NVIDIA Jetson AGX Thor Developer Kit - Ultimate Platform for Physical AI and Robotics
The NVIDIA Jetson AGX Thor Developer Kit is the most powerful edge AI platform for next-generation humanoid robotics, autonomous machines, and physical AI systems. Powered by the NVIDIA Blackwell GPU architecture and 128GB of high-speed LPDDR5X memory, this developer kit delivers an astonishing 2070 FP4 TFLOPS of AI compute performance within a compact 130W power envelope. Built on the NVIDIA T5000 system-on-module (SoM), the Jetson AGX Thor provides 7.5 times higher AI performance and 3.5 times better energy efficiency compared to the Jetson AGX Orin, making it the definitive platform for running the latest generative AI models at the edge.
Why You'll Love the NVIDIA Jetson AGX Thor
-
Blackwell GPU Power: Up to 2070 FP4 TFLOPS for seamless generative AI model inference and real-time decision-making
-
128GB LPDDR5X Memory: Massive memory bandwidth for handling large language models, vision-language models, and multi-modal AI workloads
-
14-Core Arm Neoverse-V3AE CPU: Robust processing for low-latency, real-time applications with Multi-Instance GPU (MIG) support
-
Complete Out-of-the-Box Solution: Includes 1TB NVMe SSD pre-loaded with Ubuntu 24.04 LTS and JetPack 7, Wi-Fi 6E module, and 140W power supply
-
High-Speed Sensor Integration: QSFP slot with 4 × 25GbE networking, Multi-GbE RJ45, USB 3.2, HDMI, DisplayPort, and extensive I/O for real-time sensor fusion
-
Isaac GR00T Ready: Native support for NVIDIA Isaac GR00T workflows, Metropolis, Holoscan, and agentic AI frameworks
Built for Physical AI and Humanoid Robotics
The Jetson AGX Thor represents a new class of robotic computers, architected from the ground up for humanoid robots and physical AI applications. It supports a wide range of generative AI models, including Vision Language Action (VLA) models like NVIDIA Isaac GR00T N, popular large language models (LLMs), and vision-language models (VLMs). Whether you're developing intelligent robots, autonomous vehicles, or advanced edge AI systems, Jetson Thor accelerates your innovation with cutting-edge compute performance and real-time responsiveness.
Advanced AI and Accelerators
Beyond the Blackwell GPU, the Jetson AGX Thor includes a comprehensive suite of hardware accelerators designed for demanding robotics workloads:
-
Third-Generation Programmable Vision Accelerator (PVA): Offload complex vision processing tasks
-
Dual Video Encoders and Decoders: High-throughput video processing for surveillance, streaming, and multi-camera systems
-
Optical Flow Accelerator: Real-time motion tracking and scene understanding
-
Camera Offload Engine: High-speed sensor ingestion for multi-modal perception pipelines
Seamless Cloud-to-Edge AI Workflows
The Jetson Thor runs the full NVIDIA AI software stack for physical AI, delivering a seamless cloud-to-edge experience. Develop and deploy applications using NVIDIA Isaac for robotics, NVIDIA Metropolis for visual AI, and NVIDIA Holoscan for sensor processing. Build agentic AI workflows like Video Search and Summarization (VSS) directly at the edge. With JetPack 7 pre-installed on Ubuntu 24.04 LTS, you get immediate access to CUDA, TensorRT, cuDNN, and the latest AI frameworks without lengthy configuration.
Extensive Connectivity for Multi-Modal Sensor Fusion
The developer kit's reference carrier board offers an extensive array of I/O options for high-speed sensor fusion and real-time data processing:
-
QSFP Slot: 4 × 25GbE for ultra-high-bandwidth networking
-
Multi-GbE RJ45 Connector: Wired Ethernet for reliable connectivity
-
Multiple USB 3.2 Ports and USB-C: Peripheral connectivity and power delivery
-
HDMI and DisplayPort: Dual display outputs for visualization and debugging
-
CAN Bus: Automotive and industrial communication protocol support
-
M.2 NVMe Slot: Expandable storage (1TB drive included)
-
Wi-Fi 6E Module: High-speed wireless connectivity included
Power and Efficiency
The Jetson AGX Thor is designed for flexible power management, supporting 40W to 130W configurable power modes. This enables developers to optimize for performance when plugged in or energy efficiency for battery-operated robotic platforms. The included 140W power supply ensures you have headroom for full-throttle AI workloads and multi-sensor processing.
Ideal Applications
-
Humanoid Robotics: Real-time perception, decision-making, and motor control for advanced humanoid platforms
-
Autonomous Machines: AMRs, drones, and self-driving vehicles requiring multi-modal AI
-
Physical AI Research: Frontier research on VLA models, embodied AI, and generalist robotics
-
Industrial Automation: Vision-guided robotics, quality inspection, and intelligent manufacturing
-
Smart Spaces and Surveillance: Agentic AI for video analytics, search, summarization, and edge intelligence
-
Medical Robotics: Surgical assistants, patient care robots, and medical imaging at the edge
What's in the Box
- NVIDIA Jetson AGX Thor Developer Kit (with Jetson T5000 module pre-mounted on reference carrier board)
- 1TB NVMe SSD (pre-loaded with Ubuntu 24.04 LTS and JetPack 7)
- 140W Power Supply
- Wi-Fi 6E Module (pre-installed)
- Quick Start Guide
Full Specifications
AI Performance
| AI Compute (FP4) |
2070 TFLOPS |
| Performance vs. AGX Orin |
7.5× higher AI compute |
| Energy Efficiency vs. AGX Orin |
3.5× better efficiency |
Processing
| GPU |
NVIDIA Blackwell GPU with Multi-Instance GPU (MIG) support |
| CPU |
14-core Arm Neoverse-V3AE |
| Memory |
128GB LPDDR5X |
| System-on-Module |
NVIDIA T5000 (Jetson Thor series) |
Storage and Media
| Storage |
1TB NVMe SSD (included, pre-loaded) |
| Video Encode |
Dual hardware encoders |
| Video Decode |
Dual hardware decoders |
| Optical Flow Accelerator |
Yes |
| Programmable Vision Accelerator |
3rd-generation PVA |
Networking and Connectivity
| High-Speed Networking |
QSFP slot with 4 × 25GbE |
| Ethernet |
Multi-GbE RJ45 connector |
| Wi-Fi |
Wi-Fi 6E module (included) |
| USB |
Multiple USB 3.2 Gen 2 ports, USB-C with DP Alt Mode |
| Display Outputs |
HDMI 2.1, DisplayPort 1.4a |
| CAN Bus |
Yes |
| M.2 Expansion |
M.2 Key M (NVMe) and M.2 Key E (Wi-Fi) |
Power and Physical
| Power Consumption |
40W to 130W (configurable) |
| Power Supply Included |
140W AC adapter |
| Weight |
Approximately 1.94 kg |
| Cooling |
Heatsink and active fan |
Software and Ecosystem
| Operating System |
Ubuntu 24.04 LTS (pre-loaded) |
| Software Stack |
JetPack 7 SDK (includes CUDA, TensorRT, cuDNN, VPI, OpenCV) |
| AI Frameworks |
NVIDIA Isaac (robotics), Metropolis (visual AI), Holoscan (sensor processing) |
| Model Support |
Isaac GR00T N (VLA models), LLMs, VLMs, generative AI models |
| Container Support |
Docker with NGC container registry integration |
Specifications are sourced from the official NVIDIA developer page and may be updated. The NVIDIA Jetson AGX Thor Developer Kit is designed for developers, researchers, and professionals building advanced AI and robotics solutions. For Australian power requirements, ensure you select the ANZ region-specific model (945-14070-0085-000 or 945-14070-0086-000).
