LG电子与英伟达推进机器人数据工厂合作，年内拟训练约10万小时机器人数据

LG Electronics (066570.KS) Chief Executive Ryu Jae-cheol met with Madison Huang, senior director of omniverse and robotics at Nvidia and the eldest daughter of Nvidia CEO Jensen Huang, and agreed to secure a large volume of robot data amounting to 12 years' worth this year. The high-level meeting came just four days after LG Chairman Koo Kwang-mo met with the Nvidia chief last week, accelerating the two companies' collaboration on global physical artificial intelligence.
LG Electronics said the two companies' key officials, including Ryu and Huang, met at its R&D campus in the Yangjae area of Seoul's Seocho district on the 18th to discuss ways to create synergy in robotics and physical AI. The two agreed in particular to combine LG's DataFactory — being built there as a "robot bootcamp" — with Nvidia's physical AI technology to train the systems on 100,000 hours of robot data, or roughly 12 years' worth, within the year.
Huang described the outcome as "incredible" immediately after the roughly two-hour meeting. Ryu stressed that the company would "secure competitiveness in physical AI through synergy based on 'One LG,' which brings together core capabilities across the group, and through strategic collaboration with global partners, and emerge as a total solution provider in robotics." Group affiliate heads, including LG CNS CEO Hyun Shin-kyun and LG Sciencepark head Chung Soo-heon, also attended to strengthen the partnership with Nvidia.
The large volume of data emphasized by the two companies is seen as a decisive battleground in the global robotics race. Unlike text or images, robots require unstructured and complex data, such as the movements and know-how of on-site workers. Securing such robot data is essential to gaining a technological edge.
The DataFactory that LG Electronics is building for this purpose is infrastructure where robots learn a variety of tasks and where the resulting data is verified and refined. Spanning four floors from one basement level to the third floor, with a total floor area of 10,000 square meters, the facility plans to deploy several hundred robots, including "LG CLOi," within the year to gather data and verify performance. Nvidia will apply its physical AI models and development platforms — such as the "Omniverse Library," the "Cosmos open-world model" and "Isaac open robotics" — at the site.
LG Electronics plans to accelerate development of its Robot Foundation Model (RFM), the "robot brain," by securing large volumes of data. An LG Electronics official said the combination of the high-quality data the company has accumulated over decades at manufacturing and logistics sites with Nvidia's robotics solutions "is expected to be the secret to building a world-class data cycle that sets it apart from other companies."
This DataFactory partnership is expected to create synergy with the next-generation humanoid development that Koo and Huang recently announced after meeting in person. The two met at Nvidia's headquarters in Santa Clara, California, on the 13th local time and signed a memorandum of understanding to develop a next-generation bipedal humanoid next year, with a target unveiling in the first quarter of next year. The two companies also agreed to deploy CLOi at LG Electronics' plant in Tennessee for field testing within the year and to cooperate on building an AI factory next year.
Alongside its partnership with Nvidia, LG Electronics has set a goal of becoming a total robotics solution provider offering everything from robot hardware to software by uniting its relevant internal units. To that end, it established a Robotics Business Center last month to oversee company-wide robot operations, reporting directly to the CEO. It also recently began mass production of "AXium," an actuator that functions as a robot joint.