A turning point has arrived for the humanoid robotics industry—China's leading OEMs have launched pilot projects at a scale of hundreds on manufacturing production lines, Tesla's Optimus is moving toward mass production, and Boston Dynamics is preparing deliveries to Hyundai. According to JPMorgan's latest report, the gap between industry leaders and smaller players is widening, with capital increasingly flowing toward platform-based enterprises and core component suppliers. Small OEMs without clear order visibility are seeing their window for survival close rapidly.
In the past month, the global humanoid robotics industry has seen developments occur at a higher frequency than in the entire previous year.
China's leading OEMs are already running pilot projects at triple-digit scales in logistics sorting and manufacturing production lines, with requests for expansion; Boston Dynamics is preparing inventory to deliver Atlas to Hyundai's automotive factories. $Tesla (TSLA.US)$ The announcement stated that mass production of Optimus will commence in 2026 at the Fremont factory, while the second factory in Texas is scheduled for 2027. The industry has already reached a turning point from "trade show robots" to "real-world operational deployment."
According to Trading Wind, Karen Li, an analyst covering infrastructure and industrial sectors in the Asia-Pacific region at JPMorgan, stated in her latest report that the gap between leaders and "long-tail players" is widening. Capital is increasingly concentrated in two types of entities: profitable platform companies ready for mass production, and suppliers providing high-quality components and AI/software “brains.” The research team completed a joint survey on robotics and auto shows in Beijing from April 22 to 24, hosted a webinar for global investors on the robotics industry on April 16, and attended the M+ Forum in Malaysia during the same period, engaging with Work E Robotics, the authorized Malaysian partner of Unitree Robotics.
Among these, Tesla’s position in this race is more complex than what its market valuation suggests. The company plans capital expenditures exceeding $25 billion in 2026 for AI, robotics, and self-developed chips, with the Optimus V3 design nearing readiness for mass production. However, the report maintains an "underweight" rating for Tesla, with a target price of $145, compared to its current share price of approximately $374.
By contrast, China’s leading OEMs, leveraging government procurement, supply chain advantages, and rapid hardware iteration, are advancing commercialization at a faster pace, while Boston Dynamics holds a first-mover advantage in industrial integration.
The real bottleneck is not whether the hardware can move but whether it can operate stably.
The most consistent feedback from the Beijing survey is that the industry’s main obstacle has shifted from “whether prototypes can complete tasks” to “whether they can reliably complete tasks under mass production conditions.” Reliability, maintenance cycles, and integration time with production lines were recurring themes.
A supplier of dexterous robotic hands revealed that shipments exceeded 10,000 units in 2025, with expectations to double in 2026. This figure provides a benchmark, illustrating how rapidly commercial demand is transitioning from data collection and showroom displays to real-world deployment scenarios. However, suppliers of hand components emphasized that scaling challenges now involve comprehensive testing for temperature, vibration, corrosion, and durability, rather than just “whether the robot can grasp objects.”
Physical intelligence — the “brain” of robots — is widely regarded as the core bottleneck for commercialization by 2026. Vision-Language-Action (VLA) models are responsible for mapping language and video understanding into robot actions, while world models handle reasoning, planning, and environmental understanding. These are considered the ultimate pathways for embodied AI. The transition from simulation to reality (sim-to-real) remains a universal challenge. A leading humanoid robotics company is addressing real-world data bottlenecks, particularly in force, friction, and tactile realism, using “control the world” tools and data infrastructure layers.
Commercialization strategies are also diverging. Some OEMs choose to bundle the “brain” with the robot body for sale, others sell only hardware, while some provide SDKs for customers to develop their own intelligent layers. Underlying this divergence is a key insight: much of today’s commercial traction comes from large tech companies and industrial clients using robots as data collection tools, rather than purely as labor replacements.
Tesla is betting on the right direction, but its timeline may be handing gifts to competitors.
The $250 billion capital expenditure plan is real, and mass production of Optimus is a genuine strategic priority. However, Tesla's management itself acknowledges that the initial capacity ramp-up will be slow. The Fremont factory aims for an annual capacity of one million units, with a second plant in Texas for further expansion, but these developments are targets for 2027 and beyond.
The company has deliberately suppressed public demonstrations of Optimus 3, citing intellectual property protection and prevention of competitor replication – an explanation that underscores the intensity of competition. Meanwhile, Tesla’s self-developed AI5 chip will power both Optimus and data centers, with vertically integrated chip manufacturing forming a key part of its defensive moat. However, this path requires a longer construction cycle than robot mass production.
JPMorgan characterizes Tesla as chasing leading Chinese OEMs and Boston Dynamics, rather than leading the pack. Competitive pressures and margin issues in its core EV business have not disappeared, and the current 187x forward price-to-earnings ratio for 2026 is almost entirely predicated on AI and robotics.
Chinese manufacturers are advancing faster due to government support, supply chain strengths, and rapid iteration, while Boston Dynamics is also accelerating the Atlas timeline for Hyundai's industrial applications (with initial deployments potentially starting from 2028).
The logic in Southeast Asia differs from China: it’s not about saving labor costs, but avoiding human presence in hazardous areas.
The Malaysia M+ Forum provided an intriguing cross-section. For Southeast Asian companies considering robot deployment, the primary driver is not labor cost – the low-wage environment renders this logic ineffective – but operational flexibility: 24/7 operation, replacing humans in dangerous environments, and consistent work quality.
The oil and gas industry represents the most definitive current buyer. Unitree’s B2 quadruped robot (IP protection rating, approximately 40 kg payload) is under consideration for gas leak detection and perimeter patrols. Manufacturing is seen as the next wave of adoption, with Unitree’s G1 humanoid robot (upper body plus AMR base) targeting logistics in factories with controlled ground conditions. However, scaled procurement may require two to three more years of solution maturity and further hardware cost reductions.
The business model currently revolves around outright purchases, though discussions around RaaS (Robotics as a Service) and subscription models are increasing. Work E Robotics’ management was direct: “Once hardware becomes standardized, solutions become the key.” Integration and deployment – including site modeling, LLM, and sensor stitching – will emerge as the true source of profit and customer stickiness.
The financing window is narrowing, but the way it is narrowing favors market leaders.
The primary market financing remained active in early 2026, but the landscape is shifting: valuations are being pushed higher while capital increasingly concentrates among a few platform enterprises and high-quality component suppliers. The challenge for small original equipment manufacturers (OEMs) lies in the叠加 of three cost factors—computational power for very large model training, data acquisition expenses, and manufacturing capacity ramp-up—resulting in funding requirements far exceeding what they can raise.
This divergence will likely lead to mergers and acquisitions, strategic partnerships, and structured financing rather than independent growth for smaller OEMs at the tail end of the market. The IPO pipeline serves as a key catalyst, with government procurement and public sector projects becoming major drivers of orders, particularly following local governments' establishment of data collection centers and pilot zones, which have intensified order concentration toward the backend.
At the stock level, China's robotics sector has rebounded by an average of 10% over the past month. However, JPMorgan’s logic is clear: companies with commercial momentum, strong order visibility, and differentiated technology merit overweight positions; for smaller OEMs lacking these traits, the environment will only grow more challenging.
Editor/Joe