Coral Accelerator Module
from Google
Murata and Google develop the world's smallest AI module


Accelerating the algorithms essential to AI


Google combined Murata’s compact package technology and Google’s own Edge TPU chip enabling power-efficient and high-performance machine learning inference to create a new compact AI module.
Enclosing Google’s Edge TPU ASIC in a package with a smaller footprint simplifies the design of the printed circuit board.


  • Supports operation at up to 4 TOPS (trillion operations per second).
  • Boosts power efficiency by ensuring power consumption of only 0.5 W per 1 TOPS.
  • Supports PCIe Gen 2 and USB 2.0 interfaces.

The Accelerator Module is a multi-chip module (MCM) integrating the Edge TPU and exclusive power control technology. Compactness is key to achieving extremely robust functionality in devices with limited space, because it is essential to optimize use of all space on circuit boards.

Designed by Google, the Edge TPU is a compact ASIC that speeds up model estimation times while using very little power. Murata’s compact package technology and power supply design (noise suppression) make it possible to achieve a high-performance, compact product that takes up very little space.

Model estimation time comparison

When implemented on the latest embedded CPU or the Coral Dev Board, a comparison of model estimation times using TensorFlow Lite format shows that inference time is substantially reduced and confirms that genuinely high performance can be achieved.

About Coral

Coral is a local AI platform that helps bring on-device AI application ideas from prototype to production. Coral offer a platform of hardware components, software tools, and pre-compiled models for building devices with local AI. creating a flexible development system that makes it easy to grow embedded AI products into reality. By working with creators, designers, engineers, manufacturers, and industry, Coral is helping to build truly beneficial AI for our world.

  • Object detection

    Draw a square around the location of various recognized objects in an image.

  • Pose estimation

    Estimate the poses of people in an image by identifying various body joints.

  • Image segmentation

    Identify various objects in an image and their location on a pixel-by-pixel basis.

  • Key phrase detection

    Listen to audio samples and quickly recognize known words and phrases.

Learn more, at Open in a new window

Usage settings

  • Manufacturing

    Quality control, safety monitoring, predictive maintenance, etc.

  • Healthcare

    Patient care, medical image generation,
    low-cost diagnosis, home care, etc.

  • Agriculture

    Soil analysis, sorting produce, disease detection,
    precision agriculture, etc.

  • Automotive

    Safe driving assistance, status monitoring, seamless control of vehicle systems, operation verification, etc.

System configuration

[Using pre-compiled models]

  • Either pre-compiled machine learning models or transferred learning data utilizing such models is available.

[Using custom models]

  • Classroom data is collected using automated machine learning (AutoML) and output in a binary format compatible with Coral and Edge TPU.
  • Existing machine learning model workflows are used to generate TensorFlow Lite data.

Basic specifications

  • Fast and energy-efficient learning inference (4TOPS@2W)
  • Surface-mount module incorporating Edge TPU and PMIC
  • Product size: 15.00 × 10.00 × 1.5 mm
  • Interfaces: PCle and USB 2.0
  • RoHS compliant
Download the datasheet here Open in a new window


Use the Google Coral form to make inquiries about the solution described above.

Note: Murata is unable to provide product support or respond to questions regarding the product.

For more information, contact Coral sales.

Inquiry form Open in a new window

Solution Technologies