Supply bottlenecks weigh on Orbbec's near-term outlook, but JPMorgan remains bullish on its 3D vision platform moat

Overview

The report updates forecasts based on Orbbec's 1H26 operating trends and management's post-results roadshow in Hong Kong. JPMorgan acknowledges that capacity constraints, new-product delays, and the exit from low-margin legacy businesses have caused a near-term earnings downgrade, but believes orders, the technology platform, customer coverage, and content value per unit have not weakened, and therefore maintains its positive long-term view and Overweight rating.

Core views

The core reason for the near-term forecast downgrade is supply and delivery rather than deteriorating demand. 2Q26 profit was below expectations due to factors including constrained capacity, delayed overseas expansion, and the company's proactive reduction of low-margin legacy businesses. Management lowered FY26 revenue guidance from more than Rmb1.6B to Rmb1.4B-1.5B; the overseas capacity ramp-up, originally planned for July, has been postponed to September or October, while shipment preparations for new products such as 3D scanning and physical AI data collection equipment have also progressed more slowly than expected, delaying some products by 1 to 2 months. The revised guidance implies approximately Rmb1B in sales for 2H26. Management expects sales and profit to reach a clear inflection point in 4Q26 as new production lines commence operations. The Guangdong base is ramping up, while the Vietnam factory is expected to begin operations between 4Q26 and 1Q27. The report believes humanoid robot vision and physical AI data collection are forming a mutually reinforcing growth flywheel. Orbbec launched a physical AI and robotics training data collection platform in July, targeting customers including humanoid robot OEMs, foundation model developers, and large technology companies; adoption of platforms such as EGO, UMI, and WristCamera is accelerating. The company had more than 1,200 robotics customers in 1H26, domestic embodied intelligence model companies have submitted urgent delivery requests, and the number of vision modules used in humanoid and high-end service robots is also increasing. Binocular modules are priced at approximately Rmb1,500-2,500 each, while wrist modules are priced at approximately Rmb1,800 each; more complex perception and manipulation requirements are increasing both the number of modules per robot and content value per unit. The report's logic is that greater robot deployment drives higher demand for high-quality training data, while the company's ability to provide both front-end data collection and back-end robotic perception solutions helps increase customer stickiness and strengthen its platform position. Business mix adjustment is another key driver of margin improvement. The company is shifting from the low-margin biometrics business toward the higher-margin AIoT business and reducing legacy businesses such as Touch and Pay. Gross margin reached 48.5% in 2Q26, up nearly 8 percentage points year over year. JPMorgan believes pricing power is jointly supported by an improved product mix, switching costs created once customers standardize on the platform, and integrations with NVIDIA Isaac/Jetson, Apple macOS, Microsoft, and Intel. Management expects to maintain a strong gross margin while expanding capacity and R&D investment and to improve operating leverage and reduce margin volatility as scale increases. Technology coverage and the competitive landscape constitute the platform moat. The company covers mainstream 3D vision technologies including structured light, binocular vision, ToF, and multi-sensor fusion, and continues to develop next-generation products integrating 2D, 2.5D, and LiDAR. According to management estimates, Orbbec holds approximately 95% of China's humanoid robot vision hardware market and faces virtually no substantive competition in its core 3D vision segment. The report notes that LiDAR vendors such as Hesai and RoboSense focus more on automobiles and long-distance navigation, with their functionality in humanoid robots primarily limited to obstacle avoidance and content value per unit typically only several hundred renminbi; RealSense is less competitive in terms of cost and integration. Meanwhile, the company's partnerships with new partners such as Robbyant, Linkerbot, and Honor are deepening, overseas revenue has nearly doubled year over year, and its customer base continues to expand. Forecast revisions are concentrated in FY26, while the long-term view remains largely unchanged. JPMorgan lowered FY26 adjusted EPS by 15.0% from Rmb0.77 to Rmb0.65 and FY27 adjusted EPS by 3.9% from Rmb1.49 to Rmb1.44; its model forecasts FY26, FY27, and FY28 revenue of Rmb1,347MM, Rmb2,506MM, and Rmb4,446MM, respectively, and adjusted net profit of Rmb268MM, Rmb590MM, and Rmb1,135MM, respectively. Management reiterated that its long-term 2027/28 revenue and profit targets remain on track, expects 2026 earnings to exceed Rmb200MM, and outlined a path toward Rmb800MM in earnings by 2030. JPMorgan states that its FY27/28 forecasts remain significantly above market consensus because it expects the company to fully capture the next phase of growth once capacity is released. On valuation, JPMorgan maintained its FY27 forecast P/S multiple of 23x, benchmarked against the 90th percentile of Chinese humanoid robot peers, and views Leader Drive as the closest comparable company. Due to the near-term model downgrade, the December 2027 target price was reduced by 7% from Rmb150 to Rmb140; based on the price of Rmb96.80 on August 25, 2026, the report indicates potential upside of approximately 45%. The institution maintains its Overweight rating and continues to identify Orbbec and Leader Drive as preferred names in the humanoid robot supply chain.

Analysis framework

The report first combines 1H26 operating data with management's Hong Kong roadshow to distinguish changes in demand from supply constraints. It then assesses the revenue trajectory based on orders, customer count, new-product deliveries, and capacity deployment, before updating earnings forecasts using product mix, module count, unit prices, gross margin, and operating leverage. Its analysis of long-term competitiveness focuses on technology coverage, ecosystem integration, customer switching costs, and differentiation from competitors, ultimately deriving the target price from forecast FY27 revenue and the P/S multiples of Chinese humanoid robot peers.

Methodology notes

• Industry/Sector Analysis FrameworkSupply-demand framework

  Distinguishing supply constraints from changes in demand

  The report attributes the FY26 downgrade to delays in capacity, training, and new-product deliveries, and uses evidence such as continued strong orders and customers' urgent delivery requirements to show that the issue is primarily on the supply side.

• Industry/Sector Analysis FrameworkVolume-price decomposition

  Analysis of module count per unit and module unit prices

  The report combines the increasing number of modules installed in each robot with the unit prices of binocular and wrist modules to explain how content value per unit drives revenue and gross margin.

• Competition and Strategy FrameworkMoat / competitive advantage

  Full-stack 3D vision platform and ecosystem stickiness

  The report assesses the company's competitive barriers based on its coverage of technology pathways, platform integration, rapid customization, customer switching costs, and management's estimated market share.

• Company Fundamentals and Financial FrameworkOperating/Financial Leverage Analysis

  Operating leverage from business mix and scale

  The report believes exiting low-margin businesses, increasing the share of higher-margin AIoT, and improving capacity utilization will drive margin improvement and reduce volatility.

• Valuation MethodPS valuation

  Valuation based on forecast FY27 P/S

  The target price is calculated using forecast FY27 revenue and a 23x P/S multiple, with the valuation multiple benchmarked against the 90th percentile of Chinese humanoid robot peers.

Asset mapping & comparison

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

• Orbbec - A (688322.SS)

  A direct participant in demand for humanoid robot 3D vision, physical AI data collection, and 3D scanning; the report maintains an Overweight rating.

  Strengths

  Full-stack 3D vision technology, management-estimated share of approximately 95% in China's humanoid robot vision hardware market, broad ecosystem integration, rising content value per unit, and high customer switching costs.

  Weaknesses

  Near-term capacity shortages, delays in overseas capacity expansion and new-product deliveries, and pressure on near-term revenue from exiting legacy low-margin businesses.

  Comparison

  The report considers Leader Drive the closest valuation comparable; compared with LiDAR vendors, Orbbec covers more 3D vision modalities and has higher content value per unit, while RealSense is weaker in terms of cost and integration.

  Risks

  Slower-than-expected growth in the humanoid robot, embodied intelligence, or 3D printing markets, intensified competition that reduces profitability, or a shift in vision hardware standards toward lower-priced product categories.

Key data

• FY26 revenue guidanceRmb1.4B-1.5BPreviously more than Rmb1.6B; the downgrade primarily reflects supply bottlenecks
• Implied 2H26 salesapproximately Rmb1BDerived from the revised full-year guidance
• Overseas capacity ramp-up timingSeptember or OctoberOriginally planned for July 2026
• New-product delays1 to 2 monthsInvolving products such as 3D scanning and physical AI data collection
• Vietnam factory commencement4Q26 to 1Q27To support overseas demand and supply chain resilience
• Number of robotics customers in 1H26more than 1,200Reflects expanding customer coverage in the robotics business
• Share of China's humanoid robot vision hardware marketapproximately 95%Management estimate
• Binocular module unit priceRmb1,500-2,500Used to measure vision content value per robot
• Wrist module unit priceRmb1,800/unitAn increase in the number of robot modules can raise content value per unit
• 2Q26 gross margin48.5%Up nearly 8 percentage points year over year
• Overseas revenuenearly doubled year over yearAdoption by both domestic and overseas customers is accelerating
• FY26 adjusted EPSRmb0.65Previously Rmb0.77, lowered by 15.0%
• FY27 adjusted EPSRmb1.44Previously Rmb1.49, lowered by 3.9%
• FY26/FY27/FY28 revenue forecastsRmb1,347MM/Rmb2,506MM/Rmb4,446MMCorresponding to year-over-year growth of 43.1%/86.1%/77.4%
• FY26/FY27/FY28 adjusted net profit forecastsRmb268MM/Rmb590MM/Rmb1,135MMThe model forecasts margin improvement alongside improvements in business mix and scale
• Target price valuation multipleFY27E P/S 23xBenchmarked against the 90th percentile of Chinese humanoid robot peers
• Target priceRmb140December 2027 target price, previously Rmb150, lowered by 7%

Impact & implications

The report believes the near-term revenue and profit trajectory will depend on capacity release and new-product deliveries, but strong orders imply that revenue could accelerate once capacity recovers. The combination of robotics vision and training data collection, higher content value per unit, a business mix shift toward higher-margin AIoT, and ecosystem platform standardization is expected to jointly enhance long-term revenue scalability, gross margin, and customer stickiness.

Risks

• Market growth in key applications such as humanoid robots, embodied intelligence, and 3D printing may be slower than expected.
• Intensifying competition may dilute the company's profitability.
• If vision hardware standards shift toward lower-priced product categories, content value per unit and growth potential may come under pressure.

What to watch

• Whether overseas capacity can begin ramping up in September or October 2026, and the pace of capacity release from the new Guangdong production lines.
• Whether the delayed 3D scanning and physical AI data collection products can be delivered as planned after delays of 1 to 2 months.
• Whether sales and profit can reach the clear inflection point expected by management in 4Q26 as new production lines commence operations.
• Whether the Vietnam factory can begin operations between 4Q26 and 1Q27.
• Trends in robotics orders, the number of modules per unit, data collection platform adoption, and the sustainability of the 48.5% gross margin.