LG deepens NVIDIA robotics tie-up with Seoul data factory

LG Electronics has expanded its robotics collaboration with NVIDIA, centring on a new data factory in Seoul.

Executives from both companies met at the site, now under construction at LG's Yangjae research and development campus, to review the project and discuss how to broaden the partnership. LG Group and NVIDIA have also signed a memorandum of understanding on wider strategic cooperation.

The collaboration focuses on robot learning, combining data from manufacturing and logistics operations with NVIDIA's robotics software tools to improve training.

At the Seoul facility, LG has deployed its LG CLOiD home robots to collect operational data. The site includes training areas where robots repeat tasks, along with separate spaces where the resulting data is tested and refined.

One section recreates a home environment where robots carry out cleaning tasks. Another simulates a manufacturing setting based on LG's washing machine plant in Tennessee, where robots move, stack and assemble parts.

Robots are also used in LG CNS logistics automation systems and in an area where LG Innotek trains robotic hands. Data collected across these settings is then fed into NVIDIA's robotics stack for processing and synthetic augmentation.

Data strategy

LG describes the approach as a data flywheel, in which repeated collection and reuse improve robot performance over time. It argues that decades of factory and logistics data provide a strong foundation for that effort.

NVIDIA's role includes software used from simulation to deployment. LG cited NVIDIA Omniverse libraries, NVIDIA Cosmos open world models and the NVIDIA Isaac robotics development platform as part of the system used at the facility.

The data factory spans four floors, including one basement level, and covers 10,000 square metres. LG expects it to house several hundred robots by the end of the year.

By then, the company expects to have amassed 100,000 hours of training data from a mix of material gathered at the site and synthetic data generated through NVIDIA tools. LG said that volume is equivalent to roughly 12 years of data.

It plans to use the material to develop its Robot Foundation Model, which underpins humanoid robot performance. The work is part of a broader robotics push that LG has identified as a priority this year.

Business push

LG recently set up a Robotics Business Centre that reports directly to its Chief Executive Officer. The unit is intended to oversee the group's robotics activities and improve execution across the business.

The company already operates in industrial and commercial robotics and is now aiming to expand into home robotics. It says its manufacturing base and work on components such as actuators give it a route to cover both parts and finished machines.

The tie-up with NVIDIA reflects a wider pattern in robotics, with manufacturers pairing large operational data sets with simulation and model training tools to shorten development cycles and reduce the need for real-world testing for every task.

For LG, the challenge will be turning a broad internal technology effort into products that can be used across homes, warehouses and factories. The scale of data collection planned for the Seoul site suggests the company sees training infrastructure as central to that plan.

"Through the synergy built on 'One LG' - bringing together core capabilities across the Group - and strategic collaboration with global partners, we will secure our competitiveness in physical AI and become a comprehensive robotics solutions provider," said Lyu Jae-cheol, Chief Executive Officer, LG Electronics.