(Yicai) May 29 -- China is expected to enter a phase of large-scale commercialization of robots guided by vision- and language-based models starting next year, following steady progress in overcoming technological and application challenges, according to analysts.

Following the smart robotics industry’s entry into mass production this year, commercial applications are projected to gradually materialize in 2026, experts said at the latest Lujiazui Finance Salon.

Mass production reflects manufacturing capacity, but the next stage focuses on delivering commercial value, said Zhang Yiming, director of the Embodied Artificial Intelligence Research Institute at Guosheng Securities. The key gap between the two is generalization -- the ability of robots to adapt to complex, real-world scenarios, he added.

The Lujiazui Finance Salon is a regular discussion platform launched in March in Shanghai’s Pudong New Area. Each week, it brings together participants from the financial and business sectors to explore a specific topic, with a total of 40 sessions planned for this year.

The evolution of intelligent robots has progressed from large language models to visual-language models, with the ultimate goal being vision-language-action models -- foundation models capable of guiding robotic actions through visual and verbal commands. However, Zhang noted that the success rate of these models in laboratory environments still needs improvement.

"Both the brain and the body of humanoid robots remain immature and must continue to evolve," said Wang Jiegao, a partner at Shanghai-based SEARI Private Equity Investment Management. The PE professional added that from the standpoint of replacing human labor, current humanoid robots remain more costly and less efficient than traditional industrial robots or existing smart devices.

Zhao Yuntao, co-founder of Shanghai-based Yiyou Intelligent Control Technology -- which manufactures humanoid robot components -- highlighted ongoing challenges, particularly with robot torsos. Stability and endurance are still major obstacles, he said.

Looking ahead, Zhao from Yiyou emphasized that the research and iterative development of joint modules, along with next-generation joint technologies, will be critical to advancing humanoid robotics.

Editors: Dou Shicong, Emmi Laine