot can use. The financing remains attributed to the report.
The company's Kai Ego system records a task from the human operator's viewpoint. A wearable camera array captures the surroundings, hand movements and body pose together; processing software reconstructs the scene in three dimensions and adds descriptions of the actions. That gives a robotics developer more than a video to imitate. It connects what the operator saw with how the body moved and where objects were. Kinetix's product page shows examples involving folding clothes and cartons and organizing tools, which it presents as demonstrations of its training approach.
The other half of the system is the body doing the work. Kinetix develops the full-size KaiBot humanoid and a dexterous robotic hand, alongside its models and data infrastructure. LatePost reports that the robot has 115 degrees of freedom, a measure of its independently movable joints and motions. The company argues that a human-like body is better suited to learning from human movements and operating in spaces already arranged around people. That is its design choice, not proof that a humanoid will outperform a simpler machine at every job.
Kinetix lists homes, factories, restaurants and hotels as intended settings, including sorting, carrying and handling objects. These are target uses, not a verified customer roster. Its data tools address the development work before a robot can reliably perform such jobs: collecting demonstrations, describing them consistently and testing whether the resulting behavior transfers to different hardware. The company says its training pipeline has been validated across multiple robot designs. That claim is distinct from establishing dependable operation across an entire workplace.
LatePost names Vertex, Fangguang Capital and Wanshi Capital among the investors adding capital, and says the proceeds will support joint development of the data, models and robot hardware. It describes earlier financings and a team approaching 200 people but does not disclose a valuation. Captables is recording the reported 500 million yuan threshold, rather than treating the approximate $75 million headline in later coverage as a separate dollar-denominated round.
The practical test is whether human demonstrations remain useful when the task changes slightly: a different carton, a crumpled sleeve or tools placed in unfamiliar positions. Capturing the motion is only the starting point. For Kinetix's approach to become useful factory or household equipment, the trained robot must adapt to those variations without requiring developers to rebuild the task each time. That transfer from demonstration to repeatable work is the business problem behind the financing.