Humanoid robots in China accounted for an estimated 95% of global shipments in 2025. That manufacturing lead is extraordinary—but a shipment is not the same as a productive worker.
A Reuters investigation published on 27 August 2026 found that the humanoid industry shipped about 20,000 units worldwide in 2025, with Chinese manufacturers responsible for 95%, citing BofA Global Research. China expects to build more than 100,000 humanoids in 2026, according to an industry-ministry official quoted by Reuters. (Reuters investigation: China’s humanoid boom and capability gap.)
Those figures show manufacturing momentum, not settled commercial demand. Reuters found a market strongly supported by government purchases, subsidies and demonstration projects while robots remained slow, error-prone and easily confused by changes to familiar tasks. (Reuters investigation: China’s humanoid boom and capability gap; International Federation of Robotics: China’s robotics strategy.)
How China built its humanoid robot manufacturing lead
China already operates roughly two million conventional industrial robots and installed 54% of the industrial robots deployed worldwide in the most recent International Federation of Robotics data. That creates a dense supplier base for motors, gearboxes, batteries, sensors, controllers and manufacturing equipment—many of the same building blocks required by humanoids. (International Federation of Robotics: China’s robotics strategy.)
Robotics also sits near the centre of China’s 15th Five-Year Plan for 2026–2030. National and local governments are funding factories, training centres and real-world data collection. Reuters counted at least US$230 million in fulfilled government purchases of humanoids and related systems during the first half of 2026, up sharply from prior years. (Reuters investigation: China’s humanoid boom and capability gap; International Federation of Robotics: China’s robotics strategy.)
Why shipment estimates need caution
Humanoid-market numbers vary because researchers count different products. Some include small research platforms, education robots or partially humanoid systems; others count only full-size general-purpose machines. Orders, production and delivered units are also different measures. A precise percentage can therefore conceal a less precise market definition. (Reuters investigation: China’s humanoid boom and capability gap; International Federation of Robotics: humanoids—vision and reality.)
The reliable conclusion is directionally clear: China dominates current hardware volume and supply chains. The less certain claim is that this volume proves a self-sustaining market. Many early buyers are governments, state-linked organisations, research centres or companies running pilots rather than ordinary businesses making repeat purchases from operating savings. (Reuters investigation: China’s humanoid boom and capability gap.)
Why humanoid robots in China still struggle with change
Modern humanoids can walk, dance, sort parcels and perform rehearsed factory steps. They struggle when an object, camera angle, work surface or lighting condition changes. Reuters observed robots completing simple training-centre tasks at about 20% of a human’s speed or output, according to one staff member. (Reuters investigation: China’s humanoid boom and capability gap.)
The control systems behind general-purpose robots are often called vision-language-action models. They translate images and instructions into movement, but physical training data are expensive to collect. Language models can learn from enormous amounts of internet text; robots need examples tied to real bodies, forces, objects and consequences. (Reuters investigation: China’s humanoid boom and capability gap; Reuters: Chinese robot makers seek useful work.)
Why conventional robots still win many factory jobs
A fixed industrial arm is designed around a narrow task. It can be bolted in place, guarded, programmed and optimised for speed and millimetre-level repeatability. A humanoid carries the extra complexity of legs, balance and general-purpose joints even when a job does not require them. (International Federation of Robotics: China’s robotics strategy; Reuters: Chinese robot makers seek useful work.)
Humanoids become more compelling where the workplace cannot easily be rebuilt: moving between stations, climbing stairs, using human tools or handling changing assignments. In the near term, the strongest opportunities are likely to be dangerous, repetitive and reasonably structured—not an all-purpose replacement for people. (Reuters investigation: China’s humanoid boom and capability gap; Reuters: Chinese robot makers seek useful work.)
A bubble can still produce lasting technology
China has more than 150 humanoid-robot companies, according to its top economic-planning agency. Executives and investors interviewed by Reuters expect consolidation as subsidies tighten and companies fail to find profitable deployments. Falling component prices may pressure manufacturers while also making robots more accessible. (Reuters investigation: China’s humanoid boom and capability gap.)
Overinvestment does not guarantee failure of the technology. China used intense competition and manufacturing scale to reduce costs in solar panels and electric vehicles. Humanoids are riskier because their usefulness is less mature, but the resulting supply chains and training infrastructure could survive even if many brands do not. (Reuters investigation: China’s humanoid boom and capability gap; International Federation of Robotics: China’s robotics strategy.)
What would prove real commercial value?
Watch for deployments that disclose uptime, task completion, human interventions, safety events and total cost. A robot should save more labour or reduce more risk than it adds through supervision, maintenance and downtime. Repeat orders from private customers would be stronger evidence than headline production targets. (Reuters investigation: China’s humanoid boom and capability gap; Reuters: Chinese robot makers seek useful work.)
China has already won the first race: producing humanoid hardware at scale. The next race is much harder—building machines intelligent and dependable enough that customers buy them because they work, not because a programme subsidises the purchase. (Reuters investigation: China’s humanoid boom and capability gap; International Federation of Robotics: China’s robotics strategy.)
Reporting note: This article separates demonstrated results from company forecasts and staged demonstrations. It is general information, not purchasing, employment or investment advice.
Explore more evidence-led coverage in Robotics & Humanoids.
For another perspective on robot manufacturing, read why Hyundai, Tesla and XPENG are building humanoid robots and how their automotive expertise shapes their plans.
Leave a Reply