Created for future software-defined vehicles
STMicroelectronics has developed the Stellar P3E, the first automotive microcontroller (MCU) with built-in AI acceleration for automotive edge intelligence
Created for future software-defined vehicles, the Stellar P3E can simplify multi-function integration for X-in-1 Electronic Control Units (ECUs), reducing system cost, weight, and complexity.
Luca Rodeschini, group vice president and general purpose and automotive microcontrollers division general manager, STMicroelectronics, said, “Stellar P3E sets a new benchmark for automotive electrification by combining high-performance real-time control and edge AI in a single device that meets the highest automotive safety levels. Its increased processing power, AI acceleration, large and extensible memory, rich analogue content, smart sensing capabilities, and intelligent power management functions support new applications such as virtual sensors. This better enables automakers with the tools to create safer, more efficient and more responsive driving experiences.”
The Stellar P3E features an integrated ST Neural-ART Accelerator, enabling real-time AI efficiency, making it the first MCU with an embedded neural network accelerator for the automotive industry. Powered by this dedicated neural processing unit (NPU) with an advanced data-flow architecture for AI workloads, combined with its sensing capabilities, the P3E is capable of smart sensing, facilitating new applications such as virtual sensors.
The P3E provides inference processing at microsecond speeds, reaching up to 30x the efficiency of traditional MCU core processors. This enables always-on, low-power AI that can support real-time functions, including predictive maintenance and smart sensing.
These capabilities could help to enhance charging speed and efficiency in EVs and enable rapid deployment of new features, whether in the factory or the field. OEMs can introduce new functions and more intuitive behaviours through different AI models, reducing the need for additional sensors, modules, wiring, and integration effort.
Greg Basich, associate director, Counterpoint Research, said, “Shifting neural processing from centralised hubs to the edge of the vehicle enables sub-millisecond-level decision-making, which is essential for the next generation of in-vehicle intelligence. Integrating AI hardware acceleration at the MCU level allows OEMs to deliver advanced capabilities, such as predictive maintenance for vehicle performance and smart sensing with virtual sensor applications. This enables very low latency sensing, actuation control, and other sophisticated features, without the cost and thermal burden of a full-scale SoC.”
The automotive industry is shifting toward software-defined vehicles (SDVs), and Stellar P3E’s integrated xMemory, ST’s proprietary non-volatile memory based on phase-change memory (PCM), has been designed to provide the essential scalability and flexibility for this shift. Offering twice the density of traditional embedded flash memory and qualified for automotive environments, this extension memory solution allows dynamic expansion of software storage to accommodate new features and updates without requiring hardware redesign.
The P3E is supported within the ST Edge AI Suite, a comprehensive edge-AI ecosystem spanning from dataset creation to on-device deployment for data scientists and embedded engineers. As part of this device, the NanoEdge AI Studio tool is available for the entire Stellar MCU family. The Stellar P3E is also already integrated into Stellar Studio, ST’s all-in-one development environment for automotive engineers. Together, they help to strengthen a hardware and software ecosystem designed to streamline the deployment of edge AI solutions for demanding automotive environments.
Production of the Stellar P3E is planned to start in the fourth quarter of 2026.
