Production targets for Tesla's Optimus have been significantly raised, with Chinese manufacturers also revising their shipment forecasts upwards. Global capital is pouring into the humanoid robot industry, which is now at a critical juncture for mass production. Recent reports from Nomura and UBS both point to the same conclusion: the industry's logic is shifting from "technology validation" to "scaling production." However, several key bottlenecks remain before the sector experiences its true breakthrough moment.
According to Nomura's report, Tesla has raised the annualized capacity target for its Fremont production line for the Optimus Gen 3 from the previous 50,000 units to approximately 70,000 units. The company also plans to add around 70,000 units of capacity in Austin by 2028, with a long-term combined capacity target reaching 1.5 million units. Signals from the industry supply chain indicate that Optimus shipments for 2026 are projected to be about 25,000 units (with a variance of 10,000 units). The weekly production target is expected to surge to around 1,000 units by September.
Concurrently, UBS notes that a wave of strategic capital activity in the global humanoid robot industry is concentrated in the first half of 2026. China's Ministry of Industry and Information Technology has set a goal to deploy 10,000 humanoid robots across over 100 application scenarios by the end of 2027. The Shanghai municipal government plans to deploy 100,000 units in factories around 2030.
Both institutions have observed that current market sentiment is constrained by multiple factors. UBS data shows that China's humanoid robot index has underperformed the machinery index by about 9 percentage points year-to-date. The main drags include the delayed mass production of Optimus Gen 3, capital rotation towards data centers and commercial spaceflight, and generally weak trading momentum. Nonetheless, UBS has listed potential catalysts for the second half of the year, including the official launch of Optimus Gen 3 (July-August), progress on Unitree's IPO, the World Artificial Intelligence Conference (July, Shanghai), and the World Robot Conference (August, Beijing).
Tesla's Capacity Target Increase Sets Global Production Pace
Nomura's report indicates that Tesla is advancing the Optimus Gen 3 from the initial production launch phase to the batch manufacturing stage. The annualized capacity target for the Fremont line (retrofitted from the original Model S/X line) has been raised to about 70,000 units. A second line in Austin is expected to contribute a similar scale by 2028, with the long-term combined capacity target around 1.5 million units. Based on industry research, Nomura expects the weekly production target to reach approximately 1,000 units in September.
Among non-Chinese players, Figure AI and Boston Dynamics are each tracking shipments of about 500 to 1,000 units this year, with other overseas manufacturers at around 100 to 200 units. Figure's BotQ production line has publicly achieved a cycle time of one unit per hour. Overall, the production pace of these overseas manufacturers still lags behind that of leading domestic Chinese makers.
UBS's report places the official launch window for Optimus Gen 3 between July and August, identifying it as one of the most significant market catalysts for the latter half of the year.
Chinese Shipment Forecasts Revised Upward, Driven by Government and Consumer Demand
Nomura has revised its 2026 humanoid robot shipment forecast for China upward to approximately 40,000 to 50,000 units. This is driven by accelerated government procurement for embodied intelligence bases and a potential inflection point in consumer demand in the second half of the year spurred by the launch of lower-priced products.
In terms of company tiers, Nomura estimates that the top two firms will ship about 10,000 to 15,000 units (a 2 to 3-fold year-on-year increase). Several second-tier companies are expected to ship around 3,000 units each, while third-tier companies will ship approximately 500 to 1,000 units. Regarding downstream application structure, consumer-facing applications are expected to account for about 30% of 2026 shipments, performance and entertainment for another 30%, government procurement (data collection) for 20%, education for 15%, and the remaining 3% to 5% for other commercial and industrial applications.
UBS's report also notes that Chinese policy is pushing the industry from "demonstration" towards real "work mode" deployment across factories, logistics, healthcare, and home scenarios. The Ministry of Industry and Information Technology aims to complete the deployment of 10,000 units across over 100 application scenarios by the end of 2027, while the Shanghai municipal government has set a target of deploying 100,000 units in factories by 2030.
Data Collection Paradigm Shift: Non-Robot Methods May Constrain Demand
Nomura's report highlights a structural change with profound implications for the industry landscape: data collection is accelerating its shift towards "robot-free" methods, including teleoperation, UMI (Universal Manipulation Interface), and Ego (first-person perspective) data capture. These methods cost about 20% of using physical robots and are significantly faster.
As mixed data training (approximately 90% non-robot data, 10% physical robot data) becomes the industry standard, demand for physical robot data collection is expected to contract further in 2026. The government remains the primary funder for data collection facilities. A large facility with about 1,000 units requires an investment of roughly 50 to 100 million yuan, with a payback period of about 3 to 5 years.
While this shift alleviates the upfront hardware procurement pressure on humanoid robot OEMs, it also raises the bar for data quality, consistency, and cross-platform standardization. Nomura points out that industrial deployment still faces validation bottlenecks concerning precision, cycle time, and model training costs.
Trend Towards Outsourcing Assembly is Clear, In-House Supply Chain Offers Cost Advantage
Based on field research, Nomura states that the trend towards outsourcing the assembly of complete machines is quite clear. Given the capital-intensive nature of hardware supply chains and long payback periods, most second and third-tier OEMs prefer to outsource. In the long run, the proportion of outsourced assembly may increase further. However, some OEMs that still build their own supply chains (like Unitree Technology) can directly strengthen their pricing power by reducing material costs for the complete machine.
Among potential outsourcing partners, Nomura believes those with an automotive supply chain background possess stronger long-term competitiveness due to their experience in end-to-end supply chain management and large-scale assembly. Consumer electronics manufacturers, on the other hand, hold some advantages in terms of recent brand effects.
Regarding pricing, Nomura expects mainstream full-size products to be priced between 150,000 and 300,000 yuan, with smaller-sized products ranging from 10,000 to 100,000 yuan. Price declines in 2026 have already exceeded 50% year-on-year, and Nomura estimates that the rate of price reduction will narrow in 2027.
Global Capital Accelerates Entry, UBS Favors Upstream Core Components
UBS's report outlines recent significant developments in the global humanoid robot field: NVIDIA is actively supporting an ecosystem that includes Chinese startups like Unitree and Korean companies (notably Hyundai); OpenAI formally re-entered the humanoid robotics field in June, establishing an "OpenAI Robotics" division and recruiting extensively for hardware, systems, and AI talent; Germany's Neura Robotics completed a substantial Series C funding round of up to $1.4 billion, with a reported production target of 6,000 units this year and over 10,000 units by 2027; Agility Robotics announced on June 24 its intention to go public via a merger with Churchill Capital Corp XI, with a pre-transaction valuation of $2.5 billion, expected to close in the fourth quarter of 2026.
Automotive companies are also accelerating their plans. The CEO of XPeng took personal charge of the robotics division in June 2026, with the goal of achieving mass production of the "IRON" humanoid robot by year-end. BYD Company Limited has confirmed it is developing a humanoid robot, planning to deploy up to 20,000 units internally before the end of the year, which would be one of the largest real-world scenario deployments globally.
Regarding investment strategy, UBS expressed a preference for upstream component companies with strong core businesses and high-barrier exposure to the humanoid robot field, focusing on areas like sensors, reducers, and ball screws. UBS also cautioned that, despite the gathering industry momentum, it believes it may still take some time before humanoid robots truly experience their own defining "electric vehicle moment."
Comments