Edge AI moves inference out of the data center and onto gateways, industrial PCs, cameras and even microcontrollers. The model is trained elsewhere, then deployed to hardware that sits close to the sensor. The result is lower latency, less upstream bandwidth, and the ability to keep working when the connection drops. It also keeps raw data, such as video of a shop floor, on the premises.
The cost is that edge hardware has limited memory and compute, so models are usually compressed and accelerated with an NPU or a GPU. A camera that counts people locally and reports only a number is a common example. Our guide Edge AI explained covers the hardware tiers, from TinyML on microcontrollers to server-class edge boxes.
Part of the IoT Glossary. Spotted an error or a missing term? Tell us.