China Leads as Humanoid Robotics Reach Commercial Turning Point

Over the past month the global humanoid‑robotics sector has seen developments more concentrated than in the prior year. Leading Chinese OEMs are already operating large‑scale pilot projects in logistics and manufacturing and are seeking expansion. Boston Dynamics is preparing Atlas deliveries to Hyundai factories, and Tesla has announced Optimus mass production at Fremont in 2026 with a Texas expansion in 2027. The industry has shifted from exhibition demonstrations to real‑world deployment.

JPMorgan’s Asia‑Pacific analyst Karen Li concluded that the gap between winners and smaller players is widening, with capital concentrating on profitable, production‑ready platforms and suppliers of high‑quality components and AI/software “brains.” The team conducted field research in Beijing, hosted a global investor webinar, and attended Malaysia’s M+ forum, engaging with Work E Robotics, Unitree Robotics’ authorised partner.

Tesla’s position is more complex than its valuation suggests. The company plans over $25 billion in 2026 capex for AI, robotics and chips, with Optimus V3 close to production readiness, yet JPMorgan maintains an underweight rating with a $145 target price against a current $374 share price. Chinese OEMs are advancing faster through government procurement and rapid iteration, while Boston Dynamics retains an edge in industrial integration.

Industry feedback shows the bottleneck has shifted from prototype capability to stable operation under mass‑production conditions, with reliability, maintenance cycles and integration time as key concerns. A dexterous‑hand supplier reported shipments exceeding 10,000 units in 2025 and expects to double in 2026, illustrating rapid demand growth. Component suppliers emphasised that scaling now requires comprehensive testing for durability, temperature, vibration and corrosion.

Physical intelligence is widely seen as the core bottleneck for 2026. VLA models map language and video into actions, while world models handle reasoning and planning. Sim‑to‑real transfer remains challenging, and leading firms are building tools and data infrastructure to improve fidelity of force, friction and tactile data. Commercial strategies vary, with some OEMs bundling intelligence with hardware, others selling hardware alone, and some offering SDKs. Much of current traction comes from large tech and industrial firms using robots as data‑collection tools rather than pure labour substitutes.

Tesla’s strategic direction aligns with industry trends, but its timeline may benefit rivals. Management acknowledges slow initial ramp‑up, with Fremont’s 1 million‑unit annual target and Texas expansion not materialising until 2027 or later. Public demonstrations of Optimus 3 are limited to protect intellectual property, underscoring competitive intensity. Tesla’s AI5 chip will power both Optimus and data centres, but chip‑manufacturing integration is a longer‑term project. JPMorgan views Tesla as catching up to Chinese OEMs and Boston Dynamics rather than leading, with EV competition and margin pressures persisting and valuation heavily reliant on AI and robotics.

Regional adoption differs. At Malaysia’s M+ forum, Southeast Asian firms cited operational resilience — 24/7 operation, hazardous‑environment substitution and consistent quality — as the main rationale, not labour‑cost savings. Oil and gas is the clearest buyer, with Unitree’s B2 quadruped considered for leak detection and patrol. Manufacturing is expected to follow with Unitree’s G1 humanoid for factory logistics, though large‑scale procurement may require two to three years of solution maturity and cost reduction.

Business models are dominated by outright purchases, though RaaS and subscription models are gaining traction. Work E Robotics noted that once hardware standardises, solutions become the differentiator, with integration and deployment as the true source of margin and customer stickiness.

Financing remains active but is narrowing toward leaders. Capital is concentrating on platform firms and component suppliers, while smaller OEMs face funding gaps from combined compute, data and manufacturing costs. This divergence is likely to drive M&A and partnerships rather than independent growth. Government procurement and public projects are becoming key order drivers, with local data‑collection centres and pilot parks producing concentrated demand.

China’s robotics sector has rebounded about 10% in the past month. JPMorgan’s thesis is clear: companies with commercial momentum, strong order visibility and differentiated technology warrant heavy allocation, while smaller OEMs without these traits face increasing difficulty.