This fundraising pace is indeed rare in the embodied AI sector.
Four rounds in half a year, with the valuation surging from CNY 10 billion to CNY 15 billion.
The recently announced progress by Astribot has once again drawn attention back to this robotics company, which was founded over three years ago.
However, what truly makes this company worth attention is not just the capital.
Behind it are early core members of Tencent Robotics X Laboratory who started this business; the chosen route is tendon-driven robots, a minority choice; and the goal is to connect AI models, robot bodies, and real-world task data.
Since its establishment in 2022, Astribot has been becoming an increasingly unique case among domestic embodied AI startups.
Astribot Raises CNY 1 Billion
Astribot just announced that it has completed a new round of approximately CNY 1 billion in financing, reaching a post-investment valuation of CNY 15 billion.
Institutions participating in this round of financing include Toukong Donghai, HongTai Fund, Ninghai Xingtaihe Fund, and Shanheng Capital, among others, with existing shareholder Xianmu Capital continuing to increase its investment.
Looking back, in June this year, Astribot just completed a series of Series B financing, with its valuation breaking through CNY 10 billion. In just a few short months, the company's valuation has further increased.
Looking at a longer timeline, Astribot's fundraising pace has been quite intensive:
Since 2026, Astribot has completed four rounds of financing in about half a year, from Series A+ and Series A++ to the Series B financing, and then to the latest CNY 1 billion round, with the valuation gradually entering the CNY 15 billion range from below CNY 10 billion / early stages.
Changes in the investor structure also reflect a shift in market focus.
In early robotics startups, capital focused more on technological direction and team background; as the industry enters the next stage, robot body capabilities, manufacturing capabilities, and deployment in real-world scenarios have begun to become new considerations.
The financing story of Astribot truly reflects the technological route it has chosen.
Who is Astribot? The Startup Founded by Tencent Robotics' "Employee No. 1"
The story of Astribot starts with Tencent Robotics X.
When it was founded at the end of 2022, embodied AI had not yet become the hot track it is today, but the founding team of Astribot had already accumulated many years of experience in the robotics field.
Founder Lai Jie previously participated in the Baidu Xiaodu robot project, and later joined the Tencent Robotics X Laboratory as the team's "Employee No. 1", engaging in long-term R&D related to robot systems, motion control, and smart hardware.
This experience also influenced Astribot's subsequent entrepreneurial direction.
Compared to some startups in recent years that enter robotics from large models, Astribot focused on robot bodies and control systems earlier. The reason is that once embodied AI truly enters the real-world environment, model task understanding is only the first step; robots also need to possess a body capable of executing tasks.
The challenge faced is not just whether actions can be generated, but whether continuous tasks can be stably completed in complex environments.
The direction Astribot chose to enter is robot body design, and its most distinctive technological route is tendon-driven robots.
Currently, a relatively common solution in humanoid robots is direct motor-driven joints. This type of structure has been verified by industrial robots for many years, features a mature control system, and is also the mainstream route in the current industry.
Tendon-driven mechanisms, on the other hand, adopt a different approach: transmitting power through cables, separating the driving part from the joint structure, making the overall structure closer to the relationship between muscles, tendons, and bones in the human body.
One advantage of this design is that it can reduce joint weight and improve the robot's movement flexibility and compliance. For future robots that need to handle complex objects, compliance means better contact control capabilities. For example, when grasping objects like glass cups or clothing, the robot needs to adjust its force based on contact feedback, rather than simply executing fixed movements.
Beyond the tendon-driven body, Astribot has further built a technical system of "AI model + embodied OS + robot body".
In this system, the robot body is responsible for executing actions, the AI model is responsible for understanding tasks and generating strategies, and the embodied OS takes on the connection between algorithms and hardware.
At the model level, Astribot has publicly demonstrated the Lumo embodied model and proposed the DuoCore fast-and-slow system architecture. Simply put, the high-level system is responsible for task understanding and planning, while the low-level system is responsible for rapid action response.
This design solves a practical contradiction in robot operation: complex tasks require stronger reasoning capabilities, but actions such as grasping and obstacle avoidance require low latency.
Besides the model architecture, the data system is also an important part of Astribot's technological route—exploring teleoperation and demonstration learning to obtain robot training data through human operation processes.
This direction is also reflected in the company's public research.
For example, research related to the Astribot Suite focuses on whole-body manipulation issues. Compared to training a robotic hand alone to complete grasping tasks, whole-body manipulation requires the robot to coordinate body movement, visual perception, dual-arm movements, and hand control, which is closer to complex tasks in real-world environments.
Meanwhile, StableVLA research focuses on the stability of Vision-Language-Action models. VLA models attempt to enable robots to understand language instructions and generate actions, but changes in objects, positions, and scenes in real-world environments can affect model performance. Therefore, how to improve model generalization capabilities is an important issue in current robot intelligence.
In addition, research such as Efficient-WAM explores the direction of world models, hoping that robots can predict future states before executing actions. For example, when grasping an object, it is not just about judging whether the current action is correct, but also predicting whether the object will slip and whether the force needs to be adjusted.
These research directions all revolve around a common goal: enabling robots to gradually adapt to open environments from executing fixed actions.
From a product perspective, Astribot is currently mainly promoting robot deployment through S1 and T1.
S1 first entered the public eye, with public videos demonstrating continuous tasks such as pouring water, organizing items, and operating tools. Compared to simple movement and single grasping, such tasks are closer to operational needs in real-world environments.
The subsequently launched T1 has begun to advance towards productization, emphasizing deployment and usage scenarios.
However, from an overall industry perspective, embodied AI is still in the stage of technical verification and commercial exploration. The tendon-driven route chosen by Astribot has certain characteristics in terms of flexibility and interaction capabilities, but it also faces issues such as control precision, long-term reliability, and manufacturing costs.
Similarly, although embodied models and data systems can improve robot adaptability, model generalization, data scale, and real-world environment stability remain issues that the entire industry needs to continuously address.
Financing, of course, reflects the market's continued attention to the possibility of Astribot's productization.
But Astribot has chosen a full-stack route, advancing synchronously from the tendon-driven robot body to the control system and then to the AI model, which distinguishes it from companies that only focus on a single link.
In particular, the "tendon-driven" form is now facing increasing skepticism. On the surface, it seems closer to human physiological structure and more in line with "first principles", but opposing views argue that conforming to first principles does not equal imitating human structure, but rather serving the task itself, after all, human structure is not perfect.
Not to mention issues such as the cost, reliability, and control precision of tendon-driven mechanisms.
This also means that Astribot, with a valuation of CNY 15 billion, not only needs to face challenges in multiple links such as robot hardware, algorithm models, and commercial deployment simultaneously, but also has to be a "lone brave" on a "niche" technological route, proving everyone wrong.