Walk the floor at WRC 2026 in Beijing — 300+ exhibitors, 2,000+ robots on display — and you'd think China's robot industry lives in its megacities and elite labs. It doesn't. Many of the robots being shown off were built in county towns: Pudu Robotics' largest factory sits in Jianhu County, a 600,000-resident county in northern Jiangsu, and Luerobot moved its headquarters from Beijing to Zoucheng, Shandong three years ago — then completed its Hong Kong IPO from there.
The show floor tells the glamorous story. The counties tell the real one.
Where the robots are actually made
The geography is striking once you map it:
- Xinchang, Zhejiang — known as the "hometown of Chinese bearings," with 600+ bearing manufacturers. Around 50 robot-related companies operate here, including six listed firms building robots. Wuzhou Xinchun passed Tesla's Optimus factory audit in late 2025, becoming one of seven core suppliers, and is the exclusive bearing supplier for Ant Group's Lingbo humanoid. Sanhua, another local giant, has made electromechanical actuators for bionic robots a strategic new business. The county targets RMB 10 billion in robot output by 2030.
- Beijiao, Foshan — one in every 20 domestic industrial robots is born in this town, home to 900+ manufacturers and Midea's HQ. Kepler Robotics opened its national production center here in August 2026.
- Zoucheng, Shandong — 26 robot companies, and one robot body can be assembled in 37 minutes. A "15-minute supply chain circle" has grown around Luerobot, with roughly 80% of local park enterprises feeding it.
- Jianhu, Jiangsu — 30,000+ machine tools and 2,000+ machining firms from decades of petroleum-equipment manufacturing, now hosting Pudu's largest plant plus storage-robot, service-robot and AGV projects.
- Yuhuan, Zhejiang — 4,000+ auto-parts firms and 1,300+ valve firms converted their precision-machining know-how into 182 robot-related companies, with 70% component localization.
Why counties win: the cost structure
This isn't accidental. A humanoid robot's cost is concentrated in actuators, lead screws and reducers — the "three mountains" that account for roughly 70% of the bill and are the hardest parts to make. That is precisely the expertise county clusters accumulated over decades in bearings, auto parts, home appliances and oil equipment.
The numbers bear this out. Morgan Stanley estimates Chinese companies hold 63% of the share in the global top-100 humanoid supply chain, and China's manufacturing base can cut humanoid production costs by two-thirds. In 2025, Chinese firms launched 28 humanoid models — nearly three times the US count.
The framework: brain vs. body
The AI story has been a brain story: model labs in Silicon Valley, frontier research, compute. The robot story is a body story. The US holds the brain advantage; China holds the body advantage — and the counties are to robotics what Shenzhen was to smartphones: a vertically integrated hardware ecosystem that compresses cost and time-to-iteration simultaneously.
The bottleneck of embodied AI is quietly moving from the demo stage to the delivery stage. Whoever commands the supply chain commands the unit economics — and unit economics, not demos, decide who scales.
What happens next
- Policy is an accelerant. Counties aren't passive inheritors: Xinchang runs a dedicated task force with monthly progress meetings; Yuhuan built a 2,000-mu robot core zone backed by a RMB 6.77 billion fund matrix; Jianhu is building a 2,500-mu robot park. Local governments are chasing robotics as the new engine to offset shrinking traditional industries.
- Commoditization cuts both ways. The same clusters that compress costs will compress margins. As scale grows, differentiation must migrate upward — into software, embodiment and data loops.
- Western players face a two-thirds cost gap. Design-for-China-supply-chain decisions, or partnerships with these clusters, are becoming the pragmatic path versus full in-house manufacturing.
What to watch
- For investors: upstream component makers (bearings, reducers, actuators) are the tell — supply-chain concentration precedes OEM success.
- For robotics startups: locate next to the manufacturing clusters; a 15-minute supply chain beats a remote lab for iteration speed.
- For everyone tracking embodied AI: stop measuring the industry by demo videos and count factories, yields and component lead times instead.
The robots on stage in Beijing were the headline. The counties behind them are the story that determines who delivers — and who just demos.
