36Kr reports on physical AI debate surrounding Zidong Taichu

36Kr reported on September 18, 2026, on issues in physical AI surrounding Zidong Taichu.

In physical AI, the question is not only how capable a robot is, but also how close the foundations for spatial understanding and decision-making are to practical use. 36Kr’s public coverage of Zidong Taichu shows that these issues are becoming visible in China’s public information space.

Rather than covering Zidong Taichu as the move of an individual company, 36Kr’s report is more focused on explaining the place of “spatially embodied AI” in China. What can newly be confirmed here, then, is less an event itself than that issues surrounding physical AI have reached the point of being discussed publicly. Among related articles in this newspaper, Kawasaki Heavy Industries’ announcement of Plesio focused on the usability of robots that work alongside people. By contrast, 36Kr’s report appears to place more emphasis on foundational issues, including spatial recognition and decision-making, than on the functions of robots themselves. In other words, how to connect robot hardware design with the underlying foundations of perception and learning may be a point of divergence. From an industry-structure perspective, these issues matter not only to robot makers but also to companies whose work depends on developing visual perception, spatial understanding and decision-making foundations. The fact that names such as NVIDIA, Google, World Labs and Zhongke Zidong Taichu appear together suggests that computing infrastructure, models and spatial understanding are being discussed as connected layers. However, this excerpt alone does not establish how much detail the 36Kr article provides on specific technical approaches or implementations. Readers should consult the original article for details. The next questions to examine are what scope of spatial perception Zidong Taichu handles, how it differs from existing models, and whether there are examples showing a connection to physical robots. The excerpt provides only an outline of the issues and does not indicate anything about mass production or customers.

In Japan, an issue will be how to incorporate foundations for spatial perception and decision-making alongside robot control. If these foundations are becoming a more prominent topic in China, Japanese companies may need to consider how to connect with perception and learning systems, in addition to relying on existing strengths such as joint components.