LG targets 500,000 hours of robot training data and puts home humanoids last, around 2030
On its Q3 earnings call, LG Electronics said it is building about 500,000 hours of robot training data with NVIDIA and will expand into home humanoids around 2030, after manufacturing and logistics. The home, where its CLOiD robot debuted, now comes after parts, data and factory robots.
LG Electronics is building about 500,000 hours of robot training data with NVIDIA, combining real and simulated data, Lee Jae-mo, its head of robotics strategy, said on the company's third-quarter earnings call on Oct 28. In August, LG put its year-end figure at 100,000 hours. Seoul Economic Daily and Aju Business Daily report the new number as this year's goal, which Seoul Economic Daily calls a fivefold increase; Green Economy Newspaper reports it as a goal for next year.
The same call put the home at the end of LG's robot plan. LG will "focus first on manufacturing, where the market is expected to form first," Lee said, and expand into home humanoids "around 2030, when humanoid robots really take off." LG introduced its CLOiD home robot loading a washing machine at CES in January. Its plan now starts with parts, data and factory robots, and the home humanoid comes last.
Real hours from a mock home, the rest from NVIDIA's simulators
The hours come from LG's Data Factory on its former Yangjae R&D campus in Seoul, a four-floor, 10,000-square-meter building where CLOiD robots clean a mock home and move and assemble parts on a copy of LG's Tennessee washing-machine line, according to LG's August release. Data collected there is expanded into synthetic data with NVIDIA's Omniverse, Cosmos and Isaac tools and used to train LG's robot foundation model. LG says the site has begun partial operations, and CFO Kim Chang-tae said it will run fully in the fourth quarter.
Lee said the data will go to LG's own robots first, and paid testing for outside customers is a medium- to long-term option. LG lists three business models: research robots sold as single devices, industry-specific hardware and software packages built with system integrators, and selling data and building data sets from the Data Factory.
Actuators ship before households buy
The first products are parts. LG built an actuator pilot line in Changwon in the first half, according to Green Economy Newspaper, and its earnings release says mass-production preparations are on plan while LG pursues orders from prospective global customers. Seoul Economic Daily reports that LG is also developing a bipedal humanoid with NVIDIA, which it aims to unveil in the first quarter of next year.
For the home, Lee set a price condition. The market opens when humanoids reach "a price households can actually consider buying," he said, and LG sees room in "robots designed for appliances and appliances designed for robots."
The strongest case against this reading is LG's own earlier timetable. In July, the Korea Herald reported that LG planned to commercialize CLOiD as a home robot next year, and in June The Investor reported a target of a commercial home robot by 2028. If either ships before 2030, the 2030 date covers only a full home humanoid, and LG's home business starts well before its earnings call suggested.
Sources Sources verified
- 1.LG Global Newsroom, Oct 28, 2026: LG Electronics Releases Third-Quarter 2026 Financial Resultslg.com
- 2.Seoul Economic Daily, Oct 28, 2026: LG전자 “로봇 데이터 5배 늘린다…2030년 가정용 휴머노이드 진출”m.sedaily.com
- 3.Green Economy Newspaper, Oct 28, 2026: [컨콜] LG전자 로봇, '산업→가정' 확장…내년 학습 데이터 50만 시간 확보greened.kr
- 4.Newspim, Oct 28, 2026: [컨콜] LG전자 "로봇 사업, 엔비디아와 50만시간 데이터 확보할 것"newspim.com
- 5.Aju Business Daily, Oct 28, 2026: [컨콜] LG전자 "엔비디아와 50만 시간 AI 데이터 구축"ajunews.com
- 6.LG, Aug 18, 2026: 데이터팩토리에 ‘로보틱스 핵심 파트너’ 엔비디아 초청해 사업 가속화 논의lg.co.kr
- 7.Korea Herald, Jul 16, 2026: LG Electronics maps H2 strategy around AI, roboticskoreaherald.com
- 8.The Investor, Jun 12, 2026: LG to build Korea's first humanoid 'data factory' to train robotsm.theinvestor.co.kr
- #LG
- #data factory
- #home robots
- #Korea
- #NVIDIA
Share this briefing
Know someone who follows robotics? Send it their way.
Related briefings
More in Embodied AI & Foundation ModelsAWS opens a Physical AI Toolchain with NVIDIA stack hooks
On Oct 8, 2026 AWS launched an open-source Physical AI Toolchain that stitches synthetic data, SageMaker training, simulation and edge deploy — with NVIDIA Isaac and Cosmos in the mix — after RoboMaker's 2025 shutdown.
Skild's S1 learns soccer by playing itself for 140 simulated years
After launching S1 — a robotics foundation model that copies a single video demo without fine-tuning — Skild showed Physical Self-Play: the same base policy trained against itself in Isaac Sim, then transferred to a humanoid.
NVIDIA open-sources ASPIRE, which cut Rubik's Cube turn time from 17 to 3.1 minutes in three episodes
NVIDIA's GEAR lab released ASPIRE v1.0 as open source on Oct 16, adding real-robot results to a framework that lets a coding agent write and refine robot control code and keep validated fixes as text skills. Across three episodes on a bimanual YAM robot, execution time per Rubik's Cube face turn fell from 17 minutes to 3.1 minutes and tokens per turn fell 82%.
Newsletter
Robrief Weekly
One email a week with the robotics news that matters. Free, unsubscribe anytime.