After years of developing humanoid robots that run, jump, do somersaults and perform complex movements, China’s robotics industry is entering a new phase: creating robots that not only move, but also understand, learn and make decisions.

According to a report by Reuters, Chinese companies are shifting their focus from developing robots’ physical components to artificial intelligence that will enable them to perform different tasks without needing to be programmed separately for each one.

AD PLACEHOLDER

This field is known as Embodied AI, or embodied artificial intelligence, and is considered one of the most promising areas of technological development.

Interest in this technology is growing rapidly. According to data from research firm ITjuzi, Chinese embodied AI companies secured 93.5 billion yuan, around $13 billion, through 322 investment deals in the first half of 2026 alone. This amount is five times higher than in the same period of the previous year.

Investments are increasingly flowing toward companies developing robots’ intelligent systems, as competition in manufacturing physical hardware is driving down prices and reducing the scope for profits from robot sales alone.

The development of this industry is opening a new front in technological competition between China and the United States. In the US, companies such as Physical Intelligence, Skild AI and Generalist are working on systems that enable robots to learn different skills and use them in new situations. Meanwhile, in China, companies such as X Square, Galaxea and Spirit AI are developing models that aim to combine visual perception, understanding instructions and physical action.

The US has an advantage in some areas of advanced AI research and access to powerful processors, while China relies on its extensive manufacturing capacity and direct cooperation with industrial factories.

One of the biggest challenges is training robots. Unlike chatbots, which learn primarily from vast amounts of text, robots need to understand physical interactions with the real world. They must learn how to pick up objects, use tools, manipulate different materials and adapt to changes in their environment.

To this end, companies are using videos of human activities, sensors worn on workers’ bodies and computer simulations. An interesting example described by Reuters shows a robot that could not put a pair of headphones in a cabinet that was out of its reach. After being shown videos of people solving similar problems, the robot tried another approach: it pushed a box next to the cabinet and attempted to climb onto it to reach the object.

This experiment illustrates the difference between a robot that repeats a programmed action and one that tries to find a solution on its own.

The technology is also beginning to be tested in real workplaces. According to Reuters, Spirit AI robots are being tested for connecting flexible cables during battery inspection processes at a CATL factory. Meanwhile, the Ant Group’s robotics unit, Robbyant, has said its robots process orders at two pharmacies in Shanghai, where human workers continue to check the medications before delivery.

However, these are still limited applications and do not demonstrate that robots can generally replace human labor.

Despite investment and technological advances, the industry faces significant challenges. A lack of high-quality data, the need for advanced processors and the difficulty of transferring skills learned in simulations to the real world are among the main obstacles. In addition, a robot that can perform a task in a controlled environment is not necessarily able to repeat it in another factory or under different conditions.

The race for intelligent robots is not just about technology. It is directly tied to the future of industrial production, logistics, services and the global labor market. If companies manage to create robots that learn new tasks and adapt to unfamiliar situations, automation could enter an entirely new phase.

Until now, success was measured by how well a robot could move. In the next phase, the key measure will be how well it can understand and solve problems. In this race, producing the fastest or strongest robot will not be enough. The real challenge will be creating the most intelligent robot.