[Editor's Note] Edge AI technology is emerging as a key competitive edge in the rapidly changing smart factory market. At the 2025 Smart Factory & Automation Industry Exhibition, STMicroelectronics presented an edge AI demo using the STM32N6 MCU that meets these market demands. The object detection and path tracking gimbal demo and the Auto Wake Up demo provided a glimpse into the practical applicability of edge AI technology.

We looked into the products and technology trends exhibited at this event through Hyunsoo Moon, Manager at STMicroelectronics, who is at the forefront of edge AI technology.
■ What are the market trends and key requirements?

There is increasing demand for performance to process increasingly complex AI models in real time at the edge. The market is demanding high-performance edge computing hardware that can quickly execute vision AI models or voice recognition models on edge devices.

It also requires technology to minimize power consumption while maintaining high-performance operations, various interfaces and compatibility such as MIPI CSI-2, USB, and Ethernet, and strong security functions.

To meet these market and industry requirements, the STM32N6 accelerates engineering and AI development with: △High-performance edge computing based on Cortex-M55 △600 GOPS NPU based on ST Neural-ART △Low-power design △Various interfaces support △Strong security features △ST’s comprehensive ecosystem.

■ Introduction to this exhibition demo

ST showcased two Vision AI demos based on the new STM32N6 MCU, a new product featuring ST’s self-developed neural network hardware accelerator. The NPU on the STM32N6 MCU delivers 600 GOPS of performance and can accelerate edge-level inference of neural network models such as YOLO even on compact embedded platforms based on the ARM Cortex-M55.

The first is a gimbal demo that implements object detection and human path tracking using Kalman filter, processed on STM32N6 MCU based on YOLO model.

The second is an Auto Wake Up demo using the STM32N6 and ST’s VD55G1 image sensor. When an object is detected, the STM32N6 MCU, which is in low-power mode, is triggered, and the STM32N6 MCU determines whether the detected object is a person based on the YOLO model and ST Neural-ART NPU. This demo reverts to low power mode when not in use, demonstrating the low power-based application capabilities of the MCU.
■ What are the STM32N6 MCU specifications?

The STM32N6 MCU is a new high-performance MCU from ST, based on the ARM Cortex-M55 architecture and operates at clock speeds of up to 800MHz.

In addition, ST Neural-ART, an NPU manufactured in-house by ST, provides 600 GOPS of computational performance and operates at up to 1 GHz. Equipped with a MIPI CSI-2 camera interface, it can process high-resolution image sensors, and with a built-in H.264 hardware encoder, it provides the performance to implement VISION AI and video streaming applications that were difficult to implement in existing MCUs.

It is equipped with a hardware graphics accelerator and 4.2MB of internal SRAM, allowing graphics and image-related buffers to be processed in internal memory.

■ What are the main applications?

In the field of Vision AI, it can efficiently process tasks that require high-speed calculations such as object detection, facial recognition, and gesture recognition. In the fields of speech recognition and natural language processing, it is used in applications where real-time response is important, such as keyword detection or voice command processing.

It can be effectively applied to various tasks that require △high-speed processing △improved accuracy △real-time response in the industrial and factory automation fields such as △defect detection using vision △robot automation and control △predictive maintenance.
■ What are the main benefits?

The low power and real-time performance of the STM32N6 MCU, combined with the high-speed parallel computing capability of the NPU, enables efficient processing of complex AI tasks on edge devices.

The STM32N6 supports high-resolution video processing via the MIPI CSI-2 interface and H.264 hardware encoder. The NPU with ST's Neural-ART technology provides up to 600 GOPS of computational performance, capable of handling complex Vision AI tasks.

Additionally, ST's stable software and ecosystem support enables engineers to shorten hardware and software development times.

■ Differentiation from competitors’ products

Based on the low-power design based on the STM32 MCU, it can quickly process edge computing based on the Cortex-M55 or the calculation of artificial neural network models based on the NPU in parallel even on small embedded platforms.

In addition, it provides ST’s convenient ecosystem and ST Edge AI Suite development environment that can accelerate the performance and development of STM32 MCUs. ST’s Edge AI Suite supports the process from model learning, quantization, to deployment, making Edge AI application development easier and faster.