Westlake Robotics has raised a nearly 100 million yuan Series A to develop a unified, whole-body large model for humanoid robots, adding a new university-derived contender to China’s fast-growing embodied AI sector. Gasgoo reported that Haiyuan Capital led the round and that the Hangzhou company has now raised 500 million yuan across four rounds in six months.
The financing is notable for what Westlake is trying to build. Many robotics companies focus first on the mechanical body, a single warehouse task, or a teleoperation demonstration. Westlake says it is pursuing a stack that includes a general AI “brain,” a whole-body control layer it calls a cerebellum, and its own robot body. The new funding is directed toward a unified whole-body model and an embodied-intelligence talent base, according to Gasgoo.
That focus puts Westlake in a crowded but important part of China’s AI market. Humanoid robots are becoming an increasingly visible destination for venture capital, industrial-policy support, and supply-chain investment. The difficult question is no longer whether companies can make a robot move in a controlled demonstration. It is whether they can give machines a transferable control system that works across tasks, robot bodies, and real environments.
The Series A funds a whole-body model rather than a single robot task
The nearly 100 million yuan round was led by Haiyuan Capital, Gasgoo reported. The company’s stated goal is to use the funding to develop a unified whole-body large model for humanoid robots. That wording matters because whole-body control requires coordination across balance, movement, perception, and actions. A model that works only on a fixed robot or a narrow scripted task may have limited use outside a demonstration.
AI Insider independently reported the financing and described Westlake as a Hangzhou company developing AI models, motion control, and humanoid hardware. It also noted that the company intends to expand its embodied-intelligence research team. The two reports agree on the round’s size, the lead investor, and the broad technology objective.
Westlake is described by Gasgoo as the first commercialization project to emerge from Westlake University’s independent AI and robotics research. That origin gives the company a research base, but it does not guarantee commercial scale. Turning a university project into a product company requires manufacturing, customer support, data collection, field testing, and a clear view of which tasks organizations will pay robots to perform.
China’s robot sector has already produced several high-profile public-market and private-market stories. EastFrontier recently covered how Unitree Robotics filed for a major IPO, a sign that investors are beginning to judge humanoid companies not only on technical demonstrations but also on revenue and operating discipline. Westlake’s Series A is much earlier in that path, but it shows how rapidly capital is moving toward new control-model approaches.
GAE is Westlake’s claimed foundation for motion control
Westlake’s technology pitch centers on its GAE general motion model. Gasgoo says the company claims the system can support millisecond-level teleoperation, programming-free action generalization, cross-body compatibility, and one-to-many control. These are company claims, not independent benchmark results, and should be read accordingly. The report does not provide a standardized comparison against other humanoid models or an outside test of reliability.
Still, the direction of the work is clear. If a robot needs hand-written instructions for every new movement, deployment costs rise quickly. A control model that learns from demonstrations and can transfer motions between different bodies could lower some of that programming burden. Likewise, a teleoperation system that lets one operator coordinate multiple robots could make data collection more efficient, provided it works reliably outside a laboratory setting.
The company says its architecture combines a general brain, a whole-body control layer, and a self-developed body. Gasgoo also reported a dual pre-training approach that brings together vision-language-action models and world models. Those are technically ambitious components. A vision-language-action model is meant to connect what a system perceives with instructions and actions, while a world model aims to represent possible changes in an environment. The sources do not disclose model size, training data, compute requirements, or the exact level of autonomy achieved.
Westlake’s claims must therefore be separated from demonstrated performance. The company may have a credible technical concept, but commercial customers will eventually ask whether the model works safely, repeatedly, and economically in their own facilities. That is especially important in the high-risk environments the company has identified, including industrial inspections, post-disaster rescue, and remote precision assembly.
Real-world data will determine whether the model can generalize
Gasgoo reported that Westlake has a strategic partnership with the Longyou local government to build a county-wide real-world training base. The stated purpose is to deploy the Westlake o1 humanoid robot in production and living environments and gather whole-body motion data. The report also says the company has partnerships with Alibaba, Intel, JD.com, and the Hong Kong University of Science and Technology to expand its high-quality motion dataset.
Those partnerships and deployment plans are significant because embodied AI depends heavily on data from the physical world. A language model can learn from text at internet scale. A humanoid robot needs observations of movement, objects, surfaces, tools, timing, and human behavior. Gathering that information safely and in enough variety is expensive. A real-world base can help, but the sources do not say how many robots will be deployed or when the facility will operate at scale.
China has been working to convert its manufacturing density into an advantage for robotic training and deployment. EastFrontier previously reported on China’s first automated humanoid robot factory in Guangdong. Westlake’s approach adds a different emphasis: a research-driven software and data stack designed to control the robot’s entire body.
The new funding gives Westlake more time to test that thesis. Its 500 million yuan in cumulative financing over six months signals investor confidence, but the company’s next proof point will be more concrete than another round. It will need to show that a whole-body model can move from controlled teleoperation to repeatable work in real settings. In China’s embodied AI market, that transition from impressive motion to dependable deployment is where the strongest companies will separate themselves.