LG Electronics is converting its R&D campus in Seoul’s Yangjae district into Korea’s first dedicated data factory for humanoid robots. The 33,000 square-meter facility across four floors will deploy up to 100 units of CLOiD, LG’s prototype home humanoid, as early as July, with plans to scale to 300 units by year-end. Korean media have reported that LG plans to invest more than 400 billion won – approximately $263 million – into the project through 2030. An LG official confirmed the broad plan while noting that the investment figure and specific start date have not yet been finalized.
CEO Lyu Jae-cheol is said to visit the site weekly, reflecting the strategic priority the initiative carries inside the company.
Why Data Is the Bottleneck
Generative AI systems learn from text and images that already exist on the internet. Robots cannot draw on equivalent pre-existing data. A robot cannot learn to grip a cup or open a door from a dataset – it must physically perform the motion and record the result, repeatedly, under varying conditions. The volume and quality of that physical interaction data increasingly determines how capable a robot’s foundation model becomes.
Inside the Yangjae facility, CLOiD robots will work through a simulated home environment fitted with LG appliances and a running production line, collecting the kind of interaction data that corresponds directly to the spaces the robot is eventually meant to operate in. The connection between training environment and deployment environment is deliberate – home-specific data trains home-specific capability.
From Hardware to Software Brain
LG’s historical advantage in robotics-adjacent technology is its hardware depth. The company has spent decades developing the motors behind its appliances and is beginning mass production of actuators – the components that drive robot joint movement – in the first half of 2026. The data factory represents a strategic extension from that hardware foundation into the software layer: the robot foundation model that allows a robot to perceive its surroundings and decide how to act autonomously.
CLOiD debuted at CES in January, but its movements were noted by observers as slow. CEO Lyu has described the fix as “large-scale training” – the same capability the data factory is designed to supply before CLOiD reaches the commercial market. LG is targeting a commercial home robot launch by 2028.
The Broader Data Infrastructure Race
The Yangjae facility joins a growing set of purpose-built robot training environments. China opened its first heterogeneous humanoid robot training facility in Shanghai in July, covering more than 5,000 square meters and hosting over 100 robot types from more than a dozen manufacturers. Germany’s TUM RoboGym, built in partnership with NEURA Robotics, is designed to generate physical interaction data at academic scale near Munich Airport. LG’s facility is the largest single-company domestic deployment of this kind in Korea and represents one of the most capital-intensive data generation investments by an appliance manufacturer entering humanoid robotics.
