On October 13, 2026, Bloomberg published a feature under the headline "China Is Poised to Spearhead a $5 Trillion Humanoid Robot Market." The number came from Morgan Stanley's projection for 2050 — not for this year, not for this decade. It includes the value of services and supply chains that do not yet exist. Barclays, in a separate estimate cited by Bloomberg Television, put the current humanoid robot market at $2 billion to $3 billion.
The gap between $3 billion and $5 trillion describes how far the industry has to travel. It also explains why the component that has received the least attention — the hand — has become the obstacle that determines whether the journey happens at all.
Bloomberg's feature made the point directly: despite decades of innovation, robotic hands "still fumble at shoelaces and keys," and that lack of precision is one of the main reasons the parade of spectacular robotic feats — choreographed kung fu, record-breaking 100-meter sprints — "has yet to translate into meaningful commercial deployment."
The Gap Between Shipments and Work
The headline adoption numbers look like a boom.
Smart Analytics Global estimates that global humanoid robot shipments reached about 19,100 units in the first half of 2026, more than triple the 5,100 units shipped in the same period of 2025. Chinese buyers accounted for more than 85 percent of demand. Another tally puts first-half shipments at over 22,000 units, with Chinese manufacturers taking the top five positions and a combined 86 percent share; Zhiyuan Robotics led with about 9,700 units, followed by Unitree at over 7,000.
Then the numbers get less impressive.
Of those 22,000-plus units, fewer than 18 percent went onto factory floors and into production lines. More than 60 percent went to commercial performances, research and teaching, and scenic-area tour guides. Alex Zhou, founder of LinkerBot, estimated that of the roughly 20,000 humanoid robots installed globally last year, "the number of actual working machines is in the low thousands. The rest are mainly for entertainment."
The shipment figures also include rental arrangements and government pilot programs, and many robots still depend on human supervision or pre-scripted control when performing complex tasks. At Zhaofeng Electromechanical's factory, 15 Leju robots ran continuously for over two months and about 1,000 hours, assisting in the production of more than 60,000 components, with a loading cycle stabilized at 40 seconds and a success rate of 98 percent. The detail that matters: a human doing the same loading task takes an average of 15 seconds, and the factory still needs one engineer to supervise roughly four robots.
The industry's own order book is thinner than the shipment data suggests. Public information indicates that a substantial share of humanoid robot orders in the market are non-substantive, with the real buyers being universities, vocational colleges, and local government training centers. Unitree founder Wang Xingxing has acknowledged that the company's production-line projects with automakers have not been widely replicated because the robots' overall efficiency and generalization capability are not yet sufficient.
Why the Hand Is the Bottleneck
The hand is not one component among many. It is the component that determines what a humanoid robot can actually do, and it is the most expensive single part of the machine.
Morgan Stanley's teardown of Tesla's Optimus found that the dexterous hand accounts for 17.2 percent of total unit cost — the most expensive single component. A separate supply-chain breakdown puts the hand at about 20 percent of the bill of materials, and over 30 percent of the cost within the motion system. Another analysis places it at roughly 14 percent of total unit cost, with the drive system — micro motors accounting for 38 percent of hand BOM cost — as the largest sub-component, followed by tactile sensing at over 25 percent.
The variance in those figures reflects different teardown assumptions, not different physics. The physics are the problem.
A human hand has 27 bones and 27 joints, an opposable thumb, and the ability to grasp, clench, and twist with a combination of strength and delicacy that no robotic replica has matched. Tesla's Optimus Gen 3 hand has 22 degrees of freedom, only five fewer than a human hand. Fitting 25 linear actuators into the forearm, driving the fingers through tendon cables, creates a control problem that compounds with every additional joint: tendon elasticity, friction, and coupling between multiple tendons all interact, and the hand has to move without tremor or lag.
The manufacturing problem compounds the control problem. Micro transmission systems with minimum diameters of 3.4 millimeters, sensors thinner than a ballpoint pen refill, micro motors, micro ball screws, tendons, and precision bearings all have to be packed into the volume of a human hand. Assembly hours and yield rates do not improve automatically with volume. Researchers at Physical Intelligence have argued that the dexterous hand may be a category where scale does not dilute cost — unlike solar panels or power batteries, where larger volumes produce predictable cost declines.
That is the "impossible triangle" of dexterous hands: high performance pushes cost up, low cost produces a hand that can barely grasp, and high reliability requires design simplification that sacrifices performance. All three at once is a physical constraint, not a design choice.
Tesla's experience illustrates the cost of that constraint. Optimus was originally planned for 5,000 units in 2025; the company produced only a few hundred. Hand and arm problems were the primary cause, and Tesla has set volume production at Giga Texas for the third quarter of 2027. Elon Musk said in October 2025 that manufacturing the forearm and hand was "more difficult than all the other parts of the robot combined."
The Cost Curve Is Bending
The hand is also the component where Chinese supply chains have moved fastest.
Two years ago, a dexterous hand averaged about 50,000 yuan per unit. By 2026, the average had fallen to below 10,000 yuan, with six-degree-of-freedom hands priced at 6,000 to 8,000 yuan. A separate industry estimate puts the current range at 15,000 to 20,000 yuan per hand, down from 50,000 yuan, with the domestic localization rate rising from 50 percent in 2024 to 75 percent by the end of 2025. Core components — micro motors, ball screws, tendons — have fallen to one-fifth to one-tenth of overseas prices.
The volume trajectory is steep. China's dexterous hand market sold about 19,200 units in 2025 and is projected to reach 70,200 units in 2026, 200,000 in 2028, and over 430,000 by 2030 — a compound annual growth rate of 87 percent.
Capital has followed the cost curve. Financing in China's dexterous hand sector exceeded 25 billion yuan in the first half of 2026, more than the whole of 2025. LinkerBot closed a 1.5 billion yuan Series B in February 2026 and has sold more than 10,000 bionic hands, targeting 50,000 to 100,000 units of delivery in 2026. Another estimate puts the sector's cumulative financing at over 8 billion yuan by the end of July 2026, a 1,500 percent increase over the same period of 2025.
The cost decline is real. The question is whether it is enough.
A hand that costs 6,000 to 8,000 yuan is affordable for a research lab or a demonstration unit. A humanoid robot that can work in a factory needs more than an affordable hand. It needs a hand that can grip a part with enough force to hold it, enough delicacy not to crush it, enough tactile feedback to know the difference, and enough reliability to do it 10,000 times without failing. The low-cost hands on the market today are optimized for a different set of requirements. Some of the cheapest humanoid robots have abandoned the general-purpose dexterous hand entirely — Songyan Dynamics' Bumi kept only a thumb and a simple two-finger structure, cut the lidar and depth-vision modules, and is positioned by the manufacturer for education and entertainment rather than domestic or industrial work.
The affordable hand and the capable hand are not yet the same hand.
What the Numbers Actually Support
The $5 trillion figure is a 2050 projection that includes services and supply chains. It is not a forecast of robot sales. The current market is measured in billions, not trillions. The shipment numbers are growing quickly from a small base, and most of the units shipped are not doing productive work.
The dexterous hand is the component that connects those two facts. It is the most expensive single part of the robot, the hardest to manufacture at scale, and the one where the gap between demonstration and deployment is widest. A robot that can kick a car door and backflip off a platform — as Unitree's newly released H2-X was demonstrated doing — is a robot whose hand is not the limiting factor. A robot that can tie a shoelace is a different achievement entirely.
The cost curve is bending in the right direction. The capability curve is bending more slowly. Until the two curves intersect — until an affordable hand is also a capable hand — the shipment numbers will keep growing faster than the work numbers, and the $5 trillion projection will remain a statement about 2050 rather than a description of 2026.
The bet is in the hands. The hands are not ready yet.
Sources: Bloomberg "China Is Poised to Spearhead a $5 Trillion Humanoid Robot Market" (October 13, 2026); Bloomberg Television via Perplexity summary (October 14, 2026); Barclays research via Chinastarmarket (2026) and EET-China (July 19, 2026); Smart Analytics Global via Perplexity summary (October 14, 2026); Counterpoint Research via Sgpjbg (October 19, 2026); 2026 Humanoid Robot Industry Development Report via Sgpjbg (October 19, 2026); OFweek supply-chain teardown (May 22, 2026); Sina Finance dexterous hand industry research (October 13, 2026); TMTPost "Humanoid Robots Are Still Learning to Work; Dexterous Hands Became the Hot Business First" (June 23, 2026); ReadAI / IT Juzi dexterous hand financing data (August 11, 2026); Yicai "Dexterous Hands Fall From 50,000 Yuan to Under 10,000 Yuan" (August 18, 2026); Lexotech / Toutiao "Humanoid Robots Shipped 22,000 Units in Six Months; Why Over 80 Percent Never Entered a Factory" (October 22, 2026); Unitree H2-X launch coverage via Jiemian (October 21, 2026) and PConline (October 21, 2026).
Disclaimer
The information provided in this article is for general informational and educational purposes only. It does not constitute legal, financial, or professional advice. The author and publisher are not responsible for any actions taken based on the content of this article. Readers should consult qualified professionals for advice specific to their situation. All trademarks and references to third-party products, services, or organizations are the property of their respective owners. The performance data and benchmarks discussed are based on specific research studies and may not generalize to all use cases or environments. As of the publication date, the AI landscape continues to evolve rapidly, and readers should verify current information independently.
Limitations
This analysis is based on Bloomberg reporting, third-party shipment estimates, and supply-chain research available as of the article date.
Shipment totals include rental arrangements, government pilots, and demonstration units, so unit counts overstate productive deployment.
Dexterous-hand cost ranges come from different research methodologies and are not like-for-like.
The $5 trillion figure is a 2050 projection that includes services and supply chains that do not yet exist.
Reporting standards differ across jurisdictions; figures may not be directly comparable.
Market size and contract value figures come from third-party research and are estimates.
This sector is changing quickly; claims can become outdated soon.
Sources
- Bloomberg "China Is Poised to Spearhead a $5 Trillion Humanoid Robot Market" (October 13, 2026)
- Bloomberg Television via Perplexity summary (October 14, 2026)
- Barclays research via Chinastarmarket (2026) and EET-China (July 19, 2026)
- Smart Analytics Global via Perplexity summary (October 14, 2026)
- Counterpoint Research via Sgpjbg (October 19, 2026)
- 2026 Humanoid Robot Industry Development Report via Sgpjbg (October 19, 2026)
- OFweek supply-chain teardown (May 22, 2026)
- Sina Finance dexterous hand industry research (October 13, 2026)
- TMTPost "Humanoid Robots Are Still Learning to Work
- Dexterous Hands Became the Hot Business First" (June 23, 2026)
- ReadAI / IT Juzi dexterous hand financing data (August 11, 2026)
- Yicai "Dexterous Hands Fall From 50,000 Yuan to Under 10,000 Yuan" (August 18, 2026)
- Lexotech / Toutiao "Humanoid Robots Shipped 22,000 Units in Six Months
- Why Over 80 Percent Never Entered a Factory" (October 22, 2026)
- Unitree H2-X launch coverage via Jiemian (October 21, 2026) and PConline (October 21, 2026).
The information provided in this article is for general informational and educational purposes only. It does not constitute legal, financial, or professional advice. The author and publisher are not responsible for any actions taken based on the content of this article. Readers should consult qualified professionals for advice specific to their situation. All trademarks and references to third-party products, services, or organizations are the property of their respective owners. The performance data and benchmarks discussed are based on specific research studies and may not generalize to all use cases or environments. As of the publication date, the AI landscape continues to evolve rapidly, and readers should verify current information independently.
Limitations: This analysis is based on Bloomberg reporting, third-party shipment estimates, and supply-chain research available as of the article date.; Shipment totals include rental arrangements, government pilots, and demonstration units, so unit counts overstate productive deployment.; Dexterous-hand cost ranges come from different research methodologies and are not like-for-like.; The $5 trillion figure is a 2050 projection that includes services and supply chains that do not yet exist.; Reporting standards differ across jurisdictions; figures may not be directly comparable.; Market size and contract value figures come from third-party research and are estimates.; This sector is changing quickly; claims can become outdated soon.