Zhejiang Finds a Path for Humanoid Robots from Showcases to Factory Floors
On September 9, 2026, the Hangzhou International AI and Robotics Innovation Expo (AI Show 2026) opened, with over 500 domestic and foreign brands participating, and humanoid robots took center stage. Data released at the same time showed that Zhejiang's service robot output surged 1.2-fold in the first half of the year, while core AI industry revenue grew 25.6%.
While "embodied intelligence" remains on planning documents in many places, Zhejiang is leveraging its manufacturing base and open-scene strategy to turn humanoid robots from "showcase attractions" into "factory workers."
In Ningbo, Zhejiang, a batch of embodied intelligent robots is undergoing final tests at Juneng Intelligent's workshop before heading to European production lines of leading automakers. This is not a demonstration but an order delivery. As labor costs in Europe keep rising, Chinese robots are becoming a practical choice for manufacturers to reduce costs and improve efficiency.
A more typical case is at Geely's automobile plant, where Unitree's humanoid robots perform the full automated process of grasping, transporting, and installing wheel hub trim parts. At the Zhejiang Humanoid Robot Innovation Center in Ningbo, 2,000 customized robots for garment scenarios are entering mass delivery, covering scenarios such as logistics sorting and consumer electronics, with a success rate of 99% in specific test scenarios.
Supporting these advancements is sustained investment in underlying technologies. Zhejiang has set up a major special project for humanoid robots with a total investment of 840 million yuan, focusing on core technologies like general intelligent models, high power-density drive-control, and multimodal perception. The provincial science department has launched six provincial projects dedicated to embodied intelligence, driving total funds of nearly 180 million yuan.
The Zhejiang Humanoid Robot Innovation Center is based in Ningbo rather than Hangzhou. Chief scientist Xiong Rong explained the deep logic behind Zhejiang's model: Ningbo has been guiding traditional enterprises toward the humanoid robot track through supportive policies since 2023, and it has a complete automotive and robotics industry chain, with both upstream supply and downstream application markets. Haishu District provides diverse trial scenarios, including industrial and agricultural sites, for robot algorithm training, forming a combined strategy of "industry + platform + scenario + policy."
At the provincial level, policy design also points in this direction. Government documents explicitly promote the "replacement with application" mechanism for small and medium enterprises purchasing domestic embodied intelligent equipment for scenarios like intelligent manufacturing, warehouse logistics, and eldercare services, offering a subsidy of 15% of the equipment price, up to 2,000 yuan per unit, and inclusion in government procurement catalogs.
Meanwhile, Ningbo and Hangzhou are building pilot bases and training grounds for humanoid robots, encouraging enterprises to open technical verification scenarios.
As hardware bodies mature, the "brains" still need to be "fed" with high-quality data. Zhejiang has established a provincial data group, working with six national AI application pilot bases to supply high-quality data "fuel" for embodied intelligence. Ningbo has opened up 106 manufacturing scenarios, turning production lines at Geely, Youngor, and Zhenhai Refining & Chemical into AI practice grounds, letting technologies directly confront real issues like yield improvement and production scheduling.
From the "real-world test room" at the 2026 World Robot Conference to the "real-world demonstration" at AI Show, Zhejiang is exploring a path different from most regions. Instead of pursuing flashy single-point technologies, it opens manufacturing scenarios, strengthens the data foundation, and connects the industrial chain to let robots iterate on real production lines.
Whether this path ultimately works will determine whether China's embodied intelligence industry can evolve from "being able to build robots" to "being able to use robots well."