ST’s STM32 Edge AI Microcontroller Sets New Benchmark for Low Power Computer Vision

STMicroelectronics has raised the bar for embedded AI. At Embedded World 2025, the company introduced its most capable STM32 microcontroller to date—featuring an integrated neural processing unit (NPU) that enables real-time inference for vision and machine learning, all while staying under 300 mW.

Why This Matters for Design Engineers

Edge AI is no longer a luxury—it’s rapidly becoming a requirement across industrial, consumer, and medical devices. What makes this STM32 device stand out is the combination of compute, efficiency, and accessibility.

You get a powerful NPU optimised for real-time tasks like gesture recognition and object detection, backed by proven STM32 low-power architecture. It’s already been tested with models from Ultralytics (YOLO), showcasing strong performance for computer vision at the edge.

Use Case: Smart Glasses, Real-Time AI

To demonstrate its potential, ST showed a wearables application running live on a production-ready board. The system processed visual data locally, powering features like language transcription and smart assistant functions—all on glasses that look and feel like a standard pair.

Despite the AI workload, the system runs for up to 10 hours on a 200 mAh battery. That puts it in a league far beyond most AI-capable wearable solutions on the market today.

Development Tools and Evaluation Kit

Getting started is simple. ST offers a full dev kit with:

• STM32 MCU with integrated NPU
• 5MP onboard camera
• Display and full connectivity stack
• Support for STM32Cube and the wider STM development ecosystem

The highlight? A cloud-based board farm that allows engineers to upload AI models and evaluate performance remotely—before they even touch the hardware. Additionally, ST provides a tool to estimate AI power consumption per application, helping you optimise for battery life early in the design process.

Final Thoughts

Whether you’re developing wearables, edge vision systems, or AI-enabled industrial devices, the STM32 edge AI microcontroller delivers the performance and ecosystem support engineers need. With cloud-based testing, ultra-low power usage, and strong NPU integration, it’s one of the most accessible platforms to bring embedded intelligence to real-world products.