Google Astra Chip: Low-Power Edge AI for Wearables Debuts

The google astra chip made its debut at the company’s annual Research@ event in Mountain View, California, on October 14 — a new line of low-power edge AI chips co-developed with Synaptics aimed at bringing always-on artificial intelligence to wearable devices.

The chip uses the M2, a successor to Google’s Coral neural processing unit, and consumes about 40% less energy than existing products. Rather than sending data to the cloud, edge AI runs computations directly on the device — cutting response times and reducing the risk of data leaking in transit.

Why Edge AI Needs a New Kind of Chip

Google Research VP Avinatan Hassidim said conventional edge chips are simply too slow and too power-hungry for the kind of always-on ambient intelligence the company envisions. The Astra line is designed to handle voice, video, motion and context detection in real time, directly on wearables.

Synaptics CEO Rahul Patel said the chip will be built into Google devices, though he offered no specifics on which devices or when they would arrive.

Synaptics to Detail Specs October 20

Full technical details are still to come. Synaptics plans to reveal the chip’s detailed specifications at its own event on October 20, which should show how far the collaboration pushes performance-per-watt beyond the Coral generation.

If Astra delivers on its promises, it could make always-listening wearable AI practical without the battery drain and privacy trade-offs that have held the category back.

Source: CHOSUNBIZ