Nvidia Accelerates Humanoid Robotics with Advanced AI Models and Synthetic Data

Nvidia is rapidly advancing humanoid robotics technology by launching new AI-driven tools and systems designed to accelerate robot development and deployment. At Computex 2025, Nvidia introduced the Isaac GR00T N1.5, an updated customizable foundation model for humanoid reasoning and skills, alongside the Isaac GR00T-Dreams blueprint, which generates synthetic motion data to teach robots new behaviors. These innovations aim to overcome traditional data collection bottlenecks and enhance robot adaptability in dynamic environments.

The Isaac GR00T-Dreams blueprint leverages Nvidia’s Cosmos platform to generate vast amounts of synthetic video data, enabling robots to learn complex tasks such as picking up objects or pouring liquids without direct physical training. By fine-tuning video generation models on robot-specific footage, Nvidia breaks the physical limitations of real-world data collection, multiplying training data at unprecedented scales. This approach allows robots to understand the physics and semantics of actions, improving their ability to perform new tasks accurately.

The updated GR00T N1.5 model, developed using synthetic data generated by GR00T-Dreams, significantly enhances humanoid robots’ success rates in material handling and manufacturing tasks. It offers better adaptability to new environments and improved object recognition through natural language instructions. This model will be deployable on Nvidia’s upcoming Jetson Thor edge computing platform, facilitating real-time robot operation in industrial and household settings. Early adopters like AeiRobot and Foxlink are already leveraging these capabilities to improve workflow automation and robot flexibility.

To support comprehensive robot training pipelines, Nvidia also unveiled several complementary simulation and data generation frameworks: Cosmos Reason, a world foundation model that curates high-quality synthetic data; Cosmos Predict 2, enhancing world generation accuracy; Isaac GR00T-Mimic, which creates synthetic motion trajectories from limited human demonstrations; and Isaac Sim 5.0, an open-source simulation environment. These tools collectively reduce the cost and risk of physical robot testing by enabling extensive virtual training.

Global technology leaders such as Cisco, Dell Technologies, Hewlett Packard Enterprise, Lenovo, and Supermicro are integrating Nvidia RTX PRO 6000 Blackwell-powered workstations and servers to provide unified architectures for robot development workloads. These systems enable seamless execution of training, synthetic data generation, simulation, and edge deployment. Additionally, Nvidia’s DGX Cloud and Blackwell systems offer scalable cloud compute resources, delivering up to 18 times greater performance for large-scale robotics data processing and model training.

Nvidia’s holistic approach to humanoid robotics—combining advanced AI foundation models, synthetic data generation, simulation frameworks, and powerful compute infrastructure—addresses the critical challenges of data scarcity, training time, and deployment complexity. By enabling robots to learn from synthetic experiences and adapt to diverse environments, Nvidia is laying the groundwork for the next industrial revolution driven by physical AI and robotics.

Industry leaders including Agility Robotics, Boston Dynamics, Fourier, Foxlink, Galbot, Mentee Robotics, NEURA Robotics, General Robotics, Skild AI, and XPENG Robotics are already adopting Nvidia’s Isaac platform technologies to accelerate their humanoid robot development and deployment efforts. This widespread adoption underscores the platform’s versatility and effectiveness in advancing robot intelligence and operational capabilities across sectors.

As Nvidia CEO Jensen Huang emphasized, physical AI and robotics represent the forthcoming industrial revolution. Nvidia’s integrated ecosystem—from AI brains and simulated training environments to supercomputers for foundation model training—provides essential building blocks for every stage of the robotics development journey, empowering developers to create smarter, more capable humanoid robots.