NVIDIA has taken a major leap forward in humanoid robotics with the introduction of Isaac GR00T-Dreams, a revolutionary blueprint for generating synthetic motion data that promises to transform how robots learn to interact with the world.
Revealed at Computex 2025, GR00T-Dreams enables developers to create enormous quantities of training data in a fraction of the time previously required. The technology works by first post-training NVIDIA's Cosmos Predict world foundation models for specific robots. Then, using just a single image as input, it generates videos of robots performing new tasks in various environments, extracting action tokens that teach robots how to execute these tasks.
NVIDIA Research demonstrated the technology's power by using GR00T-Dreams to develop GR00T N1.5—an update to their humanoid robot foundation model—in just 36 hours, a process that would have required nearly three months of manual data collection. This represents a 40% performance improvement compared to using only real-world data.
"Physical AI and robotics will bring about the next industrial revolution," said Jensen Huang, NVIDIA's founder and CEO, who has repeatedly described physical AI as representing "the world's next trillion-dollar industry." The company is strategically building comprehensive software and hardware infrastructure to power this revolution, from AI brains for robots to simulation environments and supercomputers for training foundation models.
Early adopters of NVIDIA's humanoid robotics technologies include industry leaders like Boston Dynamics, Agility Robotics, Foxconn, and NEURA Robotics. These companies are implementing GR00T-Dreams and related technologies to accelerate development of robots for industrial settings, with applications ranging from manufacturing and material handling to eventually household automation.
The breakthrough addresses a fundamental challenge in robotics development: the costly and time-consuming process of collecting high-quality training data. By generating synthetic data that closely mimics real-world scenarios, NVIDIA is helping to close the gap between simulation and reality, potentially accelerating the timeline for widespread humanoid robot deployment across industries.