Produced by | He Xi

Layout by | Ye Yuan

Public data shows that the material cost of building a humanoid robot starts at a minimum of $46,000. However, if you exclude the Chinese supply chain, that figure jumps to $131,000—enough of a difference to buy two more robots.

By 2026, humanoid robots may not yet be mainstream, but Chinese companies selling joints, dexterous hands, and sensors are already quietly amassing billion-dollar valuations by capitalizing on this cost gap.

An Optimus robot priced at $20,000 sees nearly half its cost go to actuators and joint modules; a single dexterous hand accounts for one-fifth of the total BOM cost. While full-system manufacturers are still ramping up for mass production, upstream component suppliers are already lining up with orders to enter the capital markets.

This article breaks down the BOM cost structure of humanoid robots, scans the three core segments—dexterous hands, joint modules, and sensors—and provides industry practitioners, supply chain managers, and tech decision-makers with a referenceable 'industry landscape.'

01 Starting with the BOM Cost Structure of Humanoid Robots

To identify who holds the core position in the supply chain, we must first understand where the money is being spent.

According to Tesla's 2025 disclosures, the BOM cost target for an Optimus robot is to drop from around $43,000 in small-batch production to under $20,000 in mass production. How is that $20,000 allocated? The Paper, based on Tesla's supply chain information, breaks down the costs into five categories:

Actuators: ~35%. Includes rotary and linear actuators, with ~28 per unit. This is the highest value-density component.

Lead Screws: ~18%. Primarily reverse planetary roller screws, with ~14 per unit, representing the highest manufacturing barrier.

Reducers: ~15%. Harmonic and planetary reducers, core components of rotary joints.

Motors: ~12%. Includes frameless torque motors and coreless motors, with ~20 coreless motors required per dexterous hand.

Sensors and Others: ~20%. Includes six-axis force sensors, tactile sensors, vision systems, etc., with force sensors seeing the fastest growth.

Further industry research by Xinchacha points out that actuators and joint modules combined account for 45%–65% of a humanoid robot's total cost, widely recognized as one of the highest value-density segments in the supply chain. Zheshang Securities calculates that linear actuators (including planetary roller screws and T-type screws) alone account for 19.4% of the value.

Consider a specific example: A teardown of Unitree's G1 base model revealed a total BOM cost of approximately ¥41,600, with core joint costs accounting for ¥27,500. Based on Unitree's 2025 shipment target of over 5,500 humanoid robots, joints alone will generate over ¥150 million in upstream procurement.

Most of this money flows to Chinese suppliers.

The Paper calculated: At the current stage in 2025, relying on the Chinese supply chain, the material cost per humanoid robot is approximately $46,000; without it, costs surge to $131,000—nearly triple the difference. China accounts for 55% of global robot production capacity and became a net exporter of industrial robots for the first time in 2025.

Without the Chinese supply chain, mass production of humanoid robots would remain just talk. Meanwhile, Chinese supply chain players are converting this advantage into technological accumulation and market share.

02 Upstream Components 'Industry Map' — A Panoramic Scan of Three Core Segments

Now that we know where the money is going, which companies hold key positions in these segments?

Dexterous Hands: The 'Last Centimeter' Battle for Billion-Dollar Valuations

The dexterous hand is the closest part of the robot to 'humanness' and one of the most technically challenging and high-value components. It accounts for ~20% of the total robot cost and over 30% of the motion system's hand-related costs.

Two representative companies have emerged in this segment:

Agilink is worth watching. Spun off from Zhiyuan Robotics in January 2026, it completed four funding rounds in under five months, reaching a valuation exceeding $1 billion. Since inception, it has delivered over 8,000 dexterous hands and tens of thousands of grippers. With a high technical starting point—derived from Tencent Robotics X Lab and Zhiyuan's dexterous hand team—it is backed by Tencent, Baidu, Hillhouse, BlueRun, and others, while leveraging Zhiyuan's '1+5+N' ecosystem for orders and scenarios.

AgileHand leads in shipment volume. It completed a nearly ¥1.5 billion Series B funding round in February, reaching a post-money valuation of over ¥10 billion, and is seeking a new round at ~¥20 billion. With monthly shipments exceeding 4,000 units, it holds over 80% of the global high-DOF dexterous hand market. Clients include leading full-system manufacturers like UBTECH and Xiaomi.

Additionally, Suzhou's ZhiXing Robotics has raised nearly ¥100 million in funding and entered the supply chains of Zhiyuan and Huawei. Zhaowei Mechanical & Electrical plans to invest ¥800 million to build a dexterous hand industrial park, accelerating its A+H listing process.

Industry Judgment: The technical routes for dexterous hands (linkage, cable-driven, direct-drive) have not yet converged, but manufacturing capabilities are already diverging. Top players are shipping thousands of units monthly, with orders shifting from 'prototypes' to 'bulk.' Whoever achieves scaling (scale) first will likely dominate standard-setting.

Joint Modules/Actuators: The 'Cash Pool' with the Highest Cost Share and Certainty

If dexterous hands are the 'crown jewel,' joint modules are the robot's 'muscle group'—accounting for the highest cost share and boasting the most mature supply chain.

Below is the current high-certainty Chinese supply chain landscape:

Actuator Assembly Players:

Top Group: Optimus Tier 0.5 supplier, with Tesla business accounting for 35%–40% of total revenue. Its Mexico factory targets 300,000 units in Q2 2026.

Sanhua Intelligent Controls: Core supplier of joint module assemblies, with $685 million in orders (~¥5 billion), and Mexico factory deliveries starting in 2026.

Precision Transmission Players:

Leader Drive: Harmonic reducer leader, with >60% domestic and >35% global market share, exclusively supplying Tesla's Mexico factory. By Q1 2025, humanoid robot-related revenue accounted for 30% of its total. Reducer prices have dropped from thousands of yuan to ¥1,300, half of Japan's Harmonic Drive.

Wuzhou Xinchun: Primary supplier of reverse planetary roller screws for high-torque applications, with Thailand factory production starting in 2026.

Shuanghuan Drive: Core RV reducer supplier, entering the Optimus supply chain indirectly through Top Group and Sanhua.

Xinjian Transmission: Unlisted but a core supplier of screws/reducers.

Joint Module Integration:

Quanzhibo: A leading joint module player, with shipments exceeding 100,000 units in 2025.

Industry Judgment: China's supply chain is deeply tied to Tesla and other leading full-system manufacturers, with rapid import substitution. Costs for core components like reducers and screws have dropped to globally low levels and continue to decline. Companies in this segment profit from 'economies of scale'—the more volume, the lower the cost, and the higher the barriers.

Sensors: The 'Winner-Takes-All' Effect of Six-Axis Force Sensors

Sensors are the robot's 'nervous system,' with six-axis force sensors being a high-barrier, high-value category. This segment shows clear head concentration (head concentration) trends:

Blue Point Touch: Completed over ¥100 million in Series C+ funding from investors including CATL's Puquan Capital, Zhiyuan Robotics, Galaxy General, and OptoTech. It held ~62% of China's 2024 humanoid robot six-axis force sensor market, rising to 72.6% in Q1–Q3 2025. By 2026, it expects to ship ~400,000 joint force sensors and 30,000–40,000 six-axis force sensors, with bulk orders covering leading companies like Zhiyuan, Xiaomi, XPeng, Galaxy General, Q-Sling, UBTECH, and Xinghai Tu.

Kunwei Technology: Completed ¥100 million in Series B+ funding, originating from aerospace research institutes. It holds >50% of the domestic six-axis force market and led the drafting of the national standard GB/T43199-2023 'Robot Multidimensional Force/Torque Sensor Detection Specifications,' filling a domestic standard gap. It supplies UBTECH, Galaxy General, Zhiyuan, and the Beijing Humanoid Robot Innovation Center.

Industry Judgment: The six-axis force sensor market has formed a dual-leader pattern (landscape), with Chinese companies dominating global standards. Notably, the phenomenon of 'customers becoming shareholders' is pronounced here—Blue Point Touch's investors include CATL, Zhiyuan, and Galaxy General, who are both clients and shareholders. This alignment reduces commercialization uncertainty for upstream players.

Embodied Brains: For Comparison

While upstream components are rapidly developing, the 'brain' segment remains an industry focal point. Galaxy General Robotics raised ¥2.5 billion at a post-money valuation exceeding ¥20 billion; Qianxun Intelligence secured nearly ¥2 billion across two rounds, breaking ¥10 billion; Zibianliang Robotics landed a ¥1 billion Series A++ round led by ByteDance and Sequoia China; and Zhipingfang secured over ¥1 billion in Series B, breaking ¥10 billion.

In comparison, the valuation gap between upstream component unicorns (Agilink at ¥7 billion, AgileHand at ¥10+ billion, Blue Point Touch, etc.) and full-system brain unicorns is narrowing. This indicates that industrial capital's attention to the upstream segment is now on par with full-system manufacturers, with upstream emerging as an independent and critical industry pole.

03 The Underlying Logic of Upstream Scale-Up

Why are upstream components scaling first? Three fundamental reasons explain this:

Logic 1: Full-system forms are unconverged, but component demand is rigid

The final form of humanoid robots—bipedal or wheeled, degree of freedom, type of dexterous hand—remains undecided. However, every full-system robot needs joints, dexterous hands, and sensors regardless of the winner. Upstream segments avoid the risk of 'betting on a single technical route' and profit from selling 'shovels' rather than 'digging for gold.'

Logic 2: China's supply chain cost advantage is converting into global orders

A threefold cost difference is no small matter. Global leading full-system manufacturers—Tesla, Unitree, Figure, Agility—are actively onboarding the Chinese supply chain. China accounts for 55% of global robot production capacity and became a net exporter of industrial robots in 2025. While tariffs and geopolitics may redirect some orders, a threefold cost gap makes rational business choices clear.

Logic 3: Industrial 'customers becoming shareholders' accelerates upstream validation and binding

Take Blue Point Touch as an example: CATL, Zhiyuan, and Galaxy General are both clients and shareholders. This alignment grants upstream players an 'entry ticket' during product validation, significantly reducing commercialization uncertainty. According to Jiemian News, one investor analyzed, 'From a commercial predictability standpoint, the upstream segment may even be stronger than the full-system side.'

Epilogue

In 2026, the most noteworthy players in the humanoid robot industry may no longer be the full-system manufacturers but those selling hands, joints, and sensors.

Upstream components are no longer mere 'appendages' of full-system robots but independent 'protagonists' with their own technological logic and industry narratives. With AgileHand shipping 4,000 dexterous hands monthly, Agilink reaching a ¥7 billion valuation in five months, and Blue Point Touch securing collective investments from clients, a clear signal has emerged: The upstream supply chain is evolving from a 'shovel seller' in the 'gold rush' into a new industrial powerhouse.

For industry practitioners, supply chain managers, and tech decision-makers, this 'industry landscape' may serve as a reference point for entering the track (segment).

As one industry expert who has long tracked this field put it: 'The upstream segment may be even stronger than the full-system side.'

PS: If you know of outstanding startups in the embodied AI space, feel free to send a private message with recommendations.