"Incredible."
This was the assessment left by Madison Huang, the eldest daughter of Nvidia CEO Jensen Huang and a senior director, after visiting LG Electronics' "robot training center" in person. Huang, a senior director, handles product and technology marketing for Nvidia's physical artificial intelligence (AI) platform. Five days after LG Group Chairman Koo Kwang-mo and CEO Jensen Huang made their cooperation in robotics more concrete, follow-up discussions continued on LG's development site.
LG Electronics plans, in collaboration with Nvidia, to secure a total of 100,000 hours of training data by the end of the year by combining data collected directly at its data factory with amplified and synthesized virtual data to boost the intelligence of its humanoids. Converted on a 24-hour-per-day basis, that amounts to about 11 years and 5 months of massive data.
Huang, the senior director, entered the LG data factory under construction in Seocho-gu, Seoul, at about 8:40 a.m. on the 18th and left at about 11:12 a.m. The visit ran more than 30 minutes longer than the roughly two hours initially scheduled. LG Electronics President Lyu Jae-cheol, LG CNS President Hyun Shin-Gyoon, and LG Sciencepark CEO Jung Soo-heon greeted Huang.
After receiving a bouquet from LG Electronics' robot CLOiD, Huang signed "AMAZING LG" on CLOiD and reviewed data augmentation and synthesis solutions with Lyu. They checked the operational status of the data factory and discussed directions for collaboration in Robotics and ways to create synergy.
Even after leaving the building, Huang continued a cheerful conversation with Lyu and Kang Sung-jin, an LG Electronics executive in charge of partnerships. In the right hand was a shopping bag printed with the phrase "LIFE's GOOD." When reporters asked for impressions of the data factory after boarding the vehicle, Huang lowered the window, replied, "Incredible," and gave a thumbs-up with one hand.
"Incredible" is also a phrase the Huang family has used to emphasize positive evaluations. Ahead of GTC 2026, Huang described physical AI and Robotics partners as an "incredible ecosystem." CEO Jensen Huang also called cooperation with Korean corporations in June an "incredible year."
Lyu said, "We will secure competitiveness in physical AI through synergy based on 'One LG,' which brings together the group's core capabilities, and strategic collaboration with global partners, and we will reinvent ourselves as a Robotics total solution provider."
◇ A 10,000-square-meter 'training center' that replicates homes and production facilities
The data factory LG Electronics is building is a facility that converts the existing Yangjae research and development (R&D) campus into a robot learning hub. It spans four floors from one basement level to three aboveground levels, with a total floor area of 10,000 square meters, and the company plans to ramp up to hundreds of robots within the year for full operation.
Inside, it recreates the environments where robots will actually work. CLOiD cleans in a home setting, and in a manufacturing space modeled after the washing machine process in Tennessee in the United States, it moves, stacks, or assembles parts. Robots are also deployed in LG CNS' logistics automation solution area and LG Innotek's robot hand learning space.
Robots must decide how to move their arms and hands whenever the position, shape, or orientation of an object changes. To work in spaces where surroundings constantly change, such as homes and factories, varied motion data under diverse conditions is needed. Recreating every situation with real robots to collect data takes considerable time and expense.
LG Electronics amplifies and synthesizes data secured on-site using Nvidia's Omniverse library, the Cosmos open-world model, and the Isaac Robotics development platform. The method expands real data into virtual data under multiple conditions and then feeds it back into robot training. The year-end target of 100,000 hours is the sum of data collected directly and the virtual data created in this way.
LG Electronics is building a "data flywheel (virtuous cycle system)" that repeats data collection → amplification and synthesis → training → performance improvement. It will use work data accumulated over decades in manufacturing and logistics sites and the know-how of skilled workers as training material, while Nvidia provides technology to expand, simulate, and train on it. The secured data will also be used to advance LG Electronics' own robot foundation model.
◇ Five days after Koo Kwang-mo and Jensen Huang reached an agreement: follow-up talks on the robot development floor
Chairman Koo and CEO Jensen Huang signed a memorandum of understanding (MOU) on the 13th (local time) at Nvidia's headquarters in Santa Clara, California, for strategic business cooperation in robotics, AI factories, and mobility. Five days later, CEO Lyu and Senior Director Huang met at the data factory to discuss data building and robot development plans. LG Electronics said it aims to use this meeting to make the cooperative framework between the two companies more concrete and speed up commercialization.
Huang has been continuously involved in Robotics collaboration with LG. In June, Huang personally introduced physical AI cooperation with LG on Nvidia's official blog and outlined plans for data generation, simulation, and training. The two companies are also developing next-generation robots using Nvidia's humanoid foundation model "Isaac GR00T" ecosystem. Last month, LG Electronics established a CEO-affiliated Robotics Business Center to strengthen its robot commercialization organization.