How to Set Up an NVIDIA Jetson Nano

The NVIDIA Jetson Nano is a small single-board computer with a GPU, which makes it a popular choice for running computer vision and machine-learning models at the edge. Setup is different from a microcontroller because you flash a full Linux image to an SD card and boot into a desktop. If you would rather have the Jetson plus a power supply, camera, and a tested project arrive together, Compoden's AI build assistant turns a described project into a parts list, wiring, and code, then ships the kit. Here is the manual setup.

What you need

You need an NVIDIA Jetson Nano 4GB developer kit, a microSD card of at least 32GB (64GB is more comfortable), a 5V supply that can deliver enough current (a 5V 4A barrel-jack supply is recommended for heavier loads), an HDMI monitor, and a USB keyboard and mouse. A computer with an SD card reader is needed to write the image.

Download the JetPack SD card image

On your computer, go to NVIDIA's Jetson Nano developer kit page and download the official JetPack SD card image for the Jetson Nano. JetPack bundles the Linux for Tegra operating system along with CUDA, cuDNN, and TensorRT, which are the libraries your AI projects will use. The download is a large zip file, so allow time on a typical Indian broadband connection.

Flash the image to the SD card

Install balenaEtcher, a free tool for writing disk images. Insert the microSD card, open Etcher, select the JetPack image (Etcher reads the zip directly), select your SD card as the target, and click Flash. This erases the card and writes the image, which takes several minutes. Do not unplug the card until Etcher reports it has finished and validated.

First boot

Insert the flashed card into the slot on the underside of the Jetson module. Connect the monitor, keyboard, and mouse, then apply power. For the barrel-jack supply you must fit the J48 jumper to switch the board off USB power. On first boot you accept the licence, choose a language and keyboard layout, create a username and password, and the system finishes setting up the Ubuntu-based desktop.

Verify the setup

Open a terminal once the desktop loads. Run nvidia-smi is not used on the Nano; instead check the JetPack version with sudo apt-cache show nvidia-jetpack, and confirm the GPU libraries are present. You can also run jetson_release if the jetson-stats tool is installed. These confirm CUDA and the L4T release are in place before you load a model.

If it does not boot

A blank screen on first boot is most often a power problem. Use a quality 5V supply and the J48 jumper for the barrel jack rather than relying on micro USB for demanding work. If the board cycles power, the supply cannot deliver enough current. If you see no display at all, reflash the SD card, since an interrupted write leaves the image incomplete.

Build it with Compoden's AI

Describe your project to Compoden's AI build assistant, for example a camera that counts people or a model that sorts objects, and it picks the Jetson Nano plus the camera, power supply, and storage you need, lays out the connections, and provides starter code and setup notes for JetPack. The kit ships to your address in India, so you spend your first session on the model rather than on sourcing parts.

Tell Compoden's AI assistant what you want to build and get a Jetson Nano kit prepared for it. You can also browse the Jetson Nano 4GB or the wider catalog.

FAQ

What size SD card do I need? Use at least 32GB, though 64GB gives more room for models and datasets. Use a fast card, UHS-1 or better, for a smoother experience.

Why won't my Jetson Nano power on? Most boot failures are power related. Use a 5V supply rated for enough current, and fit the J48 jumper when using the barrel jack instead of micro USB.

What is JetPack? JetPack is NVIDIA's software stack for Jetson. The SD card image includes the Linux for Tegra OS plus CUDA, cuDNN, and TensorRT for AI workloads.