Coral Edge TPU

Google's custom ASIC for high-performance, low-power machine learning inference at the edge: 4 TOPS at 2 TOPS/W.

The Coral Edge TPU is Google's custom Application-Specific Integrated Circuit (ASIC) designed for high-speed, power-efficient machine learning inference right on the device. This small accelerator delivers 4 trillion operations per second (TOPS), operating at an exceptional 2 TOPS per watt . It specifically accelerates TensorFlow Lite models, enabling on-device AI for tasks like real-time object detection and computer vision without relying on a constant cloud connection . Deploy this capability via the Coral Dev Board, a USB Accelerator, or a System-on-Module (SOM) for rapid prototyping and seamless production integration .

https://coral.ai