Summary of Key Points
This news article focuses on embodied intelligence (intelligent robots that can perceive their environment and perform actions like humans) and covers several key aspects:
1. The “ChatGPT moment” for embodied intelligence—when the public will directly experience a disruptive breakthrough—is still 2 to 10 years away. The critical point will be when robots can enter unfamiliar households and complete 80% of daily tasks on their own.
2. 2026 marks the beginning of the small-scale commercialization of embodied intelligence, with the Chinese market expected to exceed 1 trillion yuan in size. Humanoid robots are the fastest-growing segment.
3. The biggest bottleneck is the lack of data; there is a severe shortage of real-world scenario data, making training facilities a crucial infrastructure.
4. Training facilities are being built across the country, but they face issues such as homogenization and low data production capacity.
5. The Ministry of Industry and Information Technology has launched a special initiative to promote the implementation of embodied intelligence in certain scenarios by the end of 2026.
I. How Long Until the “ChatGPT Moment” for Embodied Intelligence?
The “ChatGPT moment” refers to the moment when people suddenly realize the power of AI, similar to when ChatGPT was launched. Wang Xingxing, the chairman of Yushu Technology, believes that this could happen in 2 to 3 years at the best, or 5 to 10 years at the worst. When will this moment arrive? It will be when robots can easily enter an unfamiliar household and complete about 80% of daily tasks on their own upon command (for example, “Fold the clothes in the living room and put them in the wardrobe”).
II. 2026: The Year of Small-Scale Commercialization of Embodied Intelligence
According to a white paper from the China Business Industry Research Institute, 2026 will be the first year for the small-scale commercialization of embodied intelligence. Competition in the industry will shift from focusing on concepts to actual product delivery. The Chinese market for embodied intelligence is expected to exceed 1 trillion yuan in 2026, with humanoid robots being the fastest-growing and most promising segment. The “broader market” includes robots, core components (such as joints and sensors), and control software.
III. The Biggest Barrier: Insufficient Data
For embodied intelligence to become intelligent, it needs to “learn from experience” in a human-like manner. However, it requires multimodal data, including visual information, motor skills (walking, picking up objects), and tactile sensations (such as the amount of force needed to hold a cup). The challenges are:
- High demand: Reaching ChatGPT-level intelligence requires 1.1 trillion tokens of data (the smallest unit of language, equivalent to a word). This amount could only be collected by 21,200 robots working continuously for a year.
- Current shortage: There are only a few hundred thousand to a million hours of high-quality real-world data available globally, which is far from enough.
- No shortcuts: While large language models can use data from the internet, embodied intelligence data has not been systematically recorded and cannot be obtained online; therefore, training facilities are essential for developing these robots.
IV. Training Facilities Are Being Built, but There Are Challenges
In the past two years, many regions have started building training facilities, with government support (more than 40 policy documents issued between 2024 and June 2026, and over 70 facilities are already in use or under planning). Zhejiang has the highest number of facilities (14), followed by Shandong and Jiangsu (8 each). These facilities are mainly aimed at industrial manufacturing (86%), home services, and healthcare. The most popular applications in industry are logistics and material handling (about 40%), while healthcare and emergency scenarios (30% and 25%, respectively) are more challenging.
However, there are issues with these facilities, such as homogenization of training scenarios, limited data diversity, and the lack of viable business models. Building facilities is just the first step; they must also be effectively utilized.
V. Government Action: Promoting the Implementation by the End of 2026
In June this year, the Ministry of Industry and Information Technology and the State-owned Assets Supervision and Administration Commission launched a special initiative to promote the practical application of embodied intelligence. The goal is for humanoid robots to be tested in representative scenarios (such as factories and hospitals) by the end of 2026 and to develop over 100 valuable use cases, leading to the commercialization of tens of thousands of robots. This initiative aims to accelerate the transition of embodied intelligence from the laboratory to real-world applications.
In summary, embodied intelligence is a major future trend, but it is still in the early stages of development. Issues related to data, use cases, and business models need to be addressed. 2026 is a critical year; whether a breakthrough will occur depends on how these challenges are overcome.