Google used Day 2 of its Cloud Next conference to roll out its latest products for leveraging data directly at the edge:The Edge TPU chip, Cloud IOT Edge software, and a development kit for the Edge TPU.
These products, according to Injong Rhee, VP of IOT Google Cloud, make Google the only cloud service provider with integrated software and a custom hardware stack for implementing machine learning IOT in the cloud and at the edge.
The Edge TPU is a purpose-built accelerator chip designed to run TensorFlow Lite machine learning models at the edge. It's so small that four of the chips can fit onto one penny, Rhee demonstrated on stage.
"Edge TPU will bring a brain to your embedded devices at extremely low cost and power efficiency without compromising performance," Rhee said. "I believe this is a game changer."
Clout IOT Edge is the software to run with the Edge TPU. It consists of two components: Edge IOT Core gateway functions and Edge ML, a TensorFlow Lite-based runtime for running models on edge devices. It can run on Android Things or Linux OS-based devices.
Lastly, the development kit includes a system on module (SOM) that combines the Edge TPU, a NXP CPU, wi-fi and Microchip's secure element. This will be available to developers in October.