Microchip Launches Full-Stack Edge AI Platforms
Microchip Technology has launched production-ready, full-stack edge AI platforms for its microcontrollers and microprocessors. The solution integrates hardware, deployable machine learning models, and development tools to simplify AI deployment in industrial and embedded environments. Key applications include anomaly detection, predictive maintenance, and smart sensing in resource-constrained scenarios.
- The solution is designed to scale from 8-bit microcontrollers to more advanced devices, utilizing tools like the MPLAB X IDE and MPLAB Harmony framework. For more computationally intensive AI tasks, the platform incorporates PolarFire® FPGAs, which can be programmed using the VectorBlox Accelerator SDK without needing deep FPGA expertise. - Microchip's VectorBlox Accelerator SDK 2.0 offers up to a 25% performance improvement over previous versions and integrates with frameworks like TensorFlow Lite, PyTorch, and Caffe. This software abstracts the underlying hardware, allowing developers to deploy models without extensive FPGA knowledge. - The new offerings include pre-trained models for specific applications such as detecting dangerous electrical arc faults, facial recognition with liveness detection, and keyword spotting for voice commands. These models can be customized and are part of a strategy to simplify the transition from prototype to production. - To facilitate development, Microchip provides hardware like the SAM9X75 Curiosity Development Board, which features a high-performance ARM926EJ-S based MPU running up to 800 MHz with integrated DDR3L memory. Another option is the PIC32CX-BZ2 and WBZ451 Curiosity Development Board for evaluating wireless MCU capabilities with Bluetooth LE and Zigbee. - The company collaborates with AI software partners like Edge Impulse, Cartesiam, and Motion Gestures to integrate their machine learning tools directly into the MPLAB X development environment. These partnerships aim to cover all phases of an AI/ML project, from data gathering to inference implementation. - For high-bandwidth sensor applications, Microchip offers the PolarFire FPGA Ethernet Sensor Bridge, which aggregates and streams data to platforms like NVIDIA Holoscan for real-time robotic vision and multi-sensor fusion. This bridge supports interfaces like MIPI CSI-2 with plans to expand to others such as 12G SDI and CoaXPress 2.0. - The technology is built on hardware like the PIC32CX-BZ2 family of MCUs, which feature an Arm® Cortex®-M4F processor and integrate Bluetooth 5.2 and Zigbee stacks. The SAM9X75 series of MPUs provides connectivity options like Gigabit Ethernet with TSN, CAN-FD, and multiple display interfaces. - An October 2025 market report from IoT Analytics, cited by Microchip, identifies embedding AI capabilities directly into MCUs as one of the top four industry trends to reduce latency and enhance data privacy.
