LG signed a memorandum of understanding with NVIDIA in Santa Clara on August 13, 2026, and announced it will unveil a bipedal humanoid robot in the first quarter of 2027. The robot will run on the open Isaac GR00T?Isaac GR00T: NVIDIA's open foundation model for humanoid robots that drives their behavior and motion. foundation model and the Jetson Thor?Jetson Thor: NVIDIA's onboard compute module combining data processing and robot motion control. compute module. The novelty is not NVIDIA’s hardware itself, but grounding the whole build on components from across the LG group.
Key takeaways
- MOU signed August 13, 2026 in Santa Clara by Kwang Mo Koo (LG) and Jensen Huang (NVIDIA).
- Bipedal humanoid unveiling planned for the first quarter of 2027.
- Onboard compute on the Jetson Thor module, control by the Isaac GR00T model, safety via NVIDIA Halos.
- Components from the LG group: actuators (LG Electronics), sensors (LG Innotek), battery cells (LG Energy Solution).
- An 80 MW AI factory in Cheonan planned for the first half of 2028.
The NVIDIA stack inside the robot
Onboard computing is handled by NVIDIA Jetson Thor, a module that combines data processing and motion control. The robot’s behavior is driven by Isaac GR00T, an open foundation model for humanoids that NVIDIA makes available to different manufacturers. The safety layer comes from NVIDIA Halos, described by the company as the first full safety framework built for robots. This is the same set of components NVIDIA promotes among other humanoid makers, so LG will not build its edge on the software alone.
| Layer | Component |
|---|---|
| AI brain | NVIDIA Isaac GR00T |
| Onboard compute | NVIDIA Jetson Thor |
| Safety | NVIDIA Halos |
| Training data | Synthetic data (PhysicalWorks) |
The One LG strategy
The supply chain is meant to make the difference. Under a model it calls One LG, the company integrates its own components instead of buying them externally. Actuators come from LG Electronics, sensor arrays from LG Innotek, and battery cells from LG Energy Solution. The goal is cost. The high price of actuators and sensors is one of the main barriers to commercializing humanoids, and vertical integration is meant to lower it. This is a different approach from most rivals, which assemble robots from components bought from outside suppliers.
NVIDIA and LG are expanding their collaboration to accelerate the next wave of physical AI. Building on years of work together, @JensenHuang and LG Group Chairman Kwang-mo Koo are advancing joint efforts across AI factories, robotics, and autonomous vehicles — combining LG’s Show more
Data from a real factory
LG also wants to feed training with data from real production. By the end of 2026, wheeled CLOi robots are to be deployed on the washing machine line at the company’s Tennessee plant to collect operational data. The stream flows into the PhysicalWorks platform, where it is used to generate synthetic data?Synthetic data: Artificially generated training data that supplements or replaces real-world data. and for continuous learning. The compute backbone is to be built in Cheonan, South Korea, where an 80 MW AI factory is planned for the first half of 2028.
Why it matters
The announcement shows where competition in humanoid robotics is shifting. Since the foundation model and the compute hardware come from NVIDIA and are available to everyone, the edge has to be built elsewhere, in production cost and access to data. LG is betting on two assets that startups lack: its own supply chain and working factories that generate real data. If vertical integration truly lowers unit cost, the conglomerate could enter the market more cheaply than specialized rivals. It is a scenario in which industrial scale, not the algorithm alone, decides the outcome.
What's next
- Bipedal humanoid unveiling announced for the first quarter of 2027.
- Wheeled CLOi robots to start on the washing machine line in Tennessee by the end of 2026.
- An 80 MW AI factory in Cheonan planned for the first half of 2028.
