Humanoid robots can already walk.
Some can run.
Others can dance, perform backflips, carry boxes, and navigate factories.
But ask one to thread a needle, peel an apple, tighten a tiny screw, or manipulate an object as naturally as a person and the problem becomes much harder.
That is why robotic hands are emerging as one of the biggest bottlenecks in humanoid robotics.
Elon Musk has publicly acknowledged the challenge. Tesla’s Optimus program has made major progress with walking and whole-body movement, but reproducing the extraordinary dexterity of the human hand remains much more difficult.
A Beijing startup called Linkerbot thinks it can help solve that problem.
Founded in 2023 and led by CEO Alex Zhou, Linkerbot builds five-fingered robotic hands designed for humanoid robots, industrial automation, research, and eventually household tasks. The company closed a Series B+ funding round in April 2026 that valued it at about $3 billion, and it is already pursuing another round targeting a valuation as high as $6 billion.
What makes investors particularly interested is not simply that Linkerbot can build sophisticated robotic hands.
It is trying to make them cheap enough to manufacture at enormous scale.
And if humanoid robots really are going to move from technology demonstrations into factories and homes, that may be just as important as the technology itself.
Why Are Robotic Hands So Difficult to Build?
Look at your own hand for a moment.
Picking up a coffee mug seems effortless.
But your hand is simultaneously calculating:
Finger position.
Grip force.
Object shape.
Surface friction.
Weight.
Balance.
Movement.
Feedback from thousands of sensory receptors.
All of this happens almost automatically.
A robot has to reproduce those capabilities mechanically and computationally.
That means engineers need tiny motors or tendon systems to move the fingers, sensors to measure pressure, software to coordinate movement, and increasingly AI to decide exactly how the fingers should respond.
The Guardian recently described dexterous hands as perhaps the final major obstacle preventing humanoid robots from moving beyond impressive demonstrations and becoming genuinely useful machines.
That is why a robot that can perform a backflip may still struggle to button a shirt.
Large movements can actually be easier than delicate ones.
Elon Musk Has Acknowledged the Hand Problem
Tesla has been developing its Optimus humanoid robot with the long-term goal of performing useful work in factories and eventually homes.
Walking has improved substantially.
Manipulation remains harder.
Forbes notes that Musk told the All-In Summit in Los Angeles that Tesla had made progress with Optimus mobility, but creating hands with human-like hardware and control remained one of the more difficult engineering problems.
That makes sense when the ultimate goal is considered.
A household robot does not become particularly valuable simply because it can walk through a kitchen.
It has to:
Open a cupboard.
Pick up a glass.
Hold a knife.
Fold clothing.
Turn a doorknob.
Use tools.
Wash dishes.
Handle fragile objects.
Those activities are overwhelmingly hand problems.
Linkerbot is betting that whoever solves dexterous manipulation at scale could become one of the most valuable suppliers in the entire humanoid-robot industry.
Linkerbot Wants to Build the “Craftsman” Hand
Zhou’s ambition goes beyond making robotic hands capable of basic grasping.
He told Forbes that the company’s objective is effectively to give robotic hands the abilities of highly skilled craftspeople.
That is an important distinction.
An industrial robot traditionally performs one carefully programmed task.
Pick this object.
Move it here.
Repeat 10,000 times.
A dexterous humanoid hand is supposed to do something fundamentally different.
Use one tool.
Then another.
Then manipulate an unfamiliar object.
Then perform a task that requires precision.
Then adapt when conditions change.
Linkerbot’s long-term goal is therefore not merely mechanical movement.
It wants a reusable library of hand skills.
That could be far more powerful.
Linkerbot Already Makes Six Different Robotic Hands
The company’s current portfolio includes six hand models covering a wide range of complexity and price.
Forbes reports prices ranging from 6,666 yuan, roughly $980, to 99,999 yuan, around $14,700.
The entry-level O6 can perform relatively simple five-finger grasping.
The L20, priced around 49,999 yuan, is currently the company’s bestseller and can perform tasks including holding cups, opening bottles, and tightening bolts.
The higher-end L30 is designed for much more precise manipulation.
Linkerbot’s own specifications show that the L30 provides 18 active and four passive degrees of freedom, giving it 22 degrees of movement overall. It can also achieve finger-joint speeds above 400 degrees per second and open or close the hand in roughly 0.2 seconds.
See Linkerbot’s L30 robotic hand specifications
Those numbers matter because robotic dexterity is often measured partly by degrees of freedom, or DoF.
What Does “Degrees of Freedom” Mean?
Think about your index finger.
It does not simply open and close.
It bends at several joints.
Your thumb moves in multiple directions.
Your wrist rotates.
Different fingers can move independently.
Each independent movement represents another degree of freedom.
A human hand has roughly 27 degrees of freedom.
Forbes, citing BofA Global Research, notes that robotic hands generally need at least around 20 degrees of freedom to perform sophisticated tasks such as folding clothes or chopping vegetables effectively.
That is why a robot gripper with two rigid fingers may be excellent at moving boxes but useless at threading a needle.
More DoF means greater flexibility.
It also creates dramatically more engineering complexity.
Every movement needs hardware.
Every movement needs control.
And all those movements need to coordinate.
Linkerbot’s L30 Can Perform Extremely Fine Tasks
The L30 is designed for precision applications.
Forbes reports that it can use tweezers to pick up tiny screws, while Linkerbot positions it for research, medical assistance, industrial precision assembly, and other applications demanding high dexterity.
That may sound like a minor demonstration.
It is actually significant.
Picking up a tiny screw with tweezers requires the robot to:
Position several fingers correctly.
Grip the tweezers.
Control pressure.
Align the tool.
Recognize the screw.
Approach it accurately.
Apply enough pressure to hold it without slipping.
Then move without dropping it.
Humans perform such actions without thinking.
Robots need enormous coordination.
That is why apparently boring tasks are often the most impressive demonstrations in robotics.
Sensors Give Robotic Fingers Something Like Touch
Movement alone is not enough.
A useful robotic hand also needs to know how hard it is gripping something.
Imagine picking up an egg.
Too little force and it falls.
Too much force and it breaks.
Human fingers constantly receive pressure information and adjust grip strength.
Robotic hands attempt to reproduce this using sensors embedded in the fingers and fingertips.
Forbes describes how robotic hands combine motors, specialized tendon-like cables, and sensing hardware that allow the system to regulate pressure during delicate tasks.
This creates one of the biggest technical differences between cheap grippers and truly advanced robotic hands.
A basic gripper knows:
Open.
Close.
An advanced hand needs to understand:
How much?
Linkerbot Still Isn’t the Most Sensitive Hand Available
The company has strong technology, but it does not dominate every performance category.
Singapore-based competitor Sharpa recently demonstrated its Wave robotic hand peeling an apple—an extraordinarily difficult task because the hand must simultaneously manipulate the fruit and tool while controlling pressure precisely.
Forbes reports that specialists consider Sharpa’s Wave more accurate at pressure control than Linkerbot’s L30 because it contains more sensing hardware.
That matters.
Linkerbot is not claiming to possess an unquestionably superior robotic hand.
Its real competitive advantage may be something else:
Price.
And that could ultimately matter more.
Linkerbot’s Biggest Weapon Is Cost
Sharpa’s Wave reportedly costs around $50,000 per hand.
London-based Shadow Robot sells advanced robotic hands for around €110,000, or roughly $125,000 each.
Linkerbot’s L30 costs about $14,700.
That difference is enormous.
A laboratory buying three experimental hands may tolerate $100,000 hardware.
A manufacturer building 500,000 humanoid robots cannot.
If each robot requires two hands costing $100,000 each, the hands alone cost more than many luxury cars.
That is economically impossible for mass adoption.
Linkerbot is therefore attacking the problem the same way Chinese manufacturers have attacked industries ranging from solar panels to electric vehicles:
Scale production until the expensive technology becomes affordable.
Why Does China Have Such an Advantage Here?
Robotic hands require many small precision components.
Motors.
Gearboxes.
Bearings.
Sensors.
Electronics.
Cables.
Machined parts.
Injection-molded pieces.
Circuit boards.
China already has enormous supply chains producing exactly these types of components for:
Smartphones.
Drones.
Electric vehicles.
Industrial automation.
Consumer electronics.
That matters because robotics does not exist in isolation.
A startup located near thousands of established component suppliers can iterate much faster than one that has to import specialized pieces from around the world.
Linkerbot takes this advantage even further by manufacturing nearly all of its major components internally, according to Forbes.
That gives the company more control over both cost and design changes.
Linkerbot Wants Robotic Hands to Cost Less Than $100 a Pair
This may be the most ambitious claim in the entire story.
Today, BofA Global Research estimates the global average hardware cost for two robotic hands at roughly $6,000, while China’s average is already closer to $1,200 per pair.
Linkerbot believes mass production could eventually bring its hardware cost below $100 per pair within three years.
Cofounder Su Yang has even discussed a potential target of approximately 500 yuan, or around $75 per pair, at sufficient production scale.
If that happens, the economics of humanoid robots could change dramatically.
Imagine reducing one of the robot’s most expensive components from thousands of dollars to under $100.
Suddenly, useful hands stop being an exotic research component.
They become something closer to a commodity.
Hands Are One of the Most Expensive Parts of a Humanoid
BofA Global Research estimates that dexterous hands can represent nearly 20% of the hardware cost of an average humanoid robot, which it estimates at roughly $28,000.
That is extraordinary when you consider how small the hands are compared with the entire robot.
Two relatively compact components can cost more than:
Large structural sections.
Leg assemblies.
Some computers.
Battery hardware.
Part of the actuator system.
That is exactly why robotic-hand manufacturing could become such an important industry.
If humanoids eventually sell by the millions, the companies supplying hands may resemble today’s suppliers of batteries, motors, or semiconductor chips.
You do not necessarily need to manufacture the entire robot to build an enormous business.
Linkerbot Is Already Expanding Production Rapidly
This is where Linkerbot stops looking like a research laboratory.
Reuters reported in May that the company was already producing close to 5,000 robotic hands per month and expected output to reach around 10,000 monthly relatively quickly.
Forbes reports that a new Beijing production line began operating in August, bringing the company’s total to four factories across China.
Linkerbot now aims to increase annual production capacity toward 100,000 hands over the next few years.
The longer-term ambition is even bigger.
Zhou told Reuters that he ultimately wants to reach one million pairs of robotic hands.
That is not research-scale manufacturing.
That is automotive-style volume.
China Could Ship 430,000 Robotic Hands Annually by 2030
Linkerbot is not alone.
China has more than two dozen companies developing robotic hands, according to TF Securities estimates cited by Forbes.
Research firm Gaogong Industry Research Institute expects Chinese robotic-hand shipments to reach around 70,000 units in 2026, nearly triple the previous year’s level.
By 2030, that figure could reach 430,000 units annually.
That creates both opportunity and danger.
Huge demand could create winners.
Huge competition could destroy margins.
China has seen this movie before.
The Industry Could Repeat China’s EV Price War
China’s electric-car industry became globally competitive partly through relentless competition.
But the same competition created destructive price wars.
Companies cut prices.
Competitors responded.
Margins collapsed.
Weaker manufacturers disappeared.
The Chinese term neijuan, often translated as “involution,” describes this type of excessive internal competition.
Forbes notes that robotics specialists worry China’s dexterous-hand industry could follow the same path.
Dozens of manufacturers may race to undercut one another until the hardware becomes extremely cheap.
That would be great for humanoid-robot companies.
It could be painful for the hand manufacturers themselves.
Linkerbot therefore needs something beyond cheap hardware.
That something is software.
LinkerSkillNet Could Be More Important Than the Hand Itself
Linkerbot is building a proprietary library called LinkerSkillNet.
The company says it already contains more than 500 hand skills.
These include activities such as:
Threading a needle.
Chopping vegetables.
Making dumplings.
Handling tools.
Playing musical instruments.
Workers collect movement data using motion-capture gloves and robotic arms.
Those movements are then converted into reusable blocks of robotic instructions.
Forbes compares the approach to Lego.
One movement block might represent bending a finger.
Another represents rotating the wrist.
Combine many blocks and the robot can perform a complex task.
That is potentially much more scalable than programming every motion manually from the beginning.
Imagine an App Store for Robot Skills
This is where humanoid robotics starts becoming easier to understand.
Today’s smartphone is valuable partly because of its hardware.
But imagine an iPhone with no applications.
Its usefulness would collapse.
Robots may evolve similarly.
Hardware provides:
Hands.
Legs.
Motors.
Cameras.
Sensors.
But software provides the actual skills.
One robot could download:
Warehouse packing.
Dish washing.
Laundry folding.
Screw assembly.
Vegetable chopping.
Tool handling.
Unitree has already launched a robot software platform resembling an app store for whole-body skills such as dancing and martial arts.
Linkerbot is pursuing a similar concept focused specifically on the hand.
That could become extremely valuable if its hardware becomes widely adopted.
Linker Genesis Could Let Robots Learn Through Language
Linkerbot is also building an AI model called Linker Genesis.
The goal is to let users control robotic hands through spoken commands rather than manually programming each movement.
Imagine telling a robot:
“Pick up that screwdriver and tighten the loose screw.”
The AI needs to translate that sentence into:
Object recognition.
Tool selection.
Finger positioning.
Grip pressure.
Arm movement.
Screw alignment.
Rotation.
Torque control.
Then completion.
That is where large AI models and physical robotics begin converging.
The language model understands the objective.
The robotics system converts understanding into physical actions.
This broader field is often called embodied AI.
Embodied AI Is China’s Next Major Technology Bet
China has made humanoid robotics and embodied artificial intelligence strategic industries.
The government has supported the sector through subsidies, tax incentives, industrial parks, and research funding.
The reason is partly economic.
China’s population is aging.
Its workforce is projected to shrink.
Factories that once depended on huge supplies of relatively inexpensive labor increasingly need automation.
A humanoid robot capable of using ordinary human tools could become extremely valuable.
Traditional industrial robots require factories to be redesigned around them.
Humanoids theoretically fit into factories already designed around human bodies.
Doors.
Stairs.
Workbenches.
Tools.
Shelves.
Machines.
All were created for humans.
A humanoid can potentially use the same environment.
But only if its hands actually work.
Linkerbot Hands Are Already Entering Factories
This is not merely theoretical.
Zhou says Chinese manufacturers already use Linkerbot hands for tasks including gripping objects and assembling components.
Chinese food company Qianweiyangchu Food announced plans in July to use Linkerbot hardware in an automated production line for frozen and prepackaged meals.
That is significant because food handling requires varied shapes and relatively delicate manipulation.
Traditional industrial grippers work beautifully when every object is identical.
A dexterous hand becomes more valuable when objects vary.
That moves robotics closer to human flexibility.
Samsung, Siemens, Stanford and Tsinghua Are Customers
Linkerbot says its customers extend well beyond Chinese humanoid startups.
Forbes reports that organizations using its technology include:
Samsung
Siemens
Stanford University
and
Tsinghua University.
Universities use the hands to experiment with robotic manipulation.
Industrial companies can explore automated assembly.
Robot manufacturers can integrate them into humanoid platforms.
That diversification matters.
Linkerbot does not want its future to depend entirely on whether consumer humanoid robots become popular next year.
Industrial robotic arms need better hands too.
The Company Is Already Worth $3 Billion
Investors clearly believe the opportunity could be large.
Linkerbot completed a Series B+ financing round in April that valued the company at approximately $3 billion.
Investors include major institutions such as Ant Group, HSG, and CICC Capital, along with other Chinese funds.
Reuters separately reported that Linkerbot is now seeking a valuation around $6 billion in its next financing round.
CEO Alex Zhou owns approximately 31% of the company, according to Forbes, giving his stake an estimated value of about $920 million at the latest valuation.
If Linkerbot reaches the next targeted valuation, Zhou could become a billionaire.
Who Is Alex Zhou?
Zhou’s path into robotics is unusual.
Forbes reports that he entered China’s Huazhong University of Science and Technology at only 14 years old through a program for exceptionally talented students.
His studies included robotics and artificial intelligence.
After university, he worked in Lenovo’s research division, developing software used in smartphone design.
He later cofounded startups in gaming and autonomous driving before selling his stakes and using some of the proceeds to finance Linkerbot during its first year.
His interest in robots goes even farther back.
Zhou says he grew up fascinated by the Japanese manga character Doraemon, a robotic cat capable of pulling seemingly endless gadgets from a magical pocket.
He now describes robotic hands almost the same way:
A universal tool capable of performing countless tasks.
Linkerbot Is Trying to Automate Its Own Factory
There is something particularly interesting about a robotics company eventually using robots to manufacture robots.
At July’s World Artificial Intelligence Conference in Shanghai, Linkerbot demonstrated an upcoming hand called the O30.
The company says the O30 is capable of helping assemble robotic hands themselves.
If that works at scale, Linkerbot could begin automating some of its own production process.
The logic becomes circular:
Build better hands.
Use those hands to assemble more hands.
Lower production cost.
Build more robots.
Use more robots to manufacture more hardware.
This is one reason investors are enthusiastic about physical AI.
Once automation becomes capable enough to help manufacture its own components, scaling economics can change rapidly.
But Linkerbot Has Serious Competition
China’s robotics industry is moving extremely quickly.
Beijing-based Inspire Robots shipped more robotic hands than Linkerbot last year, according to Gaogong figures cited by Forbes.
Shanghai’s AgiBot sells its OmniHand for roughly $2,200.
Other Chinese manufacturers include Robotera and Wuji Technology.
Outside China, competitors include:
Sharpa.
Shadow Robot.
Wonik Robotics.
And others.
Northwestern University robotics professor Kevin Lynch told Forbes that after testing several products, he does not see one unquestionable winner.
Different hands have different strengths.
That is important perspective.
Linkerbot is promising.
It has not already solved robotic manipulation.
Human Hands Are Still Far Better
The human hand remains an extraordinary machine.
Twenty-seven degrees of freedom.
Hundreds of muscles and tendons interacting through the arm.
Extremely sensitive tactile feedback.
The ability to manipulate both delicate and heavy objects.
Rapid adaptation without explicit programming.
And an enormously capable brain controlling everything.
Even advanced robotic hands remain far behind that combination.
BofA analysts caution that it will take time before machines reach truly human-level dexterity.
That means investors should be careful with claims that household humanoids are only months away from replacing large amounts of human labor.
Walking is not enough.
Hands are not enough either.
The robot needs perception, reasoning, movement, reliability, and cost to work simultaneously.
Reliability May Be Harder Than Demonstrations
This is another issue technology demonstrations often hide.
A robot picks up an egg on stage.
Everyone applauds.
But a factory asks a different question:
Can it perform the task 50,000 times without failing?
Commercial robots need:
Consistency.
Durability.
Easy maintenance.
Predictable behavior.
Affordable replacement parts.
Safe operation around humans.
A robotic hand performing a difficult task once is impressive.
Performing it continuously for three years is commercially useful.
That transition from demonstration to reliability may determine which robotics startups survive.
The U.S.-China Technology Fight Creates Another Risk
Linkerbot also faces geopolitical uncertainty.
Forbes reports that the U.S. recently introduced restrictions covering imports of certain humanoid robot models and power inverters from China.
The current rules do not directly cover robotic-hand makers, according to Linkerbot.
But future tariffs or restrictions remain possible as the U.S.-China technology rivalry expands.
That matters because Linkerbot already works with American customers including Stanford.
If robotic components become strategically sensitive, Chinese suppliers could face restrictions similar to those seen in telecommunications, semiconductors, drones, and electric vehicles.
Analysts suggest one possible response would be shifting some production outside China, potentially into Southeast Asia.
Is Linkerbot Actually Solving Tesla’s Optimus Problem?
Not literally.
The headline can make it sound as though Tesla has hired Linkerbot to fix Optimus.
There is no evidence in the Forbes reporting that Tesla is a Linkerbot customer.
Tesla is developing its own robotic-hand technology.
So the more accurate interpretation is:
Linkerbot is attacking the same engineering problem Musk has identified as one of humanoid robotics’ hardest challenges.
If Linkerbot succeeds in producing cheap, dexterous, AI-controlled hands at massive scale, the technology could influence the entire industry—even if Tesla never buys one.
Tesla is trying to vertically integrate Optimus.
Linkerbot is trying to become a supplier to everyone else.
Those are very different strategies.
Linkerbot Wants to Become the Nvidia of Robotic Hands
Zhou has an ambitious analogy.
He told Forbes that Linkerbot wants to focus on the most crucial hardware component in robotics in the same way Nvidia focused on crucial computing hardware during the AI boom.
That comparison is extremely ambitious.
But the logic is understandable.
Nvidia does not build most AI applications.
It supplies hardware underneath them.
Linkerbot does not necessarily need to build the world’s best complete humanoid.
It could supply one essential component to dozens of companies that do.
If humanoid robots eventually become a massive industry, that supplier position could be extraordinarily valuable.
The $3 Billion Valuation Is Ultimately a Bet on Humanoids
This is the largest risk.
Linkerbot’s business can serve industrial robotic arms and research laboratories.
But its enormous valuation clearly assumes humanoid robots will become commercially important.
If humanoids remain expensive demonstrations for another decade, demand for advanced hands may grow much more slowly than investors expect.
If millions of humanoid robots begin entering factories, warehouses, restaurants, and homes, the opposite happens.
Every humanoid needs two hands.
Ten million robots means:
20 million robotic hands.
Suddenly, a company capable of manufacturing dexterous hands for $100 becomes enormously important.
That is what investors are betting on.
The Hand May Determine Whether Humanoid Robots Become Useful
Robotics companies understandably love dramatic demonstrations.
Running.
Dancing.
Backflipping.
Martial arts.
Those performances demonstrate balance and motor control.
But they do not necessarily demonstrate economic usefulness.
A factory does not primarily need a robot capable of dancing.
It needs one capable of picking up a screwdriver.
Using it correctly.
Putting it down.
Picking up a cable.
Connecting it.
Handling a delicate component.
Repeating the task thousands of times.
That is why hands may ultimately matter more than spectacle.
And it explains why a startup most consumers have never heard of can already be worth $3 billion.
Linkerbot is not trying to build the humanoid robot everyone recognizes.
It is trying to build the part that makes all of those robots useful.
If Alex Zhou’s company really can combine sophisticated dexterity, AI-based skills, Chinese manufacturing scale, and hardware costs below $100 per pair, it could help remove one of the biggest barriers standing between humanoid robots and mass adoption.
That does not mean Tesla’s Optimus problem is solved.
Human hands remain vastly more capable.
Competitors are advancing rapidly.
Mass-production claims still need to be proven.
But the direction is clear.
The humanoid robot race is no longer only about who can build the best body or AI brain.
Increasingly, it may come down to something humans rarely think about at all:
whether the robot can actually use its hands.
