Bloomberg’s analysis argues that humanoid robots are emerging as the next category where China converts manufacturing depth and state-backed supply chains into global export dominance, with companies like Unitree, Fourier Intelligence, and UBTECH already undercutting Western rivals on price while deploying in domestic factories at volumes that fund the iteration cycles needed to close capability gaps.
The EV parallel is not rhetorical. China spent a decade absorbing losses in electric vehicles, subsidising domestic demand, building integrated battery supply chains through CATL and BYD, and iterating manufacturing quality at a pace that Western incumbents could not match. By the time European and American automakers recognised the threat, the cost curve had already moved. Chinese EVs now sell globally at price points that traditional manufacturers cannot match without state support of their own. The pattern in humanoid robots is following the same trajectory, only faster, because China’s robotics companies can source actuators, harmonic drives, force sensors, and battery packs from the same Shenzhen and Yangtze Delta supply clusters that already supply consumer electronics and EVs.
Unitree’s G1, priced at $16,000, is the most cited example of production economics overtaking lab prestige. Figure, Boston Dynamics, and Apptronik build more capable platforms, but none approach that price point. Unitree’s ability to reach $16,000 per unit for a bipedal robot reflects the same logic as BYD’s ability to build an EV for $10,000: vertical integration, high domestic volumes, and a supply chain ecosystem where component suppliers sit within hours of the assembly line. The same supply chain advantage applies to UBTECH, which has deployed humanoids on Nio EV assembly lines, and Fourier Intelligence, which focuses on rehabilitation and factory material handling. The use cases are modest compared to Tesla’s Optimus demos, but the deployments are real and recurring, which is what builds manufacturing confidence and cost reduction.
The stack breakdown reveals where competitive moats actually live. Actuators, specifically harmonic drive and quasi-direct drive joints, are the most critical and hardest to scale. China’s actuator supply chain is deep, with Leadshine, Hiwin, and dozens of smaller suppliers serving robotics companies at competitive prices. Battery technology is essentially solved at the chemistry level, with LFP cells cheap and energy-dense enough for 2-4 hour operational windows. Force sensors and cameras are commoditised through the same supply chain that stocks every smartphone. The two layers where Western companies currently lead are foundation models for manipulation and locomotion policies, and factory integration software that connects robotic workflows to existing ERP and MES systems. Those leads are eroding as Chinese labs like Tencent Robotics and Shanghai AI Lab accelerate on embodied foundation model research. Distribution and after-sales service remain a Western advantage for enterprise buyers outside China.
Whether humanoid robots are demos or nearing EV-style cost curves depends on the task. Carrying boxes, loading pallets, and pushing carts in warehouses and factories is technically solved today. The question is cost per productive hour versus a human worker or conventional automation. At $50,000 to $100,000 per unit with current failure rates and maintenance costs, humanoids are not yet economically competitive with industrial arms or conveyor systems for repetitive structured tasks. At $16,000 with improving reliability, the calculation is shifting. Goldman Sachs projected a $38 billion humanoid market by 2035. Morgan Stanley’s higher estimate of $1.5 trillion by 2040 assumes cost curves that parallel EV adoption. Neither figure is currently validated, but the cost trajectory is moving in the right direction and China controls most of the inputs that determine how fast it moves.
For SF readers and founders, the strategic question is not whether to compete with Unitree on price. That battle is lost before it starts for any company assembling in California or Germany. The competitive positions worth building are in the software layers that China’s hardware abundance cannot commoditise quickly: manipulation foundation models trained on real-world factory data, integration platforms that connect humanoid workflows to enterprise systems, and domain-specific deployments in healthcare, elder care, and hazardous environments where data scarcity and regulatory requirements slow Chinese market entry. Companies like Physical Intelligence, whose pi0 foundation model claims cross-robot transferability, and Sanctuary AI, focused on task generalisation, are betting on software moats that persist even as hardware costs collapse. That is the right frame: build where compute and data create defensibility rather than where actuators and supply chains decide the outcome.
European and American governments are watching and starting to respond. The US House recently introduced bills restricting Chinese robotics companies from critical infrastructure. The EU is examining standards that could limit imports on safety or data grounds. Those measures create time but not competitive advantage on their own. The companies that use that time to close the software gap while building cost-competitive hardware through non-Chinese supply chain alternatives, Vietnam, India, Mexico, will be positioned when the regulatory protection eventually normalises. Those that rely on policy protection without closing the capability and cost gap will face the same reckoning European automakers are facing now.
Also read: ParoQuant shows that better quantization math is now the fastest route to practical reasoning models • Skyroot’s unicorn status puts India’s private launchers on the global venture map • New MIT research shows automation targets wages, not just headcount, and AI startups are selling the tool
