Leading embodied intelligence company AgiBot announced that its next-generation robot AgiBot G2 will adopt NVIDIA Jetson Thor as its core domain controller.

Notably, the AgiBot G2 is the first mass-produced robot to feature the Jetson Thor platform, marking a significant milestone in the commercialization of high-performance robotic systems.

This collaboration will significantly enhance the robot's overall capabilities in computing power processing, real-time reasoning capability and technical ecosystem integration, delivering embodied intelligence solutions with stronger performance and higher reliability for diverse application scenarios across industrial and commercial fields.

At the previous World Artificial Intelligence Conference 2025, AgiBot unveiled the launch plan for its next-generation robot—the AgiBot G2. Its comprehensive upgrades in motion accuracy and environmental adaptability garnered widespread industry attention. The newly announced core hardware selection further clarifies G2’s technological roadmap. AgiBot G2 will be equipped with NVIDIA Jetson Thor, delivering an approximately 7.5X increase in AI computing power and a 3.1X boost in CPU performance compared to the previous generation, marking a significant leap in capability. This computational advancement will directly translate into advantages in the following areas:

Enhanced Execution Efficiency and Precision:

Faster model inference speeds will accelerate robot movements, improve motion fluidity, and optimize overall operational tempo. Reduced inference latency enhances motion control accuracy, minimizes deviations, and strengthens operational safety and reliability.

Strengthened On-Device Processing Capabilities:

Enhanced computing capability enables on-device execution of more complex AI models, reducing reliance on cloud services. This significantly improves real-time performance and data privacy protection while ensuring stable robot operation in low-bandwidth or offline environments.

Expanded Task Generalization Capabilities:

Higher AI compute supports larger and more complex reasoning models, enhancing the robot’s adaptability to diverse scenarios and increasing the success rate of long-horizon, complex tasks.

Based on the next-generation NVIDIABlackwell architecture, Jetson Thor delivers significant optimization in energy efficiency, offering superior performance per watt compared to the NVIDIA Ampere architecture-based NVIDIA Jetson Orin. When executing tasks of equivalent complexity, the robot’s overall power consumption will be effectively reduced, with its battery life projected to increase by 10–15%, extending effective working duration.

The adoption of Jetson Thor will also deliver significant ecosystem advantages to AgiBot.

NVIDIA's roboticecosystem will empower AgiBot to efficiently develop, optimize, and deploy end-to-end large models for robotics powered by NVIDIA Isaac platform, significantly accelerating product iteration cycles. Moreover, richer toolchains, more comprehensive documentation, and broader developer community support will greatly simplify customized feature development and scenario adaptation by partners and end-users based on AgiBot G2.

AgiBot commits to establishing the AgiBot G2 as an industry benchmark in Embodied Intelligence. The integration aims to meet the growing demand for advanced intelligent robots across domains such as industrial automation, smart logistics, commercial services, and home applications, thereby accelerating the industrialization progress of embodied intelligence technology.