Jetson context
The NVIDIA Jetson family is born in 2014 as a high-end embedded AI solution (Jetson TK1, TX1, TX2, Xavier) priced at 200-700 USD, oriented to robotics and industrial products. In 2019 NVIDIA decides to make its CUDA stack accessible at entry level too.
The launch
Jetson Nano Developer Kit is announced at GTC 2019 on 18 March 2019 at $99. First shipments April 2019.
Specs:
- NVIDIA Tegra X1 — quad-core ARM Cortex-A57 at 1.43 GHz
- NVIDIA Maxwell GPU with 128 CUDA cores
- 4 GB LPDDR4 RAM
- microSD storage
- 4× USB 3.0, Gigabit Ethernet
- HDMI 2.0 + DisplayPort
- 40 GPIO pins (Raspberry Pi-compatible layout)
- MIPI CSI-2 for 2 cameras
- M.2 Key E for Wi-Fi (Bluetooth sold separately)
- 5V 4A power (5-10 W consumption)
The software stack
The strong suit is software: Jetson Nano supports the same CUDA stack as NVIDIA servers:
- CUDA 10.x (later 11.x)
- cuDNN — neural network library
- TensorRT — inference optimisation
- DeepStream SDK — real-time AI video pipeline
- NVIDIA JetPack — complete SDK (OS + libraries)
Binary compatibility with models trained on desktop/datacenter GPUs, only with quantisation/optimisation to be efficient on 128 CUDA cores.
Use cases
Jetson Nano is used in:
- Edge computer vision — person/object recognition on IP cameras
- Robotics — SLAM, navigation, drone navigation (ROS + Jetson)
- Smart retail — customer traffic, heatmaps, auditing
- Agriculture — computer vision for agricultural robotics, crop classification
- Industry 4.0 — inline visual inspection and quality control
- Educational — university deep learning courses with real hardware
Variants
The Jetson line has expanded:
- Jetson Xavier NX (2020) — Volta GPU, 21 TOPS
- Jetson Nano 2GB (2020) — reduced version at $59
- Jetson AGX Orin (2022) — Ampere GPU, up to 275 TOPS
- Jetson Orin Nano (2023) — Nano successor, 20-40 TOPS, $499
- Jetson Orin NX (2023) — 70-100 TOPS
In the Italian context
Jetson Nano has been the preferred platform for:
- University theses on applied deep learning (Polytechnics, Universities)
- Italian robotics startups for embedded deep learning prototypes
- Industrial computer vision — Italian integrators use it for rapid prototypes before industrial deploy
- noze has used Jetson Nano and later Jetson Orin Nano for R&D projects in digital health (preliminary diagnostic imaging) and industrial computer vision
References: NVIDIA Jetson Nano Developer Kit (GTC 2019, 18 March 2019). NVIDIA Tegra X1. Maxwell GPU 128 CUDA cores. 4 GB LPDDR4. $99 launch price. JetPack SDK. CUDA 10.x, cuDNN, TensorRT. Successors: Xavier NX, AGX Orin, Orin Nano, Orin NX.