by Nathan Pulsone with Dan Mandell | 06/10/2025

AI Mini SBCs

In recent years, the growing demand for embedded AI has extended to miniaturized single board computers (SBCs), prompting hardware platform providers to develop specialized “AI Mini SBCs,” tailored for neural network inference and other workloads. Board suppliers are teaming up with a variety of embedded processor and accelerator technology providers to facilitate the development of commercial edge AI solutions. For example, the advanced AI capabilities and small form factor of devices such as Thundercomm’s Rubik Pi 3, IMDT’s V2N SBC, and the Raspberry Pi 5 with AI Kit offer flexible and competitive hardware solutions for edge AI product development.

Within the past year, several new AI Mini SBC products have emerged. Thundercomm’s Rubik Pi 3 is one such example. Powered by a Qualcomm Dragonwing processor capable of up to 12 TOPS, the board can run various AI models, such as for classification, object detection, and image segmentation, without the need for cloud connectivity. The Rubik Pi 3 can power a camera with a vision language model, performing face detection and gesture recognition. As a result, it can serve as a development platform for edge AI automation and monitoring solutions across a range of applications in industrial, medical, and retail spaces. Although the board is not explicitly designed for deterministic operations, applications needing precise timing can be enabled with real-time extensions in corresponding software supported by the Rubik Pi.

Earlier this year, IMDT introduced its latest Mini SBC, which features a Renesas RZ/V2N processor and enables up to 15 TOPS of integrated AI performance. The V2N SBC supports two MIPI CSI cameras with integrated image signal processing and features various sensors, such as an internal measurement unit (IMU) and a magnetometer for motion detection. Target applications include industrial vision, security, and surveillance. Beyond AI vision applications, the V2N SBC can also support anomaly detection in smart sensor networks, making it a bit more versatile than other popular Mini SBCs.

The Raspberry Pi 5, which is now nearly 2 years old, continues to evolve with the recently released AI Kit, featuring a Hailo M.2 acceleration module capable of 13 TOPS. AI applications for the Raspberry Pi 5 with AI Kit match the scope of the Rubik Pi and V2N SBC. With the rpicam application in the Raspbian OS natively supporting the AI module, it may be used in applications such as car and motion detection. As a long-standing leader in the small form factor prototyping and computing market, Raspberry Pi has released numerous solutions since 2012, illustrating the rapid development of Mini SBC capabilities and their expansion into new edge AI applications.

Sitting around the 10 to 15 TOPS range, applications for AI Mini SBCs are currently focused on vision processing in IoT use cases. However, there are still many other potential deployments for the form factor. Despite many deployments for AI on compact embedded systems involving some degree of cloud computing, AI Mini SBCs are already able to fulfill computing needs on the edge. These compact single-board devices offer competitive hardware solutions for IoT product development, while simplifying AI implementation with the availability of open source and proprietary application development resources, software tools, model support, and sensor compatibility

The growing space for AI Mini SBCs will be analyzed more deeply within VDC’s upcoming 2025 Embedded AI Hardware report.