Companies

Samsung and LG split on humanoid robots ahead of CES 2027

Samsung is building tendon-driven robotic hands and testing them across 100+ factories, while LG deploys CLOiD at a Tennessee plant and gathers 500,000 hours of training data with Nvidia ahead of CES 2027.

Korean tech giants take different paths in humanoid robot race
Korean tech giants take different paths in humanoid robot racePIRO4D / pixabay
By Sophie Lindqvist5 min read

Why it matters

  • A Samsung patent application published in October describes palm-housed actuators driving individual fingers via connected pulleys.
  • LG is working with Nvidia to collect approximately 500,000 hours of robot training data combining real-world and simulated data.
  • LG is testing CLOiD at its Tennessee washing machine plant, transporting top panels and sorting components.
  • Daol Securities analyst Kim Yeon-mi expects LG's robotics revenue contribution only around 2030, with key milestones in 2027.
  • Samsung plans to test robots across its network of more than 100 factories and production lines.

Samsung Electronics and LG Electronics are betting on humanoid robots with sharply different strategies, and the divergence will be fully visible when both companies present their roadmaps at CES 2027.

Samsung is going after the hardest mechanical problem in the field: the robotic hand. Its newly established Robotics Experience Division is developing tendon-driven technologies that use actuator-powered cables and pulleys to control robotic fingers and joints with precision, according to industry officials who spoke on Thursday.

LG is taking the opposite route. It is steering its CLOiD robot, first showcased as a home helper in January, into factories, starting with trials at its washing machine plant in Tennessee. The company is also working with Nvidia to collect roughly 500,000 hours of robot training data.

The stakes are straightforward. Humanoid robots that can actually perform varied work — not just repeat one rigid motion — need hands that handle tools, components, and fragile parts. Whoever solves dexterity and data collection first has a credible shot at the industrial automation market.

Why is Samsung obsessed with hands?

Dexterity is one of the most critical capabilities for expanding the practical applications of humanoid robots. Conventional robots repeat a single rigid motion. Humanoids have to perform various tasks in constantly changing environments.

That means going beyond simply grasping and moving objects. A useful humanoid must handle a wide range of tools and components with precision, and the challenge grows when the parts are delicate and fragile, demanding increasingly exact control of robotic fingers and arms.

Humanoid hands fall into two broad camps based on how they're actuated. The trade-offs matter:

  • Tendon-driven systems, used by Tesla, move finger joints through cables resembling tendons in the human hand. Fingers stay light, but the cables wear out and need replacement.
  • Joint-level actuation, used by Boston Dynamics, places actuators directly at individual joints. Motion control is precise, but motors add weight and generate heat.

Samsung appears to be exploring a tendon-inspired approach that combines motor-driven cables with active and passive pulleys to coordinate finger movements. A patent application published in October describes actuators housed in the palm that drive individual fingers, while connected pulleys coordinate movement across multiple joints.

The commitment runs to the top of the company. Last month, Executive Chairman Lee Jae-yong visited the Robotics Experience Division at Samsung's R&D campus in Seoul, where he reportedly reviewed progress on the robotic hand technologies.

The takeaway: Samsung is treating the hand as the bottleneck that decides whether humanoids become general-purpose machines or expensive demos.

What else is Samsung building?

Hardware is only part of the picture. Samsung is also researching world models — systems that predict changes in surrounding environments — along with related robot foundation models.

Those models need real-world grounding, and Samsung has an unusual asset for it: a network of more than 100 factories and production lines. The company plans to use that network to test robots' ability to perform real-world tasks and refine their capabilities.

In plain terms, Samsung can iterate on robot hands and the software that controls them inside its own manufacturing empire before selling anything to outsiders.

Where does LG's CLOiD go next?

LG Electronics first showcased CLOiD at CES 2026 in January, positioning it as a home humanoid robot. The demonstrations focused on household chores such as folding laundry and retrieving ingredients.

The company has since changed its sequencing. CLOiD currently uses a wheeled base, but industry officials expect LG to unveil a bipedal humanoid version at CES 2027.

The near-term focus is industrial. LG recently said it is deploying CLOiD at its washing machine plant in Tennessee to test the robot's ability to transport washing machine top panels and to sort and stack components.

On the data side, LG is working with Nvidia to establish a system for collecting approximately 500,000 hours of robot training data by combining real-world and simulated data. That volume matters because robot learning is notoriously data-hungry, and simulation is the only practical way to reach it at scale.

What did LG tell investors?

During its third-quarter earnings call, LG spelled out its market priorities. The company said it would "first focus our robotics business on the academic, public and manufacturing sectors, where markets are expected to develop more quickly."

Home robots haven't been abandoned. "As for home robots, we are developing robots and form factors that can naturally integrate into households by leveraging the competitive strengths we have accumulated in the home appliance business," the company said.

The logic tracks with LG's existing business. Factories and research labs will pay for robots that work today; households mostly won't until the price and capability curves cross.

When does any of this make money?

Not soon, according to Daol Securities analyst Kim Yeon-mi. She said LG's robotics business is not expected to contribute meaningfully to revenue until around 2030.

"LG Electronics' strength lies in its ability to secure early demand in both the manufacturing and R&D markets," Kim said.

"While robotics is not expected to make a meaningful contribution to revenue until around 2030, the company's key commercialization milestones are scheduled in 2027, making the value of its robotics business more apparent."

That timeline frames CES 2027 as the proof point. It's where LG's bipedal CLOiD is expected to appear and where the company's 2027 commercialization milestones start becoming visible to investors.

What the split tells us

The contrast between the two companies is now the story. Samsung is engineering from the fingertip — solving manipulation first through tendon-driven hands, world models, and factory-scale testing across more than 100 sites. LG is engineering from the deployment — putting a wheeled robot into a Tennessee plant, collecting 500,000 hours of training data with Nvidia, and letting industrial demand pull the technology forward.

Both paths converge on the same market eventually, but they price different risks. Samsung's bet assumes dexterity is the gating capability. LG's bet assumes data and early customers are. Which assumption holds will start to become clear on the CES 2027 stage, with revenue significance for both arriving, by Kim's estimate, no earlier than around 2030.

Source: feed.koreatimes.co.kr

Share this article:

More from Sophie Lindqvist

Sophie Lindqvist

Show full bio

Staff writer covering regulation, policy and their impact on artificial intelligence and machine learning at AI In Context.

823 articles

Related articles

  1. Tesla Workers Resist Training Optimus Robots Meant to Replace Them
  2. Toyota to Deploy 400,000 Robots, Spending $6.42 Billion a Year
  3. '70% success is like it doesn't work': why AI robots aren't coming to your home
  4. Amazon Quietly Builds a Home Robotics Team for a 'New-to-World' Robot
  5. Atlas set for chip factories as GlobalFoundries scales Spot

« Previous articleNext article »