Google Research unveiled a next-generation Astra AI edge chip developed with Synaptics, designed to dramatically cut power use for on-device AI processing.

The Coral NPU M2 variant is claimed to consume about 40% less energy than existing edge AI chips, enabling an ambient, always-on AI experience on wearables.

Google says the Astra family uses roughly 40% less energy than current edge AI products, addressing heat and power challenges in always-on devices like smartphones and wearables.

The article notes the content was AI-translated and invites reader feedback on the translation.

The chip is designed to process real-time, on-device data such as voice, video, motion, and context detection, reducing latency and data leakage risk.

By handling on-device processing for voice, video, motion, and context detection, the chip minimizes data transmission and strengthens privacy.

In a related move, Google announced the public availability of DeepReview, an agent that verifies research papers by checking logical consistency and mathematical propositions and offering improvements.

Rahul Patel of Synaptics said the chip will move beyond lab testing and be embedded in Google devices, with Synaptics set to reveal detailed specs at its event on the 20th.

Synaptics planned to unveil detailed specifications and live demonstrations at its event on the 20th.

Synaptics CEO Rahul Patel indicated the chip will be built into Google devices, though exact devices and release timing were not disclosed.

Google frames this as part of expanding the Edge AI market, complementing its Tensor Processing Units for data centers and signaling broader ecosystem development.

The Coral NPU platform previously powered the Coral Board, and Synaptics plans to release the Astra series version 2 on the 20th of the month.