China’s humanoid robotics sector didn’t emerge from a garage startup culture funded by venture capital. It was built on a template Beijing perfected over two decades: identify a strategic technology, flood it with state money, and let local governments compete to build the industrial base. That template first produced the world’s dominant electric vehicle industry. Now it is being redirected, almost in real time, toward humanoid robots, and the same companies, the same subsidy playbooks, and increasingly the same physical infrastructure are doing double duty across both sectors.
Understanding China’s robotics rise means understanding three actors moving in lockstep: the state, the EV industry, and the robot makers. A fourth element, training data, has emerged as the glue holding the triangle together.
From EV Subsidies to Robot Subsidies
Beijing’s approach to electric vehicles in the 2000s set the pattern still in use today. After the central government made growing the EV industry a national priority, provinces and cities competed with subsidies, tax breaks, and cheap loans, while government agencies and state-owned bus fleets became early customers before the vehicles were fully market-ready.
That same machinery is now shifting gears. As EV subsidies wind down and automaker profit margins narrow, Beijing’s policy support is moving toward embodied AI and humanoid robotics, prompting automakers to redeploy technology built for smart EVs directly into robot programs. Analysts note the pattern is already producing familiar side effects: companies with no robotics background are pivoting into the space purely to capture subsidies, echoing exactly what happened during the EV boom.
The scale of the new state commitment is substantial. China’s 15th Five-Year Plan, covering 2026 through 2030, is expected to allocate roughly $300 billion in subsidies for robotics and AI. More than $20 billion has already gone to the humanoid sector in the past year alone, and Beijing is establishing a fund worth about one trillion yuan, or roughly $137 billion, for AI and robotics startups. State procurement of humanoid robots and related technology grew sharply too, rising from 4.7 million yuan in 2023 to 214 million yuan in 2024, according to a Reuters review of tender documents. City and provincial governments have layered on their own funds and incentives as well. Shenzhen created a 10 billion yuan AI and robotics fund. Wuhan offers robot makers and component suppliers subsidies of up to 5 million yuan plus free office space once they hit sales targets. Beijing’s municipal government has offered up to 30 million yuan to help companies accelerate their first product builds.
National policy architecture backs this spending. The “Robot+” initiative and the “AI + Manufacturing” roadmap aim to build humanoid pilot production lines and double China’s manufacturing robot density by 2030. The Ministry of Industry and Information Technology has stood up a dedicated Standardization Committee for Humanoid Robots, while cities such as Shanghai run their own regional action plans targeting breakthroughs in embodied intelligence.
Where EV Makers Fit In
China’s EV sector isn’t a bystander to this shift; it is a primary vehicle for it. As the EV price war settles into a smaller set of dominant players, automakers are expanding into humanoid robotics partly out of financial necessity, but also because they hold real structural advantages: mature supply chains for batteries, sensors, actuators, and precision manufacturing, plus years of engineering experience that transfers directly to bipedal robots. As one industry observer put it, if you understand EVs, meaning sensors, chips, batteries, charging, and manufacturing, the step to building a robot is comparatively small.
That overlap shows up concretely. Nio, the EV maker best known for its battery swap network, has partnered with robot manufacturer UBTech while simultaneously building an in house humanoid R&D team. More broadly, China’s EV sector is backing a large share of the country’s sixty plus homegrown humanoid manufacturers, out of more than 150 nationwide, leaning on the same component suppliers for batteries, sensors, and actuators that already serve the auto industry. In the Yangtze River Delta, this overlap has produced one of the most vertically integrated humanoid supply chains in the world, with companies like Unitree manufacturing motors, reducers, and sensors in house, and EV linked component suppliers located within a two hour logistics radius.
The result is that China’s top humanoid makers, led by Unitree and AgiBot in 2025 shipments and followed by UBTech, Leju Robotics, Engine AI, and Fourier, have inherited much of their production capacity, cost structure, and speed to market directly from the EV manufacturing base that came before them. Chinese firms shipped an estimated 80 percent of the world’s humanoid robots in 2025, and that share is projected to grow further in 2026.
Capital Markets Are Catching Up
Investment and valuations tell a more mixed story than subsidy volume alone suggests. Humanoid startups such as Galbot and AgiBot, both founded in 2023, have together raised close to $2 billion in venture funding. Unitree, Galbot, and AgiBot have each individually crossed $1 billion in valuation, and Unitree secured regulatory approval in mid-2026 for an IPO expected to raise roughly 4.2 billion yuan, or about $619 million, valuing the company near $7 billion.
That figure is notably conservative next to comparable companies in the United States. Figure AI, for instance, raised a $1 billion Series C at a $39 billion valuation without having shipped a commercial product, while Unitree’s IPO values it well below that despite having shipped 5,500 units in 2025. The gap illustrates a structural difference between the two ecosystems. Silicon Valley’s robotics boom is largely venture capital driven, betting on future breakthroughs, while China’s is turbocharged by industrial policy and state demand, betting on scale and deployment speed instead.
Training Data: The Fourth Corner of the Triangle
Hardware has advanced quickly, but robot “cerebellums,” the motion control and manipulation skills that let a robot function reliably in the real world, remain the industry’s weak point. Unlike large language models, humanoid robots cannot be trained primarily on scraped internet text; they need embodied, physical demonstration data covering vision, touch, force, and motion, and that data is scarce almost everywhere.
Beijing has responded by treating data collection as national infrastructure, much the way it once treated EV charging networks. The 15th Five-Year Plan explicitly incorporates the systematic development of training environments for embodied AI, and the results are already visible on the ground. More than forty state backed robot data collection centers had been announced by the end of 2025, with roughly two dozen already operational. Cities including Beijing, Zigong, Liuzhou, Jiujiang, Wuxi, Wuhan, Shaoxing, and Zhengzhou have each stood up their own embodied intelligence data collection centers.
A flagship facility in Beijing’s Shijingshan district, built by the local government in partnership with humanoid maker Leju, spans more than 10,000 square meters and covers 16 distinct training scenarios, from mock car assembly lines to household tasks. A newer center in Shanghai’s Zhangjiang district covers roughly 5,000 square meters and will host more than 100 robot models from over a dozen companies. Output at these facilities is already substantial. One Shandong center generates around 6 million data points annually, the highest in the country, with trained robots reaching a 95 percent task success rate across more than 20 functions. A Hubei center runs about 100 humanoid robots through repetitive folding, wiping, and sorting tasks purely to harvest movement data.
Much of this data collection is still surprisingly manual. At these centers, human trainers wearing VR headsets guide robots through the same pick and place motions over and over via teleoperation, a process one industry figure described as essentially teaching robots to think for themselves, one repetition at a time. This has effectively created a new labor category and a new commercial service layer, as data collection, cleaning, annotation, model training, and validation evolve from in house R&D functions into a specialized value chain in their own right, linking AI development directly to the physical economy.
Private companies are folded directly into this state built infrastructure rather than building parallel systems of their own. PaXini Tech, for example, partners directly with the government to operate large scale data collection factories in Shanghai, Tianjin, and Fujian, producing open source datasets that support model development across the industry, infrastructure no single company could easily fund alone.
How the Triangle Closes
Put together, the financial relationship between the three players resembles a closed loop rather than a simple funding pipeline. The state subsidizes hardware development, funds physical data collection infrastructure, and acts as an early, guaranteed buyer through public procurement, exactly as it did with EVs. EV makers supply manufacturing scale, component supply chains, and engineering talent, repurposing infrastructure built for cars into infrastructure for robots, often under financial pressure from a saturated EV market. Robot makers, increasingly EV adjacent themselves, consume subsidized capital and government built data infrastructure to accelerate development, in exchange for operating training centers and generating the datasets the wider ecosystem depends on.
Each better robot justifies further state investment. Each new data center lowers the cost of entry for the next wave of companies. Each EV maker’s pivot into robotics deepens the supply chain overlap between the two industries. It is, in effect, the EV playbook running a second time, with the same risks analysts flagged the first time around. China’s industrial policy model has a documented history of producing overcapacity, wasted resources, and brutal price wars once a sector matures, and some analysts already expect humanoid robot costs to follow the EV trajectory, falling from roughly $35,000 in bill of materials cost today to as little as $17,000 by 2030 if the supply chain remains China centric.
