LG puts robotics under the CEO — and starts building a training-data factory

LG Electronics established a Robotics Business Center effective July 1, 2026, reporting directly to CEO Lyu Jae-cheol. The unit consolidates business development, supply chain and manufacturing for robot commercialization, which had been spread across the company.

The structure covers three pillars: industrial robots through Robostar, commercial service robots through Bear Robotics, and residential and home robots. LG invested $60 million in Bear Robotics in March 2024 for about 21%, then exercised a call option to reach a controlling 51% by January 2025.

The data factory is the unusual part

Alongside the reorganisation, LG is building a large-scale robot training data factory at its Yangjae R&D campus in Seocho-gu, Seoul, targeted to begin operating before the end of 2026. That is a deliberate copy of a Chinese playbook — dedicated floors of operators producing demonstration hours — rather than of the American one, where the same data is bought through teleoperated services or scraped from deployed fleets.

LG is the second Korean conglomerate in three weeks to pull robotics up to the CEO line. Samsung Electronics created its Robotics eXperience division under CEO TM Roh on July 21. The convergent move suggests both companies concluded that robotics could not be run as a division of an appliance or handset business.

Why it matters

Every serious physical AI program eventually confronts the same bottleneck: manipulation data does not exist at the scale the models want, and nobody has found a way to buy it cheaply. LG's answer is to build a facility whose output is data rather than product — a capital line item that only makes sense if you believe the models, not the hardware, are the constraint.

It also tells you what LG thinks it is buying with Bear Robotics. A fleet of serving robots in restaurants is a modest business on its own. As a source of continuous real-world interaction data feeding a training pipeline, it is something else.