China's Embodied Intelligence Enters an Accelerating Phase of Mass Production and Commercial Validation

Overview

The report focuses on China's physical AI and embodied intelligence, with emphasis on the mass-production progress of the humanoid robot industry, value distribution across the supply chain, localization opportunities, policy support, and training infrastructure. It argues that China's manufacturing depth, supply-chain costs, policy coordination, and accumulation of real-world data could accelerate the commercialization of humanoid robots.

Core views

Humanoid robots currently remain small within China's embodied intelligence market, but have substantial future growth potential. In the near term, value capture is concentrated primarily in hardware, especially actuation-layer components, precision transmission systems, sensors, chips, materials, and equipment. As scale expands, lower component costs, more training data, and improved battery performance will drive adoption; however, large-scale commercialization will still depend on measurable productivity gains and whether usage costs can approach labor costs.

Analysis framework

The report conducts thematic research on the commercialization path of physical AI and investment beneficiary segments by combining global shipment forecasts, China's supply-chain segmentation, component localization rates, comparisons of supply-chain financial metrics, policy documents, and progress in training infrastructure.

Methodology notes

• Supply Chain AnalysisPerception-Planning-Actuation Three-Layer Framework

  Classifying the value chain by robot functional modules

  The report divides the ecosystem into perception, planning, and actuation layers, and notes that the actuation layer accounts for the highest share of the bill of materials and is currently the core of hardware value capture.

• Commercialization AssessmentCost and Productivity Thresholds

  Measuring adoption pace through annual usage costs and quantifiable productivity

  The report considers usage cost to be the key determinant of commercialization speed; commercial viability will improve when annual usage costs fall to approximately RMB80,000, close to manufacturing worker costs.

• Technology PathwayDual-Track Sim-to-Real Training

  Combining virtual simulation training with real-world scenario fine-tuning

  Robots first acquire foundational skills at low cost in simulated environments, then undergo fine-tuning and validation in real-world settings such as factories, warehouses, and service environments to shorten the cycle from R&D to deployment.

Asset mapping & comparison

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

• China Humanoid Robot Core Components

  Primary beneficiary segment

  Strengths

  High value contribution from the actuation layer; opportunities for localization and rising demand in screws, force sensors, bearings, chips, and materials.

  Weaknesses

  Gaps remain in high-end precision, reliability, and technology accumulation, and the conversion of scaled orders remains uncertain.

  Comparison

  Compared with OEMs and solution providers, core components may offer greater value capture and domestic-substitution upside during the mass-production ramp-up phase.

  Risks

  Commercialization progressing more slowly than expected, price competition, changes in technology pathways, and failure to meet expectations in key performance validation.

• China Robot OEMs and System Integrators

  Commercialization beneficiary segment

  Strengths

  Can directly benefit from policy support, real-world scenario deployment, and market expansion while accumulating operating data.

  Weaknesses

  Large-scale profitability and differentiated competitiveness remain limited, while hardware costs and application deployment pressures are high.

  Comparison

  Compared with upstream components, OEMs are more dependent on end-demand, scenario validation, and ongoing service capabilities.

  Risks

  Insufficient unit economics, tighter safety and compliance requirements, delayed customer procurement, and intensifying industry competition.

• Artificial Intelligence Computing Power, Training Platforms, and Software Ecosystem

  Medium- to long-term infrastructure beneficiary segment

  Strengths

  Domestic computing power, robot training facilities, and real-world data networks can support model iteration and industry self-sufficiency.

  Weaknesses

  Returns on investment and business models are still emerging, while gaps remain in foundation models and high-performance chip capabilities.

  Comparison

  Value realization in this segment depends heavily on the scale of industry deployment and the maturity of data closed loops, so its realization period may be longer than that for components.

  Risks

  Computing-power supply, chip performance, data compliance, model capabilities, and returns on capital expenditure falling short of expectations.

Key data

• Global Humanoid Robot Shipment Forecast18,000 units in 2025; 60,000 units in 2026; 1.75 million units in 2030The report forecasts a compound annual growth rate of approximately 150% for shipments from 2025 to 2030.
• China's Share of Global ShipmentsMore than 50%The report expects China to maintain a share of more than half of global humanoid robot shipments.
• Planetary Roller Screw Localization RateApproximately 40% to 50%This component accounts for approximately 22% of the bill of materials and is regarded as a segment with significant localization opportunities.
• Six-Axis Force Sensor Localization RateBelow 30%This component accounts for approximately 10% of the bill of materials, leaving substantial room for domestic substitution.
• Embodied Intelligence Market SizeApproximately RMB1 trillionChina's embodied intelligence market size cited by the report.
• Humanoid Robot Market SizeApproximately RMB3 billion to RMB8.5 billionThe report notes that it currently represents only a small portion of the embodied intelligence market.
• Commercialization Cost ReferenceApproximately RMB80,000 per yearThe usage-cost threshold cited by the report, close to the average annual cost of a manufacturing worker.

Impact & implications

In the early stage of rising industry shipments, suppliers of technically demanding core components with low localization rates may benefit more directly than downstream original equipment manufacturers, solution providers, and foundation-model participants. Over the medium to long term, data generated through real-world deployment can accelerate model iteration and expand application scenarios, but industry competition will shift from capacity expansion toward core technology, reliability, cost, and commercialization capabilities.

Risks

• Humanoid robot scaled commercialization and end-market procurement progress may fall short of expectations.
• Technology gaps in core areas, including high-end reducers, servo motors, advanced sensors, foundational algorithms, and high-performance AI chips, may persist.
• Equipment prices remain significantly above labor costs, potentially limiting adoption in industrial and service scenarios.
• Requirements concerning safety, data privacy, liability allocation, military-use ethics, and human-robot relationship governance may raise deployment barriers.
• Displacement of low-skilled jobs and retraining pressures may create social adaptation and policy implementation risks.

What to watch

• Actual global and China humanoid robot shipments, orders, and mass-production ramp-up speed.
• Whether annual usage costs move toward the labor-cost benchmark of approximately RMB80,000.
• Localization rates and profitability improvement for planetary roller screws, six-axis force sensors, bearings, chips, and materials.
• Implementation progress of MIIT and SASAC real-world scenario training initiatives and the "six networks" infrastructure.
• Robot counts, scenario coverage, and data accumulation at robot training facilities in Shanghai, Beijing, Hangzhou, and other locations.
• Validation of quantifiable productivity in industrial, logistics, service, healthcare, eldercare, and public-safety scenarios.