NVIDIA has officially launched the Jetson AGX Thor developer kit and Jetson T5000 module, a potent new tier of robotics computing designed to bring advanced AI — including generative reasoning and real-time multimodal processing — directly to robots and edge devices.
Built on the Blackwell GPU architecture and featuring a massive 128 GB memory layout, Jetson Thor delivers up to 2,070 FP4 TFLOPS of AI performance, all within a 130 W power envelope — around 7.5× more compute and 3.5× better energy efficiency than its predecessor, Jetson AGX Orin.
Leading the Robotics Revolution
This platform is designed to power real-time reasoning and fast interactions in robotics. Major robotics innovators like Agility Robotics, Boston Dynamics, Meta, Amazon Robotics, Caterpillar, Medtronic, and Figure are already early adopters. Meanwhile, John Deere and OpenAI are evaluating the system for future use.
For instance, Agility Robotics plans to integrate Jetson Thor into its next-gen digit robot to enhance real-time perception and decision-making. Boston Dynamics will bring server-grade AI compute directly on Atlas using Jetson Thor.
Key Specifications & Capabilities
| Feature | Specification |
|---|---|
| Compute Power | Up to 2,070 FP4 TFLOPS (2070 TFLOPS) |
| GPU Architecture | Blackwell with 2560-core GPU, 96 5th-gen Tensor Cores, MIG support |
| CPU | 14-core ARM Neoverse-V3AE (up to 2.6 GHz) |
| Memory & Bandwidth | 128 GB LPDDR5X @ 273 GB/s |
| Networking | 4× 25 GbE + 1× 5 GbE, high-speed sensor data fusion |
| Video/Camera Support | Multi-4K/8K encode-decode, supports up to 20 cameras via high-speed interfaces |
| Power Envelope | 40 W–130 W configurable, optimized for edge deployment |
| Software Ecosystem | Fully compatible with Isaac, Metropolis, Holoscan platforms for robotics and AI workflows |
Market Reaction & Coverage
NVIDIA frames Jetson Thor as the supercomputer for “physical AI” — a leap toward fully intelligent robots capable of nuanced interaction at the edge.
Tech media calls it the most powerful mini-PC yet — tailored not for gamers, but for developers pushing physical intelligence and robotics forward.TechRadar
Final Thoughts
With the Jetson Thor debut, NVIDIA is elevating robotics — delivering vast, energy-efficient AI compute right where it’s needed: in devices, machines, and robots in real-world environments. This brings us closer to responsive, intelligent agents shaping how industries operate and people interact with machines.