Humanoid robots on the way: Nvidia and LG collaboration revealed
Nvidia and LG are deepening their joint efforts to develop robotics and physical artificial intelligence, establishing a new 'data factory' in Seoul that will train hundreds of robots simultaneously.
LG Electronics and Nvidia are deepening their strategic collaboration to accelerate the development and commercialization of solutions in the field of robotics and physical artificial intelligence (Physical AI). Both companies are working to create an infrastructure that will allow robots to learn faster by combining LG's industrial experience with Nvidia's artificial intelligence and simulation platforms.
The collaboration gained significant momentum after a strategic memorandum of understanding was signed on August 13 at Nvidia's headquarters in California. Just four days later, executives from both companies met again at LG's new facility in Seoul, known as the "data factory." The pace of these meetings highlights the importance the companies place on accelerating activity in the robotics sector.
LG has defined the current year as a launch point for business expansion in the field and has established a dedicated business center for robotics.
At the center of the collaboration is LG's new "data factory" in Yangjae, located within the company's research campus in Seoul. The facility spans approximately 10,000 square meters across four floors, with hundreds of robots expected to be operational there by the end of the year.
The facility is used for robot training, data collection, generation, and verification. As part of the activity, environments are being built to simulate various situations in which robots are required to operate. One of these is a home environment where the LG CLOiD robot practices cleaning tasks.
Another environment simulates an LG washing machine factory, where robots train on tasks related to moving, loading, and assembling parts. The goal is to enable robots to handle complex tasks in conditions that simulate real-world use.
LG estimates that by the end of the year, about 100,000 hours of training data will have accumulated at the facility. The data will include information collected on-site alongside synthetic data generated using Nvidia's models. According to the company, this volume is equivalent to about 12 years of data collection using standard methods.
The data is expected to be used to advance LG's robotic infrastructure model, with a special emphasis on the development of humanoid robots.
One of the key components of the collaboration is the creation of a "data flywheel." LG believes that a significant advantage in the field of physical artificial intelligence is the ability to continuously collect high-quality data, use it to improve robots, and in turn generate new data.
The company plans to integrate information collected over the years in its production and logistics environments with Nvidia's artificial intelligence platforms. The data collected at the facility will be processed and adapted using Nvidia's robotics technologies to create high-quality datasets for training.
The collaboration includes Nvidia Omniverse technologies for simulations, Nvidia Cosmos, which includes open-world models, and the Nvidia Isaac platform for robotics development.
The connection between the two companies effectively combines Nvidia's technological infrastructure with LG's practical experience in industry, manufacturing, and logistics. While Nvidia provides the artificial intelligence and simulation tools, LG brings real physical environments, industrial expertise, and products intended for home use.
The collaboration also marks LG's attempt to expand robotics activity beyond factories. Alongside solutions for industrial automation, the company is working to promote robots for home use and humanoid robots. LG's goal is to become a "comprehensive provider of robotics solutions" and establish a competitive advantage in a rapidly developing field.
