China’s Carmakers Race Into Humanoid Robots
China’s automotive and embodied intelligence sectors are converging at speed, blurring the line between the car and the humanoid robot. As of the end of August 2026, nearly 20 major global automakers had entered the embodied robot business through in-house development, incubation, or investment, with cumulative planned spending exceeding 100 billion yuan, according to Xinhua News. Chinese automakers are the driving force behind the wave.
The clearest signal came on September 9, when XPeng officially launched its robot production line in Guangzhou. The company said the world’s first high-order general-purpose humanoid robot completed automated final assembly and walked off the line on its own — a leap from R&D trial production to full line manufacturing.
From Four Wheels to None
The logic behind the shift is technological overlap. Former Deputy Director of the Policy and Regulation Department at China’s State Post Bureau, Jin Bing, said intelligent vehicles and embodied robots share more than 70% of their underlying technology, empowering each other in both directions. That commonality spans hardware — motors, sensors, and thermal management — as well as software algorithms, manufacturing processes, and quality management systems. Automakers’ accumulated mass-production capability and supply chains can be repurposed for robots without rebuilding from scratch.
“For the automotive industry, we have done the work on four wheels well; the next focus is to do well the work without wheels,” said Zhang Yongwei, chairman of the China EV100 think tank, at the Aggregated Intelligence Industry Development Conference (2026) in Wuhan. Zhang argued the convergence is a necessity for both sides: abandoning the embodied intelligence track would mean carmakers miss a major opportunity to extend their industrial capabilities, while the robotics industry would see its path to commercialization slowed and its development cycle greatly extended without deep integration with the automotive technology chain.
XPeng Group Chairman and CEO He Xiaopeng framed the production milestone as a first. “The robot production line was created from scratch, with no precedent to follow,” he said. “XPeng is creating a production line for a completely new product category.” The company plans to enter scaled mass production of its humanoid robots by the end of 2026, starting with commercial applications in its own stores and campuses, and targets an official market launch and delivery in China and overseas markets in 2027.
Capital and Policy Fuel the Race
The convergence is being amplified by two forces: policy and money. Embodied intelligence was listed as a future industry in China’s 15th Five-Year Plan outline and the 2026 Government Work Report, alongside quantum technology, brain-computer interfaces, and 6G. A State Council Development Research Center report projects the nation’s embodied intelligence market could reach 400 billion yuan by 2030 and exceed 1 trillion yuan by 2035.
Investors have responded. XPeng’s robot business closed its first equity financing round in August, raising over $900 million at a post-money valuation above $6.3 billion — a record for a single private equity round in China’s embodied intelligence industry, as Wallstreetcn reported. The round was led by IDG Capital, with Tencent and Alibaba joining as strategic investors. Elsewhere, Unitree’s valuation on the STAR Market far exceeds that of established EV makers NIO, XPeng, and Li Auto — a vivid illustration of the industry’s appeal.
The same pattern is playing out globally. Tesla discontinued its Model S and Model X, dismantled the production lines at its Fremont, California factory, and converted them into a dedicated Optimus humanoid robot production line, targeting 100,000 units by the end of 2026. CEO Elon Musk described the move as a strategic shift toward becoming a “physical world AI company.” For automakers facing the reality that “selling cars alone is no longer profitable,” embodied intelligence offers a compelling second growth curve.
Technical Barriers Remain
Despite the enthusiasm, the path to scale is riddled with obstacles. Core technical barriers cluster in four areas: motors, transmission, thermal management, and perception. Robot joint motors differ fundamentally in structure from automotive motors. Robot joints — especially harmonic reducers — pursue flexible deformation, high reduction ratios, and zero backlash, whereas automotive transmissions prioritize rigidity, high rotational speed, and unidirectional durability. Robots are far more dependent on high-precision, multi-dimensional force sensors than cars.
Cost, reliability, and lifespan pose even stiffer challenges. Robot dexterous hand joints have a lifespan of only one to three months, and high-load joint durability is insufficient for long-term stable operation. System integration is difficult — a humanoid robot carries more than 20 degrees of freedom and hundreds of sensors that must work together in a confined space. Core robot components have historically relied heavily on imports, leaving domestic substitution in its early stages.
“The two industries share deep technological roots, but their product application scenarios and performance requirements differ markedly, leaving cross-sector transformation with no shortage of challenges,” the Xinhua report noted.
Building an Aggregated Intelligence Ecosystem
Industry leaders argue the answer lies in deep, multi-dimensional collaboration rather than simple technology stacking. Zhang proposed a broad agenda: accelerate co-building of AI-based technology foundations and extend automotive large models to embodied intelligence; attack foundational hard tech in storage, computing, connectivity, and perception; and open up the materials-to-actuators mechanical chain so mature automotive-grade supply chain capabilities can migrate to robotics.
Others emphasized shared infrastructure. “Intelligent vehicles are currently the wheeled embodied intelligent agents with the largest commercialization scale and the most mature landing conditions — they are the mother industry and the best testing ground for scaled stress-testing,” said You Zheng, deputy general manager of Dongfeng Motor Group. Liu Hongqian, general manager of Puhua Basic Software, added that operating systems that grew up on intelligent vehicles are already positioned for reuse in embodied robots and low-altitude aircraft, so the next generation of industry operating systems need not start from zero.
What’s Next
With XPeng targeting scaled robot production by the end of 2026 and Tesla aiming for 100,000 Optimus units in the same window, the coming months will test whether automakers can translate manufacturing muscle into reliable, affordable humanoid robots. The 21st Century Business Herald noted that capital is now betting on the full chain — from core technology and mass production to global commercial deployment — rather than on single-point breakthroughs.
The open questions are stark. Can robot joints reach automotive-grade durability? Can domestic component substitution mature quickly enough to secure supply chains? And can the convergence deliver on its promise of a genuine second growth curve for an industry under margin pressure? The China Economic Net account of Zhang’s remarks captured the mood best: the move from four wheels to none is no longer a theoretical question — it is now industrial practice. Whether China’s carmakers can master the work without wheels will define the next chapter of the country’s manufacturing ambitions.