Google Astra Chip: 40% Less Power for Edge AI WearablesTechnology

<h2>Always-on AI ambition</h2> Avinatan Hassidim, vice president of Google Research, said conventional edge chips were slow and used too much energy, making an always-on ambient AI experience impossible. The new chip uses the M2, a successor to the Coral Neural Processing Unit, an AI compute engine Google introduced last year. The chip is expected to handle voice, video, motion and context detection in real time on wearable devices, allowing phones and glasses to understand their surroundings continuously without constantly waking power-hungry main processors.

Rahul Patel, the chief executive of Synaptics, said the chip would not stay in the laboratory and would be built into Google devices, although he did not specify which products or when they would be released. Synaptics plans to reveal detailed specifications and demonstrate the chip at its own event on October 20. The partnership extends the two companies&#8217; earlier collaboration: Synaptics unveiled the first Astra chip series last October, the first to commercialise Google&#8217;s Coral neural processing platform, and earlier this year built the Coral Board development platform around it.

At the same event, Google Research also shared other projects, including Google Earth AI for predicting environmental and public health conditions, Deep Review for checking the logic and factual accuracy of technical documents, and TimesFM, a forecasting model for time-series data. The Astra announcement reflects a broader industry shift toward on-device intelligence, as companies race to put AI assistants into wearables that users can wear all day without draining the battery.