In the first half of 2026, China shipped more than 40,000 humanoid robots, accounting for 97% of global supply. This is a number capable of changing the rules of competition: embodied AI—AI systems that enable robots to perceive their environment and perform physical tasks—is moving from the stage of demonstrating capabilities to the stage of delivering at scale, but it is still far from replacing existing systems at scale.
What This Is
Mass production means more than building prototypes. It requires a complete system spanning component supply, whole-machine integration, reliability validation, spare parts and maintenance. Zhishen Technology has cumulatively mass-produced more than 15,000 units, with deployments in power, petrochemical, inspection, firefighting and security. Its strategy has not been to pursue general-purpose robots, but to begin with highly structured, dangerous or hard-to-staff tasks and deliver through a combination of hardware, maintenance and data.
The Ministry of Industry and Information Technology has proposed creating the capacity for deployments at the 10,000-unit scale by the end of 2026 and building more than 100 high-value application scenarios. The more important question is how the ministry defines “high value”: in scenarios such as material handling and inspection, robots must be able to offset costs for equipment, maintenance and labor. The robot’s own purchase price is not the only barrier.
Industry View
We believe the 40,000-unit figure demonstrates that China has established a relatively complete humanoid robot supply chain, but it does not yet prove that the economics work. Compared with China’s installed base of tens of millions of industrial robots, 40,000 units remains a tiny scale. How many are actually entering production environments and operating continuously, rather than being delivered for a one-off project, still needs to be observed.
The risks are equally clear. MTBF (mean time between failures), recognition accuracy in non-ideal environments, BOM cost (bill of materials cost), spare-parts response times and system integration capabilities will all create gaps between companies. The data definitions also warrant caution: Zhishen Technology’s 15,000 units represent “cumulative mass production,” not shipments in a single year. Once demonstration videos become less common, the real screening metrics will be repurchase rates, failure rates and customers’ return on investment.
Impact on Regular People
For enterprise IT: procurement priorities will shift from asking what a robot can do to determining whether it can integrate with existing MES (manufacturing execution systems), data platforms and maintenance workflows. IT departments will become involved earlier in procurement and acceptance testing.
For individual careers: the near-term impact will be concentrated in job support and restructuring in areas such as inspection, firefighting and security. Dangerous, repetitive and hard-to-staff roles are more likely to adopt robots first; the pace of substitution will still depend on maintenance systems and returns on investment.
For the consumer market: home robots may remain at the demonstration and pilot stage, with cost, reliability and after-sales networks remaining the main constraints. Large-scale commercial adoption is more likely to occur first in B2B (enterprise) scenarios.