News

Google's New Chip Designed for AI (and IoT) at the Edge

• By Becky Nagel
• 08/15/2018

Google recently announced Edge TPU, an application-specific integrated circuit (ASIC) designed to do exactly what its name suggests: run AI "on the edge" -- something often needed for large machine learning and training projects, as well as time-sensitive Internet of Things (IoT) projects.

Edge computing is a method of distributed computing designed to achieve increased real-time performance and bypass bottlenecks by having the processing take place as close to where the data is stored as possible, instead of first moving the data to, for example, a centralized cloud location. It is often seen as an ideal solution for high-data projects with intense and time-sensitive processing.

"With the explosive growth of connected devices, combined with a demand for privacy/confidentiality, low latency and bandwidth constraints, AI models trained in the cloud increasingly need to be run at the edge," Google commented on the chip's announcement Web site. "It delivers high performance in a small physical and power footprint, enabling the deployment of high-accuracy AI at the edge."

According to Google, the Edge TPU's open nature complements a number of uses and architectures: "It augments Google's Cloud TPU and Cloud IoT to provide an end-to-end (cloud-to-edge, hardware + software) infrastructure."

Google noted that when Edge TPU is released (no date was given), it will be designed to work with the company's open source TensorFlow Lite and will be supported by its Cloud IoT Edge product.

Interested parties can request early access (currently underway) here.