Optimus capacity expansion and China's volume ramp resonate together, while the humanoid robot cost curve continues downward

Overview

Based on Nomura's advanced manufacturing field research and recent industry surveys, this report discusses the mass-production cycle of humanoid robots, the paradigm shift toward non-robot data collection, and the trend toward complete-system contract manufacturing. The report believes that overseas Optimus is moving from SOP into scaled manufacturing, while China's humanoid robot shipments are being revised up under the push from government procurement and consumer demand; at the same time, falling costs and a maturing contract manufacturing system will drive industry scale-up.

Core views

The core views include: first, Tesla Optimus capacity expansion has been revised upward, with a target of about 1,000 units per week by September 2026, and supply chain signals point to around 25k shipments in 2026; second, China's humanoid robot shipments in 2026 are expected to reach about 40k-50k units, with domestic manufacturers overall running ahead of overseas peers; third, the 2026 volume ramp will be driven mainly by consumer demand rather than data collection, and non-robot data collection will begin to compress demand for real-robot collection from 2027 onward; fourth, the share of complete-system contract manufacturing may rise over the long term, and the few OEMs that build their own supply chains may gain pricing advantages by lowering BOM costs; fifth, industrial applications remain constrained by validation bottlenecks in precision, takt time, and hardware cost.

Analysis framework

The report uses a combination of industry research, supply-chain order signals, communication with company management teams, and application-scenario breakdowns to compare Tesla, Figure AI, Boston Dynamics, and Chinese manufacturers in terms of capacity, shipments, costs, application mix, and supply-chain organization.

Methodology notes

• Industry researchAdvanced manufacturing field research

  Industry conclusions are formed through visits to Unitree, Deep Robotics, Shenhao Technology, Mech-Mind Robotics, and companies in Tesla's supply chain.

  This method is used to assess complete-system capability, supply-chain maturity, application deployment, and contract manufacturing trends, but the conclusions depend on research samples and supply-chain feedback.

• Supply chain verificationCross-validation of order signals and capacity ramp-up

  Use production-line expansion guidance, order signals, and weekly production targets to infer 2026 shipments for Optimus and other overseas manufacturers.

  The report combines Fremont and Austin production-line plans, the September ramp target, and supply-chain order signals to arrive at an estimate of about 25k ±10k Optimus shipments in 2026.

• Cost curve analysisData collection cost substitution framework

  Compare the cost and speed of real-robot data collection versus non-robot data collection such as teleoperation, UMI, and Ego.

  The report believes non-robot collection costs about 20% of real-robot collection and is faster, so future training data structures may shift to roughly 90% non-robot data and 10% real-robot data.

Asset mapping & comparison

Structured mapping from thesis to named assets (strengths, weaknesses, peers, risks).

• Tesla (TSLA US)

  Core overseas complete-system manufacturer for Optimus; labeled Not rated in the report.

  Strengths

  Capacity planning has been revised upward, the Fremont line target has increased, and the planned second Austin line provides room for further expansion.

  Weaknesses

  Its overseas mass-production pace is still considered slower than that of domestic Chinese manufacturers.

  Comparison

  Compared with Figure AI and Boston Dynamics, Optimus shows higher 2026 shipment signals; compared with Chinese manufacturers, its production pace is still slower.

  Risks

  Production-line ramp-up below expectations, insufficient supply-chain delivery, and software and application-scenario validation progressing more slowly than expected.

• Chinese humanoid robot OEMs

  The main volume-growth market and manufacturing cluster highlighted by the report.

  Strengths

  2026 shipments are expected to reach about 40k-50k units, with leading manufacturers potentially reaching 10k-15k units each, supported by falling costs and local supply chains.

  Weaknesses

  Breadth of B-end applications and model capability still need validation, and industrial applications remain in the exploration and validation stage.

  Comparison

  Overall shipments and production pace are ahead of most overseas manufacturers.

  Risks

  Insufficient realization of low-cost consumer demand, changes in the pace of government procurement, and weaker-than-expected data quality and standardization.

• Unitree (Unlisted)

  One of the interviewed Chinese robotics manufacturers and cited in the report as an example of a self-built supply chain.

  Strengths

  A self-built supply chain may reduce complete-system BOM costs and strengthen pricing power.

  Weaknesses

  Model capability, breadth of B-end applications, and long-term scalability still need validation.

  Comparison

  Compared with more second- and third-tier OEMs that outsource, a self-built supply chain offers advantages in cost and control.

  Risks

  Higher capex and operational complexity for a self-built supply chain, and demand volatility could affect the payback period.

• Figure AI (Unlisted)

  An overseas humanoid robot manufacturer used for comparison with Optimus and Chinese manufacturers.

  Strengths

  Its BotQ line has publicly reached one unit per hour, implying a 2026 pace of about 500-1,000 units.

  Weaknesses

  Overall shipment scale is significantly smaller than that of Optimus and the Chinese market.

  Comparison

  Like Boston Dynamics, it is in the second tier of overseas mass-production pace and trails leading Chinese manufacturers.

  Risks

  Uncertainty around mass-production ramp-up, cost declines, and the speed of commercial application deployment.

• Boston Dynamics (Unlisted)

  An overseas humanoid robot/robotics company used for output comparison.

  Strengths

  Has accumulated robotics technology, with about 500-1,000 units expected in 2026.

  Weaknesses

  Mass-production scale remains limited.

  Comparison

  Similar to Figure AI, but its overall production pace still lags Chinese manufacturers.

  Risks

  Uncertainty around commercial product positioning, cost, and expansion of downstream demand.

• Complete-system contract manufacturers

  An industry-chain segment benefiting from the outsourcing trend among second- and third-tier OEMs.

  Strengths

  Entrants from the automotive supply chain have end-to-end supply-chain management and large-scale assembly experience, while 3C manufacturers have short-term brand advantages.

  Weaknesses

  Hardware supply chains are capital-intensive with long payback periods, and order volatility may affect capacity utilization.

  Comparison

  In the long run, manufacturers with automotive supply-chain backgrounds may be stronger than typical 3C manufacturers, while in the short run 3C manufacturers may find it easier to gain customer recognition.

  Risks

  A higher-than-expected share of OEM self-building, price declines squeezing contract-manufacturing margins, and rising quality and delivery standards.

Key data

• Optimus 2026 shipmentsabout 25k ±10k unitsDerived from industry-chain order signals; Tesla Optimus remains a Not rated related name.
• Optimus September 2026 ramp targetabout 1,000 units/weekThe report says the upgraded capacity plan is supported by the Fremont retrofit line and the subsequent Austin line.
• Fremont production-line targetannualized capacity of about 70k units in 2027Higher than the previous target of about 50k units.
• Austin second-line targetabout 70k units in 2028The report says the long-term combined capacity target has been raised further.
• China 2026 humanoid robot shipmentsabout 40k-50k unitsDriven by procurement from government embodied-intelligence bases and an inflection in demand for low-cost consumer products in 2H26F.
• China downstream demand mixconsumer about 30%, performance/entertainment about 30%, government procurement about 20%, education about 15%, other about 3%-5%The report's estimate of the 2026 downstream mix.
• Non-robot data collection costabout 20% of the cost of real-robot dataIncluding routes such as teleoperation, UMI, and Ego data.
• Mixed-data training mixabout 90% non-robot data and 10% real-robot dataThe report expects this structure to become the standard, compressing demand for real-robot collection.
• Investment in a large data collection factoryabout CNY50-100mnA single large facility with about 1,000 units in scale, with a payback period of about 3-5 years.
• Mainstream product price rangefull-size about CNY150k-300k, small-size about CNY10k-100k2026F prices have already fallen by more than 50% year over year, and the report expects the decline to narrow in 2027F.

Impact & implications

The industry impact is that pressure on hardware procurement may decline because of non-robot data collection, while the importance of data quality, consistency, and cross-platform standardization rises; meanwhile, rapidly falling costs and a maturing contract manufacturing system will lower the barrier to scale, but may also compress the differentiation space for complete-system makers. For the industry chain, contract manufacturers with an automotive supply-chain background may gain long-term advantages through end-to-end management and large-scale assembly experience, while 3C manufacturers may have short-term advantages in brand and manufacturing recognition.

Risks

• Industrial deployment is still constrained by precision, takt time, and hardware costs, and the validation cycle may be longer than expected.
• If the demand inflection for low-cost consumer products does not materialize, the upward revision to China's 2026 shipments may face downward adjustment.
• Although non-robot data collection reduces costs, the difficulty of data quality, consistency, and cross-platform standardization increases.
• Changes in the pace of government procurement and investment in data collection bases may affect orders and cash flow.
• Rapid declines in complete-system prices may compress OEM and supply-chain profit margins.
• The report involves many Not rated or unlisted companies, lacking continuous rating coverage and public financial data.

What to watch

• Whether Optimus reaches the ramp target of about 1,000 units per week by September 2026.
• Whether Fremont and Austin production-line expansions proceed as planned.
• Whether leading Chinese manufacturers achieve shipments of 10k-15k units in 2026.
• Actual orders and repurchase trends after the launch of low-cost consumer products in 2H26F.
• The share of non-robot data collection in training data, quality standards, and cross-platform compatibility.
• Progress in software iteration and commercial deployment in scenarios such as retail, guided tours, exhibitions, and multi-SKU picking.
• Changes in the share of complete-system contract manufacturing orders between automotive supply-chain players and 3C manufacturers.