Chinese electric car maker Nio is said to have assembled a team of about 20 people dedicated to researching a robot dog project.

According to Chinese media reports, the project is headed by former Momenta algorithm expert Xu Kang.

Xu joined Nio's self-driving team in 2020, and this shift to the robot dog project marks Nio's further exploration in the field of Embodied AI.

Although robot dogs are less expensive to develop than embodied robots, the development of embodied intelligence is still in its early stages.

"Intelligent cars as a limited-scene robot, the development of intelligent driving all still need time, with a wider use of the scene of embodied intelligence must be much more difficult," said an industry insider.

In the automotive industry, Tesla is the leader in the field of embodied intelligence, and in 2022, Tesla demonstrated its first humanoid robot "Optimus" at AI DAY, which shares the fully automated driving system FSD with Tesla's cars, and possesses the same visual perception and computational capabilities.

China's Xpeng followed suit, acquiring robotics startup Dogotix in 2020, establishing Pxing Intelligence, and launching Iron, an AI robot capable of performing production tasks, in 2024.

However, Li Auto, one of Nio's rivals, is taking a cautious approach to robotics.

Li Auto CEO Li Xiang said a month ago that Li Auto will definitely get involved in humanoid robots in the future, but its current priority is to solve the technical difficulties of level 4 (L4) self-driving self-driving cars.

Analysts believe that autonomous driving as an important scene of embodied intelligence, its "perception - decision-making - control" algorithmic architecture and embodied intelligence, highly similar technology path, and algorithms and components can be reused.

Driven by GAC Group, BYD, Nio and other automakers, embodied intelligence has become one of the hottest tracks in China's capital market.

According to incomplete statistics, at least 60 financing rounds have been completed in China's embodied intelligence sector in the past year, totaling more than 5 billion yuan.

At CES 2025 in early January, Nvidia CEO Jensen Huang showed 14 humanoid robots, six of which came from Chinese companies, namely Unitree Robotics, AgiBot, Fourier, Xpeng, Galbot, and RobotEra.

Among them, Xpeng's AI robot, Iron, is close to adult size and uses a Turing AI chip with 3,000T of computing power.

Unitree Robotics's first general-purpose humanoid robot, H1, weighs about 47kg, has a maximum torque of 360N.m, and moves at a speed of 3.3m/s.

Galbot lent Huang to show Galbot, a wheeled, two-armed, body-foldable humanoid robot.

Fourier, on the other hand, demonstrated GR-2, a new-generation humanoid robot equipped with the new FSA2.0 actuator and 12-degree-of-freedom dexterity hand.

Star1, a humanoid robot equipped with a fully direct-drive five-fingered dexterous hand and a maximum joint torque of 400Nm, was demonstrated by RobotEra.

AgiBot demonstrated the Expedition A2 robot, which is based on a speech grand model, enhancing real-time dialog capabilities while supporting L4-level autonomous mobility.