China Shipped 82% of All Humanoid Robots in 2025. Here is Why.

In 2025, the global humanoid robot industry shipped approximately 12,800 units. Of those, roughly 10,500 came from Chinese manufacturers. That is 82% of the global total, and the number surprised even the analysts who had been tracking China’s robotics push for years.

This is not a story about one lucky company. It is a story about industrial policy, manufacturing ecosystems, and a fundamentally different approach to bringing new technology to market.

The MIIT blueprint

China’s dominance did not happen by accident. It was planned.

In November 2023, China’s Ministry of Industry and Information Technology (MIIT) published its “Humanoid Robot Innovation and Development Guidelines,” a policy document that laid out a national strategy for becoming the world leader in humanoid robotics by 2027. The guidelines called for breakthroughs in key components like actuators, sensors, and AI chips, and set explicit production targets for Chinese manufacturers.

The document was not just aspirational. It came with money. Provincial and municipal governments across China began offering subsidies, tax incentives, and cheap land to humanoid robot startups. Shanghai alone allocated over 10 billion yuan (roughly $1.4 billion) to its robotics industrial zone. Shenzhen, Beijing, and Hangzhou followed with their own incentive packages.

By contrast, the United States had no coordinated federal policy on humanoid robotics in 2025. American companies relied on venture capital and, in Tesla’s case, internal funding from other business lines. The CHIPS Act and Inflation Reduction Act included provisions for advanced manufacturing, but nothing specifically targeting humanoid robots.

The companies driving the numbers

China’s humanoid shipment numbers are not evenly distributed. A handful of companies account for the vast majority.

AgiBot: from zero to market leader in 18 months

The biggest surprise in 2025 was AgiBot, a Shanghai-based company that went from near-obscurity to the world’s top humanoid robot shipper in under two years.

AgiBot’s secret was not a technological breakthrough. It was a manufacturing breakthrough. The company built a vertically integrated factory in Shanghai’s Lingang Special Area that could produce humanoid robots at a pace and cost that no Western competitor could match. Their Genie series robots, while not as technically sophisticated as Figure’s 02 or Tesla’s Optimus, were good enough for the logistics and light manufacturing tasks that made up the bulk of early demand.

AgiBot also benefited from something uniquely Chinese: a massive domestic market of companies willing to pilot humanoid robots even before the technology was fully mature. Chinese manufacturers, warehouse operators, and hospitality chains deployed AgiBot units in real-world settings, providing the company with invaluable field data to improve their products.

Unitree: the price disruptor

Unitree, based in Hangzhou, took a different approach. Rather than targeting enterprise deployments, they positioned their G1 as an accessible platform for education, research, and early commercial use. Their starting price of roughly $16,000 made them the “gateway drug” of humanoid robotics, and it worked. Over 3,200 G1 units shipped in 2025, finding homes in universities, research labs, and pilot programs worldwide.

Unitree’s strategy was deliberately modeled on DJI’s early approach to drones: sell affordable hardware, build a developer ecosystem, and iterate faster than anyone else. The comparison is not perfect, but the playbook is recognizable.

The cost advantage

Price is where China’s advantage is most visible and most durable.

The cost difference is not just about labor. Chinese humanoid robot manufacturers benefit from a mature supply chain for actuators, motors, batteries, and sensors that was originally built to serve the country’s electric vehicle and consumer electronics industries. The same Nanjing factory that makes servo motors for EV steering systems can supply actuators for humanoid robot joints at a fraction of the cost of sourcing equivalent components in the US or Europe.

China also leads in battery technology, which is a critical bottleneck for humanoid robots. CATL, BYD, and other Chinese battery makers produce the lithium-iron-phosphate cells that power most Chinese humanoid robots at 30-40% lower cost than equivalent cells from Samsung SDI or LG Energy.

The timeline of China’s robotics push

What this means for the US and the rest of the world

The 82% number should be a wake-up call, but it needs context.

China’s lead is in volume, not in all aspects of technology. American companies like Figure AI, Tesla, and Apptronik are arguably ahead in AI capabilities, particularly in foundation models for manipulation and task learning. The gap is that these AI advantages have not yet translated into shipped products at scale.

There are also real questions about the quality and reliability of Chinese humanoid robots. Many of the 10,500 units shipped in 2025 were deployed in controlled environments with limited task requirements. Performing repetitive pick-and-place operations in a warehouse is very different from the kind of flexible, multi-task autonomy that Figure AI is targeting.

The data flywheel problem

This is the part that should worry American competitors the most. Every deployed humanoid robot generates data: about locomotion, manipulation, navigation, and human interaction. That data feeds back into training better AI models. More deployed robots means more data, which means better AI, which means more deployable robots.

China has over 10,000 humanoid robots operating in real-world environments, generating real-world training data every day. American companies have maybe 1,500. If the data flywheel thesis is correct, and most AI researchers believe it is, then China’s volume lead will eventually translate into an AI lead as well.

This is the same dynamic that gave Chinese companies an edge in areas like facial recognition and autonomous driving in Chinese cities. Deployment at scale, even with imperfect technology, generates the data needed to make the technology better.

What happens next

The MIIT’s 2026 target is 20,000 domestic humanoid robot shipments, nearly double the 2025 number. Based on production capacity announcements from AgiBot, Unitree, UBTECH, and Fourier Intelligence, that target looks achievable.

For the United States, the window to respond is narrowing but not yet closed. The Biden administration’s executive orders on AI included some provisions for robotics R&D funding. Several bills in Congress would create tax incentives for domestic humanoid robot manufacturing. But nothing has passed yet, and the gap keeps growing.

The humanoid robot race is not over. The technology is still immature, the market is still tiny, and the winning business model is far from settled. But the manufacturing race has a clear leader, and pretending otherwise does nobody any favors. China shipped 82% of all humanoid robots in 2025 because it decided this technology was strategically important and acted on that decision with money, policy, and urgency.

Whether the rest of the world responds with the same level of seriousness will determine the shape of this industry for decades to come.