LinkerHand O30: 20 Direct-Drive Joints and a 30 kg Max Load, What the Specs Say

The LinkerHand O30 is a five-finger robot hand from Beijing's Linkerbot with 20 active joints, each driven directly by its own motor, and a company-stated maximum load of 30 kg. It weighs about 730 g. Humanoid.guide lists it at 6,000 USD and describes it as in production, as of 13 October 2026.

Every figure below is company-stated or relayed by media and retailers. None of these figures has been independently measured.

What is the LinkerHand O30?

The O30 is a dexterous robot hand meant to be bolted onto an arm or a humanoid. Linkerbot unveiled it at the World AI Conference in Shanghai on 17 July 2026, calling it the world's only fully direct-drive dexterous hand mass-produced on a dexterous-hand production line. That "only" is the company's wording.

In plain terms: each of the 20 joints has its own motor, so there is no tendon or linkage between a distant motor and the fingertip. Humanoid.guide calls it a step up from Linkerbot's linkage-driven L-series hands. Interesting Engineering says the target uses are high-precision manufacturing such as electronics, automotive and aerospace, plus research and use as an end-effector.

What is Linkerbot's thesis?

Linkerbot's bet is that hands, not whole robots, are the scarce part, and that a company that can build them in volume sets the standard. The evidence is the company's own claims. In a December 2025 release, it said it had shipped its 10,000th dexterous hand and claimed over 80% global share of the high-degree-of-freedom segment.

At WAIC it also showed a "hands assembling hands" production line and claimed 93% in-house core parts. Both are company statements. The same report positions the O30 for Sim2Real training, automotive precision installation and aerospace tasks.

How is Linkerbot funded?

Linkerbot has raised several rounds in quick succession, though few amounts are public. Here is what the sources give:

Early backers named by KrASIA are Ant Group and HongShan. The Next Web reports that the latest round added Zhongguancun Science Park Fund, Bank of China Asset Management and Fosun Capital. The $6bn figure is a target reported by Reuters, not a closed round.

What are the LinkerHand O30's specifications?

All values are company-stated or relayed, as of 13 October 2026.

Sources: Interesting Engineering, Humanoid.guide and 10jqka.

How does the LinkerHand O30 work?

Each joint pairs a brushless motor with a worm gear, according to Humanoid.guide. A worm gear is hard to push backward, so the finger holds its position without the motor drawing current. The WAIC report describes this as a self-locking design with zero-power gripping, and says the direct drive removes transmission backlash.

The direct drive also gives closed-loop control of position and torque at each joint, per Interesting Engineering. In practice, that means the controller reads and commands every joint separately, rather than inferring finger motion from a cable's tension. The tradeoff is packaging: 20 motors must fit inside a palm-sized hand.

For control, the hand takes CAN or CAN FD commands at up to 500 per second. Interesting Engineering shows it performing three magic tricks (a card change, pulling a toy car from a card, and a vanishing coin) as a precision demonstration. That video is a company-supplied demonstration, not a test.

What's new compared with Linkerbot's earlier hands?

The O30 is the direct-drive line, while the L-series uses other drives. KrASIA lists the L30 at 22 DoF and the L20 at 21 DoF, and The Next Web says the Linker Hand family spans 6 to 42 degrees of freedom. Humanoid.guide describes the O30's 30 kg as the highest static payload in Linkerbot's current hand lineup.

Note that the L30 has more listed degrees of freedom than the O30. The O30's change is the drive method and load, not a higher joint count. DoF counts also vary between listings, so compare them with care.

Where will people actually use this?

The sources point to industry and research rather than homes.

• Electronics, automotive and aerospace assembly: Interesting Engineering names these as precision manufacturing targets.
• Automotive precision installation: the WAIC report names this task specifically.
• Sim2Real training: researchers can train a hand in simulation and test it on hardware with a known joint layout.
• Research platform: a 20-joint hand is a testbed for manipulation work.
• Arm end-effector: Interesting Engineering lists this as a use, so a factory arm gains a hand without a full humanoid.

For a normal buyer, there is no home use case in the sources.

What are the limitations?

First, nothing has been independently measured. The 30 kg rating, 0.2 mm repeatability and 70 N grasp are company-stated. A 30 kg static rating also does not mean the fingertips can push 30 kg. The listed fingertip forces are 24 N for the thumb, 30 N for a four-finger pinch and 70 N for a five-finger grasp. That is about 7 kg-force in total. The 30 kg is a holding rating, which the self-locking worm gear makes plausible, but dynamic lifting is a different claim.

Second, the sources conflict on small points: weight is 730 g or 740 g, and Zhou Yong is called CEO by The Next Web but "founder and CTO" by KrASIA. Third, 20 motors in one hand raise questions about heat, wear and cost that no source addresses.

What's still unclear?

• Tactile sensing: Humanoid.guide lists it as N/A. Whether the O30 senses touch at the fingertips is not confirmed.
• Price: only a third-party listing at 6,000 USD, with no manufacturer figure.
• Dimensions: not published.
• Lifetime and reliability: no cycle-life or failure data.
• Market share: the 80% claim comes from Linkerbot and relies on its definition of the "high-DoF" segment.

Should you buy it or wait?

If you are a robotics lab or integrator with a CAN-based arm, the O30 is worth a quote request. The specifications fit the use case, and Humanoid.guide lists it as in production. Ask for a datasheet with dimensions, a tactile option and a test report for the 30 kg rating.

If you want a hand for hobby work, wait. At a listed 6,000 USD, it costs more than most arms. Independent measurements, which do not yet exist, should come first.