NVIDIA Jetson AGX Thor Review: Specs and Uses of the Robot AI Computer
Updated: 7 days ago
The NVIDIA Jetson AGX Thor is an embedded computer made to run AI models inside robots. The manufacturer does not publish a price. This review covers its specifications, features and uses.
Quick Verdict
Jetson Thor is the computer NVIDIA built for humanoids and other robots that run large AI models on board. The T5000 module in the developer kit delivers 2070 FP4 TFLOPS with a Blackwell-generation GPU of 2560 cores, a 14-core Arm Neoverse-V3AE processor and 128 GB of memory, drawing 40 to 130 W. That memory is what allows vision-language-action models to run locally without a network link. A smaller T4000 module offers 1200 TFLOPS and 64 GB at 40 to 70 W. The page does not list a price; the kit is sold through partners.
NVIDIA Jetson AGX Thor at a Glance
AI compute: 2070 FP4 TFLOPS
Memory: 128 GB LPDDR5X
CPU: 14-core Arm Neoverse-V3AE
GPU: 2560-core Blackwell
Power: 40 W to 130 W
Price: on request
NVIDIA Jetson AGX Thor Technical Specifications
The table below lists what the manufacturer states.
Specification | NVIDIA Jetson AGX Thor |
Manufacturer | NVIDIA |
AI compute | 2070 FP4 TFLOPS |
Memory | 128 GB LPDDR5X |
CPU | 14-core Arm Neoverse-V3AE |
GPU | 2560-core Blackwell |
Power | 40 W to 130 W |
Memory bandwidth | 273 GB/s |
T4000 module | 1200 TFLOPS; 64 GB; 12-core CPU; 40 to 70 W |
The developer kit price is not listed on the page; it is sold through partners.
Design and Build Quality
The developer kit packages the T5000 module with cooling and ports for prototyping; production robots embed the module on their own carrier board.
Software and Integration
Software Stack
The platform runs NVIDIA's Jetson software with the Isaac robotics libraries.
Models On Board
Large memory allows foundation models to run on the robot.
NVIDIA Jetson AGX Thor Applications and Use Cases
Humanoid Robots
On-board reasoning and control.
Mobile Manipulators
Perception and planning without the cloud.
Industrial Vision and Inspection
Many camera streams with AI at the edge.
NVIDIA Jetson AGX Thor Price and Value
NVIDIA does not list a price for the Jetson AGX Thor developer kit on the product page; it is sold through partners.
For the company behind the product, read the NVIDIA Robotics Company Profile: Isaac Platform, Structure, Financials, Leadership and Stock.
Pros and Cons
Pros
2070 TFLOPS on board
128 GB of memory
Scales from 40 to 130 W
Smaller T4000 option
Large software ecosystem
Cons
Up to 130 W needs cooling and battery budget
Price not on the page
Tied to NVIDIA's software stack
No published price
Who Should Buy the NVIDIA Jetson AGX Thor?
It fits robot makers and research teams that run large AI models on the machine itself.
NVIDIA Jetson AGX Thor FAQ
How much does the NVIDIA Jetson AGX Thor cost?
NVIDIA does not list a price for the Jetson AGX Thor developer kit on the product page; it is sold through partners.
What is the NVIDIA Jetson AGX Thor?
An embedded AI computer for robots.
How powerful is the Jetson AGX Thor?
2070 FP4 TFLOPS of AI compute.
How much memory does the Jetson AGX Thor have?
128 GB of LPDDR5X.
What CPU does the Jetson AGX Thor use?
A 14-core Arm Neoverse-V3AE.
How much power does the Jetson AGX Thor use?
40 W to 130 W.
What is the Jetson T4000?
A smaller module with 1200 TFLOPS, 64 GB and 40 to 70 W.
What is Jetson Thor used for?
Humanoid robots, mobile manipulators and edge AI.
Specifications are based on manufacturer information and may change.
