Nvidia's Jim Fan Previews Autonomous Humanoid Demo
Fan, who co-leads Project GR00T, will present the teaser at the CoRL 2026 keynote in Austin, Texas, on November 9, 2026.
Updated on Nov. 7, 2026 in Robotics
Jim Fan, who co-leads Nvidia's Project GR00T, has confirmed that the teaser for his CoRL 2026 keynote shows real humanoid hardware operating autonomously at normal speed. The keynote is scheduled for November 9, 2026, in Austin, Texas.
Why it matters
The teaser's autonomous, real-time footage is intended to show progress on a central obstacle in humanoid robotics: robots lack an internet-sized library of physical movement data to train on. Nvidia's GR00T effort targets that gap directly.
The methodology expands limited teleoperation data by over 1,000x using simulations and generative world models, and the GR00T N1 model is 2B parameters. The keynote's specific task benchmarks and performance metrics have not yet been disclosed.
The players
Jim Fan
Jim Fan is a robotics researcher who co-leads Nvidia's Project GR00T, an effort to build foundation models for humanoid robots.
Nvidia
Nvidia is a semiconductor and AI computing company that develops the GR00T humanoid robot foundation models and platforms such as Isaac Lab and Cosmos.
Jensen Huang
Jensen Huang is the chief executive of Nvidia, who unveiled Project GR00T at the GTC 2024 conference.
The details
Fan will show the range of tasks the robots handle and describe how the training pipeline works, using platforms such as Isaac Lab and Cosmos. The robot in the teaser runs autonomously rather than under teleoperation.
Timeline
March 2025: Nvidia introduced the open-weight GR00T N1 model.
2025: Follow-up versions N1.5 and N1.6 were expected to roll out.
November 9, 2026: Fan's CoRL 2026 keynote is scheduled in Austin, Texas.
The Tech Race
The teaser extends Nvidia's Project GR00T, which Jensen Huang unveiled at GTC 2024 and which released the open-weight GR00T N1 in March 2025. It signals Nvidia's push to build foundation models for humanoid robots as a competitive platform in the field.
Autonomous humanoid robots that learn from simulated and synthetic data could eventually reduce the cost and time of training machines for physical tasks. The keynote will show what tasks robots can already handle, which signals how close such systems are to practical use.
The takeaway
Training robots on simulated and generated data rather than slow, costly real-world collection is the approach Nvidia is betting on. Watching the CoRL 2026 keynote will show how far that approach has progressed in practice.
What happens next
Fan's CoRL 2026 keynote on November 9, 2026, in Austin, Texas, is where the teaser's full presentation and training pipeline are expected to be detailed.
Further reading
For broader context on humanoid machines, explore Robotics coverage on Wisevoter.
