LG, Nvidia Deepen Robotics Push with Seoul Data Factory

LG’s 10,000-square-meter facility will train hundreds of robots on real-world and simulated tasks as the company targets 100,000 hours of robotics data by the end of 2026.

Robots learn much like people do: through practice.

At LG Electronics’ new Data Factory in Seoul, that means repeatedly cleaning rooms, moving objects, stacking components and assembling parts while the company collects the data needed to make those machines better at performing real-world tasks.

LG is combining decades of manufacturing and logistics experience with NVIDIA’s robotics technologies as it accelerates its push into physical AI and robotics.

The Data Factory, currently under construction at LG’s Yangjae R&D Campus, spans approximately 10,000 square meters across four floors and is scheduled to become fully operational by the end of 2026.

LG expects the facility to house several hundred robots by year-end.

Where robots learn by doing

The facility is designed as more than a conventional testing laboratory.

LG has created dedicated environments where robots can perform the same tasks repeatedly while the company generates, validates and refines the resulting training data.

In a replicated home environment, the company’s LG CLOiD™ home robots practice cleaning.

Another section recreates conditions from LG’s washing machine plant in Tennessee, where CLOiD units learn to move, stack and assemble different parts.

Robots are also being used in LG CNS logistics automation solutions, while another area supports LG Innotek’s training of robotic hands.

The goal is to turn every successful or unsuccessful attempt into useful information that can improve future robot performance.

Building a robotics ‘data flywheel’

LG describes the process as a “data flywheel.”

Data collected from real-world and simulated environments is repeatedly secured, refined and used to train robots, which can then generate additional data as their capabilities improve.

LG is also combining this new information with knowledge accumulated through decades of manufacturing and logistics operations.

The Data Factory will serve as a base for expanding and synthesizing robot-learning data gathered from LG manufacturing facilities, logistics sites and home appliances worldwide.

For robotics, the scale of that data matters.

By the end of 2026, LG expects directly collected information from the Data Factory, together with synthetic and augmented data, to reach 100,000 hours.

LG says that is equivalent to roughly 12 years of data.

NVIDIA technologies support simulation and training

The collaboration brings several parts of NVIDIA’s physical AI ecosystem into LG’s robotics development process.

According to LG, the facility incorporates NVIDIA Omniverse libraries, NVIDIA Cosmos open world models and the open NVIDIA Isaac robotics development platform across stages ranging from data generation to robot deployment.

NVIDIA Cosmos will also be used to synthetically generate and augment training data.

LG plans to use the resulting dataset to advance its Robot Foundation Model, or RFM, which the company says underpins the performance of humanoid robots.

This approach allows robots to learn not only from what happens physically inside a training environment, but also from additional scenarios created digitally.

For a robot expected to eventually operate in a home, factory or logistics environment, exposure to more situations can help improve how it reacts when conditions change.

LG and NVIDIA deepen strategic collaboration

The Data Factory initiative follows a broader agreement between the two companies.

LG Group and NVIDIA signed a memorandum of understanding for strategic collaboration on future business initiatives at NVIDIA’s headquarters in Santa Clara, California, on August 13, 2026.

Four days later, senior executives from LG Electronics and other LG Group affiliates met with NVIDIA officials at the Seoul Data Factory to review the collaboration and discuss additional opportunities.

LG said the quick follow-up reflects the companies’ intention to strengthen their strategic cooperation and accelerate commercialization of the robotics business.

From industrial robots to the home

The investment is also part of a broader shift inside LG.

The company has identified 2026 as the starting point of a larger robotics business push and established a dedicated Robotics Business Center reporting directly to the CEO.

The unit oversees robotics activities across the company as LG seeks to improve execution and expand beyond its existing industrial and commercial applications.

The next target is the home.

LG is combining its experience in manufacturing, logistics, robotic components such as actuators, and large-scale learning data as it works toward becoming a robotics solutions provider with both hardware and software capabilities.

“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, CEO of LG Electronics.

Why it matters

The bigger story is not simply that LG is building more robots.

It is building the infrastructure to help robots learn faster and from a much broader range of situations.

For consumers, that could eventually mean home robots capable of handling increasingly complex everyday tasks. For businesses, the same learning systems could support automation across factories, logistics operations and other commercial environments.

By bringing real-world practice, simulated environments, large-scale training data and NVIDIA’s physical AI technologies together, LG is laying the groundwork for a robotics business that stretches from the factory floor to the home.