The shape of the deal

Under an agreement signed by LG Corp CEO Kwang Mo Koo and Nvidia CEO Jensen Huang, a new humanoid robot will be developed. The robot is planned to be shown publicly in early 2027, a date expected to coincide with CES 2027.

The division of labour is clear: Nvidia provides the robot's AI infrastructure, while LG brings in technologies from across its companies. Actuators, sensors and battery technologies developed by LG Electronics, LG Innotek and LG Energy Solution will be used in the robot's physical platform.

The software side

The robot's software foundation will use the vision-language-action model from Isaac GR00T. That model analyses multiple inputs such as image and language to determine the movements the robot needs to perform.

In practice this changes how development works. Instead of training a model from scratch for every task, developers can train for different tasks and robots on top of an existing foundation model. Robotics, in other words, is moving toward the fine-tuning-on-a-foundation pattern already established in language models.

The hardware and platform Nvidia highlights for this work:

  • The Jetson Thor and Isaac GR00T platform is designed for low-latency AI processing in humanoid robots.
  • The system offers 2070 FP4 TFLOPS of AI compute.
  • It carries 128GB of memory.

The role of simulation

Nvidia previously introduced the Isaac GR00T platform to speed up humanoid robot development. It allows data to be generated, models trained and different tasks tried in simulation before testing on a real robot.

That detail changes where cost accumulates in robotics. Experimenting with a physical robot is slow and expensive; every mistake means hardware wear. Shifting most of the experimentation into simulation is the main factor shortening development time.

Why it matters

The collaboration shows how the division of labour in humanoid robotics is settling: the company that builds the hardware and the one that supplies the brain are different. LG combines its hardware expertise with Nvidia's AI and robotics technologies.

The companies expect physical AI to accelerate the development of robots usable in settings from homes to factories. Even so, the picture should not be read as further along than it is: what exists is an agreement and a reveal date, not a product. In humanoid robotics the distance between a demonstration and commercial use has always been long; the real test begins after the early 2027 showing, with what the robot can do off a controlled stage.