LG Electronics Targets 500,000 Robot Training Hours

LG Electronics said on a third-quarter 2026 earnings call that it would continue collaborating with Nvidia on robot training data and physical AI, aiming to accumulate 500,000 hours of training data by next year. Its Robot Data Factory in Umyeon-dong, Seocho-gu, Seoul, will first be used to improve LG’s own robots. Paid trials for outside customers and data supply are medium- to long-term initiatives.

On the 28th, Lee Jae-mo, an executive overseeing robotics strategy at LG Electronics, said the company was continuing work with Nvidia to advance the technology using reference robots. The companies are collaborating across physical AI, including simulation, world models and robot development platforms. A world model is an AI system that learns physical environments and movement to help robots predict the results of their actions.

LG is building a system that gathers data in settings resembling real work environments, drawing on its experience in homes and manufacturing, and combines it with training data created in virtual environments. Lee said this approach would yield about 500,000 cumulative hours of real and virtual data by next year.

In August, LG set a target of 100,000 hours of data by the end of this year. On the earnings call, it raised next year’s target to 500,000 hours, five times this year’s goal. Potential external businesses for the data factory include standalone robots for research, hardware and software solutions for specific industrial and service sectors developed with systems integration partners, data sales and dataset development.

LG said improving its own robots takes priority over commercializing the data factory for outside customers. The company is considering paid trials for external customers over the medium to long term, and Lee said it would share details once it has developed a specific business model.

LG Electronics said on a third-quarter 2026 earnings call that it would continue collaborating with Nvidia on robot training data and physical AI, aiming to accumulate 500,000 hours of training data by next year. Its Robot Data Factory in Umyeon-dong, Seocho-gu, Seoul, will first be used to improve LG’s own robots. Paid trials for outside customers and data supply are medium- to long-term initiatives.

On the 28th, Lee Jae-mo, an executive overseeing robotics strategy at LG Electronics, said the company was continuing work with Nvidia to advance the technology using reference robots. The companies are collaborating across physical AI, including simulation, world models and robot development platforms. A world model is an AI system that learns physical environments and movement to help robots predict the results of their actions.

LG is building a system that gathers data in settings resembling real work environments, drawing on its experience in homes and manufacturing, and combines it with training data created in virtual environments. Lee said this approach would yield about 500,000 cumulative hours of real and virtual data by next year.

In August, LG set a target of 100,000 hours of data by the end of this year. On the earnings call, it raised next year’s target to 500,000 hours, five times this year’s goal. Potential external businesses for the data factory include standalone robots for research, hardware and software solutions for specific industrial and service sectors developed with systems integration partners, data sales and dataset development.

LG said improving its own robots takes priority over commercializing the data factory for outside customers. The company is considering paid trials for external customers over the medium to long term, and Lee said it would share details once it has developed a specific business model.