Analysis for informational purposes only. Capital at risk.
The headlines for China’s humanoid robots focus on televised dancing and running.
The reality is much quieter: a rapid build-out of local supply chains and early factory deployments.
To look past the show-floor demonstrations and understand the actual market, we assess these key questions:
- Value Chain: Why is China dominating the supply chain?
- Utility: What can these robots actually do today?
- Constraints: What engineering hurdles are delaying adoption?
- Timeline: When will mass commercial deployment happen?
- Unit Economics: What is the financial break-even point for replacing human labour?
Humanoid Volume: The China Monopoly
In 2025, five Chinese startups (Unitree, AgiBot, UBTech, Leju, EngineAI and Fourier) accounted for roughly 85% of humanoid robot volume, according to Omdia.
U.S. players (Agility, Figure AI, Tesla) represented just 3%.
Europe’s Silent Adoption
Recognising this hardware lead, European industrial incumbents are already deploying Chinese humanoids.
Airbus is deploying UBTech’s Walker S2 directly into its aircraft manufacturing facilities.
BMW is piloting the Swedish-built “Aeon” humanoid, which relies heavily on core components—such as five-finger dexterous grippers—sourced from Chinese suppliers.
German conglomerate Schaeffler has partnered with Leju and ROKAE to deploy humanoids and establish a robotics subsidiary in China, while German integrator EnduX has pivoted to importing Chinese humanoid platforms for its European clients.
The EV Playbook: The Supply Chain Moat
This massive volume lead is the direct result of a highly domesticated supply chain.
Manufacturing advantage: China’s existing ecosystems for smartphones, EVs and consumer electronics have the core inputs for humanoids — batteries, servos, sensors, cameras, etc. The supplier base is already in place.
Economies of scale: In 2025, UBTech partnered with Foxconn to pilot Walker S1 robots on iPhone assembly lines. Lingyi iTech (a Foxconn supplier) is building a Beijing “super‑factory” targeting 10,000 humanoids/year by end‑2026 and 500,000/year by 2030.
Price competitiveness: Unitree lists the G1 at ~$16k. Tesla’s Optimus, which leverages its own Shanghai supply chain, targets $20k–$30k. In contrast, Boston Dynamics’ Atlas remains many times pricier, underscoring a severe Chinese cost advantage.
The Reality Check: Three Tiers of Utility
Having said that, volume does not equal capability.
While acrobatics look great on video, factory floors are unpredictable. To understand actual utility, we break robotic skills down into three tiers of difficulty:
L1: The structured factory (Happening now): Repeatable motions in a rigid, tightly controlled environment. Tasks have high repeatability with minimal sensing or decision‑making.
- Example: A robot in a fenced car assembly station that picks the door panel from a fixed conveyor position and places it onto a welding jig. In addition, Figure AI recently hosted a livestream showing a team of humanoid robots (F.03) sorting packages for over 40 hours.
L2: The shared workspace (Testing): Multi‑step tasks in environments where humans and robots coexist. Robots have basic perception, path planning and simple error recovery. Robots operate with pre‑mapped layouts and modest variability.
- Example: A logistics robot that navigates aisles, identifies a cardboard box of expected dimensions, retrieves it and carries it to a packing station — slowing or stopping safely if a human crosses its path.
L3: The unpredictable real world (Future): Robots handle soft, squishy, or delicate items. They can operate in completely messy, unsupervised environments. Most importantly, if a totally new problem arises, they can figure out a solution on its own without needing a human programmer to intervene.
- Example: A robot deployed in a hospital or eldercare facility. It can walk into a messy, unfamiliar bedroom, gently sort and fold a pile of soft laundry, and then safely hand a paper cup of water directly to a patient.
Figure AI robots sorting packages
Robots currently sits between L1 and L2: basic multimodal perception, execution, and reasoning. Locomotion has advanced significantly but manipulation is a different matter. Fine motor tasks that constitute most blue-collar work require environmental generalization that current models have not achieved.
As a current industry benchmark, UBTECH’s Walker series performs parcel-handling and logistics tasks at roughly 30% to 50% of a human worker’s efficiency.
The Engineering Bottleneck: What Delays Mass Adoption
Reaching that final tier of unpredictable real-world utility is currently constrained by several bottlenecks:
- Hands: The human hand has 27 degrees of freedom. The combination of precision, reliability, and cost has not been solved at production scale.
- Sensor: Six-axis force-torque sensors which give a robot the ability to feel force while handling objects are still expensive for mass deployment.
- Power: Typical continuous operation is 2–4 hours; swappable batteries help but current battery energy density limits heavy‑duty, continuous lifting.
- Cognition: Fragmented technological approaches (Vision‑Language‑Action vs. World Models) mean there is no unified, robust architecture for physical reasoning yet. In addition, training data remains insufficient and real-world, simulated, and video data are being combined to compensate this.
- Reliability: Current units need regular maintenance; industrial 24/7 uptime remains a target, not a reality.
The Accelerated Rollout: Mapping the Adoption Curve
Despite these hurdles, the market is not waiting for perfection.
While the market assumes widespread adoption is still many years away, recent industry targets point to a much faster timeline. We expect the rollout to happen in three phases:
2026–2027: Factory use and early home models
Carmakers move from testing to production (UBTech at BYD/Nio; Figure AI at BMW; Tesla using Optimus internally). Simultaneously, US firms like 1X and Tesla plan limited, expensive early consumer launches by 2027.
2028–2029: Warehouses and broader home appeal
Broad use in logistics and warehouses, contingent on robots running reliably for full eight-hour shifts. In the consumer space, prices dropping to the $12k–$18k range allows robots to transition from tech novelty to household purchase.
2030+: Mainstream public and care roles
Everyday use in unpredictable settings like retail, hospitals, and eldercare. This requires near-perfect safety and autonomous problem-solving capabilities.
The Tipping Point: When Humanoids Beat Human Wages
Financially, achieving mass commercial deployment hinges heavily on unit economics crossing the break-even threshold.
- Break‑even sketch: Assuming a Chinese manufacturing wage of USD 10k/year, 20% downtime and 20% annual maintenance, net annual savings per robot ≈ USD 6k. That implies a 2–3 year payback when purchase price is roughly USD 12–18k.
Product vs. economics: Unitree’s lineup illustrates the gap — entry and mid models may approach parts of the break‑even window, but true heavy‑duty models (H2/H1) remain materially more expensive. Until heavy‑duty humanoids fall into the USD ~12–18k range for single‑shift equivalence, mass replacement of human labour is unlikely.
The consumer crossover: For the average consumer, the break-even price range of a robot is comparable to an entry-level car, making it a financeable household expense. Once the hardware achieves the cognitive AI required to navigate unstructured domestic environments, humanoids will transition from industrial assets to consumer durables. This expands the Total Addressable Market (TAM) from factory labour to domestic services.
Navigating the Value Chain
As the sector moves from the lab to the factory floor, we assess beneficiaries along the value chain, each offering a distinct risk-reward profile:
Component Suppliers
Suppliers benefit from sector-wide volume growth regardless of which robot brand ultimately dominates the market. In addition, there is a systematic displacement of Japanese and European component monopolies across the humanoid supply chain.
However, near-term financials remain tied to legacy segments like industrials, as robotic revenue remains a fraction of total turnover.
Reducers: It converts high-speed motor rotation into the precise torque needed for robot joints. Harmonic Drive Systems (Japan) held the dominant position. Leaderdrive (688017 CH) now holds over 60% of the domestic market and ranks second globally. Shuanghuan Transmission (002472 CH) serves lower-body joints where higher torque is required.
Coreless motors: Previously Switzerland’s Maxon and Germany’s Faulhaber territory. Moons Industries (603728 CH) has broken through: 35% domestic market share in humanoid dexterous hand motors, pricing at one-third to one-half of Maxon’s equivalents. It has entered the supply chains of both Unitree and Tesla Optimus.
Planetary roller screws: Critical for knee and elbow joints — were dominated by European producers Rollvis and GSA, who held over 80% of global supply. Hengli Hydraulic (601100 CH) is now passing client validation at multiple humanoid integrators.
At system integration level: Sanhua Intelligent Controls (002050 CH) is the exclusive supplier of 14 rotary joint assemblies to Tesla Optimus, representing approximately RMB28,000 in hardware value per unit. Tuopu Group (601689 CH) functions as a Tier 0.5 supplier, producing integrated actuator assemblies rather than individual parts.
Robot Manufacturers
Direct exposure but heavy capex and margin pressure until scale is achieved.
UBTECH (9880.HK) is a listed pure-play manufacturer currently, with peers Unitree and AgiBot planning IPOs in 2026. In addition, Xiaomi (1810.HK) and Midea (330.HK) are expanding into the robotic market.
Downstream Users
Early adopters such as BYD (1211.HK), NIO (9866.HK), CATL (3750 HK) could drive margin expansion through 24/7 productivity and labour replacement. However, material financial impacts are likely years away.
This article is a “periodical publication” for information only and is not investment advice or a solicitation to buy or sell securities. This article does not constitute a “personal recommendation” or “investment advice” under UK FCA regulations. Investing in equities involves significant risk. The author holds NO position in the securities mentioned. There is no warranty as to completeness or correctness. Please do your own due diligence or consult a licensed financial adviser. Please read the Full Disclaimer before acting on any information. Images created with the assistance of Gemini AI.
Article Provided by Asia Pulse