Google Research and Synaptics have announced the new Coralboard edge AI platform to enable on-device artificial intelligence processing. This compact single-board computer runs machine learning models locally without relying on cloud connectivity. Consequently, developers can build private and power-efficient applications for small devices.
Hardware Specifications and Architecture
The hardware is powered by the 2GHz Synaptics Astra SL2619 dual-core system-on-chip. Notably, this processor includes 2GB DDR4 DRAM and a 1 TOPS neural processing unit subsystem. This architecture prioritizes matrix operations over scalar compute to run neural networks efficiently.
In addition, the board supports CSI camera input, DSI display connectivity, RGB LEDs, microphones, and a buzzer. Developers can also expand the platform using an M.2 module for wireless connectivity. These physical interfaces make the board suitable for smart devices and robotics.
Coralboard edge AI Capabilities
Billy Rutledge, Director at Google Research, stated that the platform simplifies bringing advanced machine learning to the physical world.
“By pairing efficient, AI-native hardware with an open software stack, this platform lowers the barrier to building private, always-on experiences that run all day in power-constrained environments.”
Billy Rutledge, Director at Google Research
Software Integration and Developer Tools
Specifically, the platform integrates Google’s Coral NPU technology to support the Gemma 3 270M open-source model. This Coralboard edge AI integration allows developers to run artificial intelligence workloads directly on the hardware. Meanwhile, the open-source compiler and runtime tools help optimize these models for low-power environments.
During the Google I/O 2026 event, the companies demonstrated these capabilities through an installation called Jellectronica. This setup used a YOLOv8 object detection model to track jellyfish movements from a live stream. As a result, the system converted motion data into real-time music using generative audio models.
Future Outlook and Availability
While a limited-edition version was distributed to event attendees, general availability of the Coralboard edge AI platform will begin this summer. Furthermore, the partners plan to announce pricing and wider distribution details later in 2026. This release aims to support developers building applications for wearables, smart home appliances, and industrial automation.
Notably, this collaboration between Google Research and Synaptics represents a shift toward open standards in hardware development. By open-sourcing the underlying architecture, the companies aim to reduce fragmentation in the hardware market. Consequently, developers can avoid proprietary toolchains when building intelligent, context-aware devices.
Source: X (@googlegemma) — researched
