China leads humanoid volume
Industry analysis suggests China accounted for roughly 87% of global humanoid shipments in 2025, shifting the competition from demos to manufacturing and supplier depth. Reporting argues Chinese scale plus state support has built out actuator, sensor and assembly capacity while Taiwanese firms are racing to differentiate on tactile sensors for real‑world interaction. The practical upshot is that component‑level engineering — sensors, reducers, battery and actuator supply chains — is becoming a strategic bottleneck for commercial humanoids. ( )
A humanoid robot is basically a stack of parts pretending to be a worker: motors move the joints, gears multiply force, batteries feed power, and sensors tell the machine what it just touched. In 2025, the country that shipped the most of those stacks was China, with 13,317 global humanoid units shipped and Chinese manufacturers taking about 87% of that volume. (ibtimes.com.au) That number changes the story from “who has the best demo video” to “who can actually build thousands of machines.” Industry reporting now describes humanoids as moving onto factory floors, into trailer unloading, and into pilot home tasks like folding laundry. (ibtimes.com.au) China did not get there by accident. A Ministry of Industry and Information Technology guideline set a goal for a preliminary humanoid innovation system by 2025 and a secure, reliable industrial and supply chain system by 2027. (english.www.gov.cn) Local governments pushed the same way. Beijing’s 2023-2025 robot action plan explicitly targeted self-development across the supply chain in key technology areas instead of just assembling final machines from imported parts. (english.beijing.gov.cn) The bottleneck is not the metal body shell. The bottleneck is the “inside the joints” hardware, because one humanoid can need more than 20 reducers, and a harmonic reducer is the precision gearbox that lets a motor move like a wrist instead of a jackhammer. (wuhan.gov.cn) Batteries are another hard limit. TrendForce said in January 2026 that most humanoid robots still run only 2 to 4 hours, so longer shifts depend on hot-swappable packs or higher-energy-density batteries such as solid-state designs. (trendforce.com) Even high-profile builders are running into that wall. TrendForce reported in July 2025 that Tesla’s Optimus program faced bottlenecks in battery life, hardware-software integration, joint motors, and transmission systems. (trendforce.com) That is why Taiwan is trying to win a different layer of the stack. DigiTimes reported on April 8, 2026 that Taiwanese firms are racing into tactile sensors, which are the pressure-and-contact sensors that let a robot tell the difference between grabbing a bolt and crushing a paper cup. (digitimes.com) Tactile sensing is becoming valuable because warehouses are structured and repetitive, while service work is messy and full of surprises. DigiTimes tied that push to labor shortages in service sectors and to the need for robots that can handle unstructured environments instead of fixed production lines. (digitimes.com) Research firms are now framing the market the same way suppliers do. TrendForce’s latest humanoid research highlights batteries and tactile sensing as core challenges, while DigiTimes Research says hardware cost will keep adoption narrow for the next 3 to 5 years. (trendforce.com, digitimes.com) So the race is no longer just about whose robot can walk onstage without falling over. It is about who controls the boring parts underneath: reducers, actuators, sensors, battery packs, and the factories that can turn those parts into thousands of identical machines on schedule. (ibtimes.com.au, english.www.gov.cn, digitimes.com)
