Humanoid Robot OEMs Struggle as Dexterous Hand Profits Shift Upstream
The finances of humanoid robot OEMs are out of sync with their funding headlines. A humanoid unit priced at 85,000 yuan requires the brand to handle design, assembly, algorithm development, and after-sales service, leaving hardware gross margins around 10%.
Unitree Technology stands out with a 60.13% gross margin in 2025, but its prospectus lists falling margins amid rapid revenue growth as a risk due to price wars. OEMs are cutting prices and expanding capacity, yet most remain unprofitable.
Dexterous hands are falling in price much faster than complete humanoid robots. They account for 10% to 15% of a humanoid robot's total cost, not the largest share among components, but their price decline is worth examining separately.
The price of a single hand has dropped from around 100,000 yuan in the early days to under 10,000 yuan, a decline of nearly an order of magnitude. This pace is far faster than that of complete robots, indicating that upstream scale effects are first appearing in dexterous hands.
Complete robot price cuts are constrained by brand premiums, channel costs, and after-sales systems. Dexterous hands are more standardized components, so once production capacity rises, costs are diluted faster.
Market growth data also illustrates the trend. In 2025, China's dexterous hand sales were about 19,200 units, up 236.84% year-on-year. Sales are expected to reach 70,200 units in 2026, up 265.63%, and could exceed 430,000 units by 2030.
That growth rate is an order of magnitude higher than the growth rate of complete robot shipments. Complete robot shipments are climbing from the thousand-unit level to the ten-thousand-unit level, while dexterous hands are climbing from the ten-thousand level to the hundred-thousand level. The slopes are different.
Upstream components indeed have higher gross margins than complete robots. Bearings used in humanoid robots require more than two months to process 1,000 sets, with a gross margin of about 30%. Ordinary bearings require only one month to process 100,000 sets, with a gross margin of only 10% to 20%.
Processing difficulty is inversely related to batch size. Humanoid robot components are produced in small batches with high precision requirements, so the per-unit profit space is larger. The logic for dexterous hands is similar.
Dexterous hands involve micro motors, reducers, sensors, and tendon-driven transmission. Each link demands higher processing precision than other parts of the complete robot, while supply volumes are far smaller than consumer-electronics-level batches. Pricing power therefore rests with suppliers.
This structure is not favorable for complete robot manufacturers. A considerable portion of the money an OEM earns from selling one robot must be shared with upstream suppliers. Suppliers of core components such as dexterous hands, joint modules, and force sensors already obtain stable gross margins while OEMs are still losing money or earning thin profits.
The OEM's role is more like that of a system integrator and brand owner, while the truly profitable link is upstream. The fiercer the competition in embodied intelligence complete robots and the more intense the price war, the stronger the bargaining power of upstream components becomes.
The next question is whether OEMs will develop dexterous hands in-house to take back this profit. If they develop in-house, short-term costs will rise, but long-term gross margins may improve. If they continue to outsource, profits will remain upstream. This choice will determine the future profit distribution between OEMs and component suppliers.