Search  for anything...

NVIDIA Jetson Thor Developer Kit

Based on 33 reviews
Condition: New
Checking for the best price...
$5,283.26
Holiday Deal · 4% off was $5,499.00

Buy Now, Pay Later


As low as / mo
  • – Up to 36-month term if approved
  • – No impact on credit to apply
  • – Instant approval decision
  • – Secure and straightforward checkout

Ready to go? Add this product to your cart and select a plan during checkout.

Payment plans are offered through our trusted finance partners Klarna, Affirm, Afterpay, Zip, Apple Pay, and Google Pay. No-credit-needed leasing options through Acima may also be available at checkout.

Learn more about financing & leasing here.

Free shipping on this product
This item's return window has been extended for the holiday season: FREE refund/replacement until Jan 31, 2027

To qualify for a full refund, items must be returned in their original, unused condition. If an item is returned in a used, damaged, or materially different state, you may be granted a partial refund.

To initiate a return, please visit our Returns Center.

View our full returns policy here.


Availability: Only 3 left in stock, order soon!
Fulfilled by Amazon

Arrives Saturday, Nov 21
Order within 20 hours and 49 minutes

Available payment plans shown during checkout

Protect Your Purchase

Easy claims online, fast replacements, no fees

Checking for protection plans...

At a glance

Condition
New
Model
945-14070-0080-000
Weight
6.5 lb
Graphics Coprocessor
NVIDIA Blackwell GPU
Graphics Ram Size
128 GB
GPU Clock Speed
1.57 gigahertz
Video Output Interface
DisplayPort, HDMI
Warranty
1 year warranty for development use only
Works with
Humanoid Robots, Specialized AI/robotics Systems

Features 2

  • GPU:2560-core NVIDIA Blackwell architecture GPU with 96 fifth-gen Tensor Cores
  • AI Performance:2070 TFLOPS

Specs 23

Graphics Coprocessor NVIDIA Blackwell GPU
Brand NVIDIA
Graphics Ram Size 128 GB
GPU Clock Speed 1.57 gigahertz
Video Output Interface DisplayPort, HDMI
Graphics Card Ram 128 GB
Graphics Ram Type GDDR6X
Graphics Card Interface PCI-Express x16

Frequently asked questions

If you place your order now, the estimated arrival date for this product is: Saturday, Nov 21

Yes, absolutely! This item's return window has been extended for the holiday season: FREE refund/replacement until Jan 31, 2027.

To initiate a return, please visit our Returns Center.

View our full returns policy here.

  • Klarna Financing
  • Affirm Pay in 4
  • Affirm Financing
  • Afterpay Financing
  • Zip Pay in 4
  • Financing through Apple Pay
  • Financing through Google Pay
Leasing options through Acima may also be available during checkout.

Learn more about financing & leasing here.

Top Amazon Reviews


Absolute beast | Couldn’t be more happier
Absolute beast, flashing OS was easier than I thought. I’m hosting Nemotron Nano 3 30b 8bit quantized and getting 53 tokens/sec.
Reviewed in the United States on October 6, 2026 by Arpan Pathak

Slow bandwidth, still amazing
Writing this June 23rd 2026. Anemic memory bandwidth, calibrate your expectations accordingly. It is still amazing to see things running locally and doing real work pretty much immediately.
Reviewed in the United States on June 23, 2026 by Gary

One to rule them all
If you are looking for a great server to help with your AI tasks, this is the one. This is the best of the best. Read the specs. The performance is a great value for your money. Nvidia did most of the heavy lifting for you. If you know what you doing, get this one. This is one of those if you know you know situations. ... show more
Reviewed in the United States on May 20, 2026 by Chris Lewis

Nvidia’s best
Amazing product! Arrived fast with no issues
Reviewed in the United States on July 6, 2026 by Amazon Customer

Expensive junk that does not work.
Does not work, flashing does not work, libraries don't work. This is an overpriced, useless junk. Get Mac mini, or Mac Studio, or just build a MicroATX, it will work better and faster.
Reviewed in the United States on April 11, 2026 by Gene

Whack-a-mole!
Observed behavior: - usbipd “endless attach loop” on a BUSID is insufficient because BUSID changes when the USB identity changes. - Device may vanish from Connected list and remain only as Persisted, indicating the USB device is not enumerating on the host (cable/port/power management sensitivity). - Host component installs (Nsight Perf SDK) fail due to missing utilities (unzip) or symlink collisions (file exists), creating false “installation failed” signals unrelated to the device wipe objective. ... show more
Reviewed in the United States on January 10, 2026 by Kim

It works
This device is taken by those who know why they take it. So the review “it works”.
Reviewed in the United States on April 20, 2026 by Kirill Keker

Incomplete disorganized documentation, SDK lackluster
Disclaimer: I know this is intended for robotics with companies like Anduril, Lockheed, Raytheon, etc. likely being the target demo but the marketing copy had me convinced this would be an easy button for running containers with CUDA and unified memory. Well, that’s only true if you like running one-off commands from CLI or using desktop GUIs. Don’t get your hopes up if you use containers, unless you love “docker run” syntax and don’t care about security. The documentation is all over the place. Important repos disappear from their site, for which you then have to find a random mirror some private citizen put on GitHub and pray it works. The SDK (Jetpack) and jetson-containers project are designed to wrap docker, NOT podman|systemd|kubernetes|etc. so if your preference is not docker or baremetal be prepared to suffer. “Jetson OS” (Linux for Tegra or L4T rebranded) is built on Ubuntu LTS 22.04 with so many outdated packages for container support - be prepared to build low level dependencies and common tool chains from source while spending hours stumbling through Nvidia’s forums to find a stray comment from an employee that offers a clue to resolving a permissions issue introduced by an undocumented security patch. Hey Nvidia I want a functional declarative container stack, not being limited to “docker run” please. dusty-nv’s “jetson-containers” repo is cool for finding prebuilt container images designed to work with the CUDA samples already in place for the T5000/Blackwell, but I don’t want to use DOCKER! Let alone run on bare metal... You need to improve general LXC support and keep the repos up to date with latest dependencies and toolchains. At the very least you should ship your own flavor of minikube or similar with jetson-containers bundled, to keep the official packages from getting trampled while us users hack around the broken pieces… I should be able to copy my existing declarative container configs (YAML|TOML|JSON, k8s for compose|quadlet|podlet, don’t care which) to this machine and run them with minimal adaption necessary. The Nvidia hooks/wrappers and syntax for using the CTK need to be collocated and/or config paths clearly documented in one KB page. Overall, if you buy this instead of a MINISFORUM/Beelink/Mac Studio, you’re paying a premium for CUDA support that’s only officially offered two ways: Baremetal or Docker (just Docker-anything else you’re on your own). The added complexity of ARM builds always being an afterthought for many projects is almost enough to make me wish I did a custom multi-4090 GPU x86 rig instead. P.S. Some announcement about who is taking over dusty-nv’s work and what the future of Jetson looks like would be nice. ... show more
Reviewed in the United States on January 8, 2026 by FFW
Checking for best price...