Presentation of an open modular platform for creating Coral NPU AI accelerators
11:50, 17.10.2025
Google's open Coral NPU platform was recently unveiled. It offers access to an open hardware accelerator for machine learning models and the necessary tools for using it with AI engines.
Features of the Coral NPU platform
This project can be used to create more energy-efficient systems on a chip for edge computing, the Internet of Things, and low-power devices such as smart watches and headphones.
The NPU uses a 32-bit architecture, AXI4 bus, and a four-stage instruction pipeline. The processor is equipped with 8 KB ITCM and 32 KB DTCM memory and supports SIMD operations.
With the Coral NPU, you can guarantee the continuous operation of AI applications with the lowest possible power consumption. The basic implementation guarantees performance of up to 512 GOPS and power consumption of several milliwatts.
Synaptics will be the first manufacturer, and the company has already announced a line of processors specifically designed for Astra SL2610 devices.
In addition, the NPU can be used for context awareness, audio, and image processing. It enables the functioning of applications on devices for object recognition, language models, gesture control, visual search, user activity detection, and much more.