China robot sector and embodied AI robotics: AI-enabled traditional robotics should advance before humanoids reach mass adoption

Overview

This China robotics sector report assesses how embodied AI changes demand across humanoid and established robot categories. UBS sees faster-than-expected early humanoid adoption but believes AI-enabled traditional robots and upstream component suppliers offer clearer three- to five-year earnings exposure.

Core views

UBS raises its 2026-30 global humanoid-demand forecast by 73-123%, driven by stronger-than-expected demand from research institutes, data-collection centres and commercial applications with relatively simple tasks. Global humanoid sales rose to 18,315 units in 2025 from around 2,200 in 2024; Unitree and AgiBot accounted for about 30% and 28% of global shipments, respectively. UBS estimates that commercial applications represented 52% of 2025 sales, with research institutes and data-collection centres contributing 27% and 18%. In China, Interact Analysis identifies 64 operating data-collection centres and 26 under construction across 23 provinces. UBS nevertheless lowers its short-term expectations for industrial and household humanoid uses because testing cycles are longer and generalisation remains inadequate. The report forecasts global humanoid demand of 305,858 units in 2030, from 51,899 in 2026, and expects the global humanoid market to expand from Rmb9.2bn in 2025 to Rmb90bn in 2030. Its forecast retains a 10% 2025-30 CAGR decline in humanoid ASPs, with faster volume growth driving market expansion. Longer term, UBS raises its 2050 demand forecast by 5% to 90m units and its 2050 market-size estimate to US$1,765bn, including rental and after-sales services. The higher long-run outlook is based mainly on greater commercial penetration initially, followed by industrial and household catch-up if generalisation improves. UBS argues that humanoids remain technologically immature because effective generalisation requires more capable AI models, much more real-world training data, reliable real-time multimodal coordination and low error rates in physical tasks. The largest robotics datasets are only at the million-trajectory scale, versus 36 trillion text tokens for the largest text datasets. The report highlights continuing sim-to-real gaps, high cost of manipulation-data collection, uncertainty over appropriate data mix and quality, latency trade-offs, and the fact that physical errors can be irreversible or compound across industrial workflows. Model development has progressed from task-specific robot transformers toward vision-language-action models and world models, but UBS says the industry has yet to converge on the ultimate model architecture or training-data mix. These constraints underpin its view that a humanoid-sector "EV moment" will not arrive before 2030. AI should nevertheless accelerate adoption in traditional robotics by improving perception, planning, ease of programming and human-machine interaction. UBS forecasts 2025-30 China unit-demand CAGRs of 10% for industrial robots, 25% for collaborative robots, 15% for mobile robots, 20% for service robots and 60% for quadrupeds. AI contributes roughly 3 percentage points to industrial and mobile-robot demand CAGRs, about 10 points to cobots and service robots, and about 35 points to quadrupeds. Industrial robots should remain China’s largest non-humanoid market in 2030 at Rmb69bn, followed by mobile robots at Rmb35bn, cobots at Rmb7bn, quadrupeds at Rmb5.3bn and service robots at Rmb3.7bn. UBS therefore expects traditional robotics to dominate sector revenue in the near to medium term. For positioning, UBS prefers upstream components over OEMs because component suppliers can serve several robot categories and OEM customers, potentially allowing earlier revenue scale and breakeven. It estimates China’s key robotics component TAM at Rmb125bn in 2030. Perception systems and reducers are expected to be the largest markets at Rmb26bn and Rmb25bn; sensors account for about two-thirds of perception-system value, or Rmb17bn and 14% of total bill of materials. Screws have a smaller Rmb5bn TAM but are especially leveraged to rising humanoid penetration, while chips could reach Rmb15bn and around 12% of bill of materials by 2030 as higher-tier processors gain adoption. UBS is more cautious on Chinese robot OEMs, citing intense competition, lower in-house manufacturing capability, capital requirements for hardware and AI-brain development, and uncertainty over which companies can combine technology, applications, capital strength and governance successfully. It favors established traditional OEMs over pure-play or transitioning humanoid OEMs in the short to medium term because of better earnings visibility. Sector and humanoid indices have underperformed the machinery index year to date, which UBS attributes to delayed commercialization, comparatively high valuations and investor preference for AI segments with clearer earnings and order visibility. UBS sees sentiment improving only with clearer Tesla Optimus V3 mass-production milestones, tangible AI-generalisation progress, or significant hardware advances. Among covered names, UBS initiates Unitree with a Neutral rating, viewing its hardware and motion-control leadership as supportive of early commercial, research and data-collection demand but considering its valuation fair at 43x 2027E P/S and 225x 2027E PE. UBS maintains Buy ratings on OPT, Kedali and Hengli Hydraulic. It highlights OPT’s machine-vision exposure, industrial-AI platform and expansion into semiconductors, AI servers, optical modules, autos and robotics; it expects Hengli’s established hydraulic operations to provide earnings visibility while ball and roller screws add growth options. UBS maintains a Sell rating on Leader Drive despite raising its 2026-28E net-profit forecasts by 16-59% and its price target from Rmb52.9 per share to Rmb216 per share, as it considers 238x forward PE demanding after a more-than-50% year-to-date rally and expects capacity expansion to pressure margins.

Analysis framework

UBS combines shipment, application and market-size forecasts with an assessment of AI-model development and robotics deployment constraints. It compares robotics categories by locomotion, manipulation and generalisation capabilities; quantifies AI-related demand uplift and component addressable markets; then evaluates OEMs and suppliers by technology, applications, capital, governance, profitability and valuation.

Methodology notes

• Industry AnalysisSupply-demand framework

  Demand forecasting by robot category, downstream application and component market.

  UBS estimates shipment growth, ASP trends and addressable market size across humanoid and traditional robots, then links demand growth to component content.

• Industry AnalysisUpstream-Midstream-Downstream Transmission

  Robotics supply-chain analysis from components through OEMs to downstream uses.

  The report argues that broader component exposure allows suppliers to benefit earlier and across more robot types than individual OEMs.

• Valuation methodsSOTP (Sum-of-the-Parts) Valuation

  Sum-of-the-parts valuation for Leader Drive.

  UBS changed Leader Drive’s valuation approach from PE to SOTP when raising its target price, separating the value of its businesses rather than applying one earnings multiple.

Asset mapping & comparison

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

• OPT (688686.SS)

  Preferred machine-vision and sensor supplier benefiting from industrial AI and diversified downstream demand.

  Strengths

  Exposure to Apple’s supply chain, a full-stack industrial-AI platform, and expansion into semiconductor, AI-server, optical-module, automotive and robotics applications.

  Comparison

  UBS identifies OPT as one of its top component picks in China.

• Hengli Hydraulic (601100.SS)

  Preferred roller-screw supplier with traditional hydraulic earnings support.

  Strengths

  Construction-machinery recovery, market-share gains, overseas capacity ramp-up, mass supply of machine-tool ball screws and initial roller-screw production.

  Comparison

  UBS identifies Hengli Hydraulic as one of its top component picks in China.

• Kedali (002850.SZ)

  Preferred component supplier exposed to reducers, roller screws and dexterous-hand components.

  Strengths

  Robust non-robotics business outlook in UBS’s thesis table.

  Comparison

  UBS identifies Kedali as one of its top component picks in China.

• Unitree (688836.SS)

  Covered humanoid and quadruped OEM.

  Strengths

  Hardware and motion-control leadership; 5,215 humanoid and 23,037 quadruped robots sold in 2025.

  Weaknesses

  Humanoid generalisation remains an industry-wide constraint and valuation appears fair to UBS.

  Comparison

  Neutral-rated, unlike UBS’s preferred component suppliers.

  Risks

  Order growth, AI-model progress and margins.

• Leader Drive (688017.SS)

  Covered reducer supplier with embodied-AI exposure.

  Strengths

  Price-performance, responsive service, R&D capability and harmonic-reducer capacity expansion.

  Weaknesses

  UBS considers valuation demanding at 238x forward PE and expects capacity expansion to pressure margins.

  Comparison

  Least favored in UBS’s thesis map.

  Risks

  Weaker traditional-robot demand, slower humanoid mass production, and delayed capacity, product or customer ramp-ups.

Key data

• Global humanoid shipments18,315 units in 2025Up from around 2,200 units in 2024.
• Global humanoid demand forecast revision+73% to +123% for 2026-30Driven by research, data-collection and commercial applications.
• Global humanoid market sizeRmb90bn in 2030Versus Rmb9.2bn in 2025.
• Global humanoid demand305,858 units in 2030UBS forecasts a 76% 2025-30 CAGR.
• China robotics component TAMRmb125bn in 2030Perception systems and reducers are estimated at Rmb26bn and Rmb25bn.
• Long-term humanoid demand90m units in 2050A 5% increase from UBS’s prior forecast.

Impact & implications

UBS believes near- and medium-term robotics growth should be led by AI-enabled traditional robots and component suppliers with broad OEM exposure, technical barriers and resilient core businesses. Humanoid OEMs retain greater long-term optionality, but their commercial outcome depends on resolving AI-generalisation and deployment bottlenecks.

Risks

• Downstream manufacturers may expand capacity less than expected.
• Domestic and overseas renewable-energy demand, including lithium-battery and solar demand, may be weaker than expected.
• Competition may intensify and product ASPs may decline sharply.
• Humanoid deployment may be delayed by inadequate AI generalisation, limited data, reliability issues and high data-collection costs.
• Leader Drive faces risks from weaker traditional-robot demand, slower humanoid mass production and delayed capacity, product or customer ramp-ups.

What to watch

• Progress toward mass production of Tesla Optimus V3, including the stated end-2026 target.
• Tangible improvement in humanoid AI-model generalisation and robotics data collection.
• Hardware advances demonstrated at industry conferences.
• Evidence of genuine commercial orders for robots.
• Third-quarter 2026 results from A-share humanoid-related companies, including Unitree.