STMicroelectronics (NYSE: STM), a global semiconductor leader serving customers across the spectrum of electronics applications, today announced the launch of a new series of microcontrollers (MCUs) that are the first to integrate accelerated machine learning capabilities.

ST said that its new MCUs that support embedded AI implementations can provide high-performance functions in small embedded systems while utilizing computer vision, audio processing, sound analysis, and other algorithms in cost- and power-sensitive consumer and industrial products.

ST claims that its STM32N6 microcontroller (MCU) series is its most powerful to date, with the first devices to feature the Neural-ART AcceleratorTM Neural Processing Unit (NPU) that it claims to deliver 600x more machine-learning performance than the latest high-end STM32 MCUs.

The STM32N6 has been available to select key customers since October 2023 and is reportedly available in volume quantities now. Already, various companies such as LG Electronics, Lenovo, Alps Alpine, Carlo Gavazzi, MetaBounds, and Autotrack have adopted the STM32N6 early.

“The STM32N6 delivers outstanding AI performance and outstanding flexibility in a compact silicon package that is ideal for embedded systems and wearable devices,” said Yehan Ahn, CTO of LG Electronics, commenting on the early adoption. “Furthermore, the inference speed driven by the ST Neural-ART accelerator exceeded our expectations, and ST’s easy-to-use software tools enabled developers to more seamlessly integrate AI models into the MCU.”

In addition, Alps Alpine is a global company that develops automotive infotainment and electronic devices, and said that the introduction of the STM32N6 is 'well-suited for developing compact devices that execute AI inference based on multiple sensors and provide diverse and enhanced user experiences.'

MetaBounds said in the development of the consumer AR glass, “The STM32N6 enabled us to deliver outstanding features such as an embedded Neural-ART accelerator, an enhanced camera interface, and a dedicated Image Signal Processor (ISP) in an ultra-light and compact form factor, while improving the user experience without compromising battery life.” The ultra-light, high-performance, and low-power chip was highly evaluated as a key element for wearable products.
With up to 600x better machine learning performance than typical high-end STM32 MCUs, the STM32N6 features Neural-ART accelerators. It contains approximately 300 configurable MAC (Multiply-Accumulate) units, delivering performance of up to 600 GOPS (Giga Operations per Second).

It is equipped with an 800MHz Arm® Cortex®-M55 core, provides 4.2MB RAM, supports ISP integration and ISP IQTune software, supports ST Edge AI Suite, and supports AI model library (Model Zoo) to accelerate development and shorten time to market.

“ST is on the cusp of a major transformation at the small edge,” said Remi El-Ouazzane, President of ST’s MCU, Digital IC and RF Products Group. “Our customers’ workloads will increasingly be complemented or replaced by AI models. Today, these models are used for tasks such as segmentation, classification and recognition, but in the future they will be applied to new applications that have not yet been developed.”

“A common misconception about AI is that it will only apply to power-hungry big data center applications,” said Tom Hackenberg, senior analyst in Yole’s Memory and Compute Group. “However, today’s IoT edge applications require many of the same types of analytics capabilities that AI provides.”

“The STM32N6 is a prime example of a new era that will combine the workloads of energy-efficient MCUs with the analytics capabilities of AI to deliver computer vision and large-scale sensor-based performance, significantly reducing the total cost of ownership of modern equipment,” he explained.