The chips are designed to support applications where power consumption and form factor constraints have limited the deployment of always-on computer vision.

Mosaic SoC said its architecture uses a multi-core design to improve performance per watt, enabling continuous perception capabilities such as object tracking, spatial mapping and event detection while maintaining low energy use. 

The system can create real-time maps of environments and track objects within them, allowing devices to perform tasks such as identifying when specific conditions are met or recalling where items were last observed. 

In addition to hardware, the company is developing software tools and application layers that allow device manufacturers to integrate perception capabilities without building systems from scratch.

While initially targeted at consumer electronics, the approach reflects broader interest in embedding intelligence directly at the edge, reducing reliance on centralized computing and enabling faster, more autonomous decision-making in connected systems.

Mosaic SoC was founded by researchers from ETH Zurich and plans to expand from chip development into a broader platform supporting deployment and optimization of spatial intelligence applications.