On the afternoon of October 22, Wang Xingxing, founder of Unitree Robotics, delivered a speech at the 2026 World Physical AI Conference, reviewing the company's progress over the past year and sharing his views on the development trends of physical AI.
Wang Xingxing revealed that the Unitree G1 humanoid robot is currently the world's best-selling humanoid robot, with cumulative production of approximately 10,000 to 20,000 units. "Today, other robotics companies around the world have successively launched similar products, and some products look almost indistinguishable," he said.
Discussing the impact of GPT-6's release on the robotics field, Wang Xingxing stated that Unitree's tests showed that with completely zero training, the model was already able to directly control robots to complete tasks such as painting, cleaning brushes, and playing Go. The robot could change colors on its own, clean its brushes by itself, and "the precision of placing stones is extremely high."
He judged that in the coming years, large model companies will continue to increase their attention to machine intelligence and the robotics industry. "At the fastest, one to two years; at the slowest, two to three years, the industry is expected to see more progress."
Wang Xingxing proposed that he hopes future robots will be able to complete about 80% of tasks using only voice and language commands. "At the fastest, there is a chance to achieve this next year; to be more conservative, it might take until the year after." He pointed out that the main issues not yet well resolved are the interaction between robots and real objects, as well as the alignment between AI model inputs and outputs and real robots, a large portion of which are engineering and technical problems.
Regarding the key directions for large-scale deployment of physical AI, Wang Xingxing believes there are three points. First, enhancing the large-scale manufacturing capacity of robot hardware. "We currently produce roughly tens of thousands of robots per year, as well as hundreds of thousands of joint motors and modules, but if we want to achieve truly large-scale popularization, the current production capacity is far from sufficient." Second, promoting independent research and development of physical AI models and their industrial deployment. Unitree will increase its self-research investment while cooperating with large model companies both domestically and internationally. Third, building larger-scale server clusters and computing power clusters, and possessing the corresponding manufacturing capacity for computing power chips.
Wang Xingxing also revealed that Unitree is developing "Unitree Preview," a self-evolution system for physical AI robots that allows AI models to independently collect papers, write code, train in simulation environments, and then test in real environments. The test results are scored and fed back by the AI model, forming a self-circulating loop.
At the end of his speech, Wang Xingxing admitted, "I am truly afraid of wasting every single day," and said, "Our generation truly has the opportunity to realize superintelligence or AGI, and also has the chance to elevate humanoid robot productivity to the next level."
Comments