In line with the initiative of the National Science and Technology Council (NSTC) and the Ministry of Economic Affairs (MOEA) to build Taiwan's Intelligent Systems and Manufacturing Integration Platform, Nuvoton follows ITRI's three key pillars for AI development —data, computing power and algorithms— along with a six-dimensional readiness framework: AI strategy, organizational culture, talent and skills, infrastructure, data governance and risk management.
Based on this framework, Nuvoton is modularizing its toolchains, AI models, and development boards, and collaborating with ITRI’s Chip and Systems Integration Services Program to establish a TinyML microcomputing platform. This platform enables small and medium-sized enterprises (SMEs) to complete proofs of concept (PoCs) with minimal barriers to entry, advance to pilot production, and scale through replication. Simultaneously, it fosters “dual-brain collaboration” between AI experts and domain specialists, increasing project success rates and supporting the government’s vision of transforming Taiwan into an “AI Island.”.
As one of the few entry-level AI solutions on the market, the M55M1 integrates an Arm® Cortex®-M55 core (up to 220 MHz) with an Arm Ethos™-U55 micro-NPU onto a single chip, delivering approximately 110 GOP/s of acceleration for CNN/DNN inference. The chip includes up to 1.5 MB of SRAM and 2 MB of Flash, and can be expanded via HyperBus to support HyperRAM/HyperFlash, enabling real-time, offline, low-power AI inference directly at the edge.
Together with Nuvoton's NuML Toolkit and ready-made models such as facial recognition, object detection, speech/command recognition, and anomaly detection, developers can quickly get started using standard MCU development flows, significantly lowering the barrier to AI adoption.
Nuvoton and ITRI will initially focus on three key scenarios:
Edge inspection on production lines: using CCAP for image preprocessing and U55 for inference, enabling defect detection and predictive analysis of equipment condition.
People flow detection and energy control in smart buildings: PIR, ToF, or low-resolution image sensors combined with zone and time-based control to optimize lighting and HVAC.
Edge medical alerts: posture and fall detection directly on end devices, sending only relevant events to protect privacy.
Nuvoton and ITRI will continue to leverage Taiwan's local supply chain and its strength in hardware-software integration to bring AI to real-world environments. The M55M1 enables SMEs to adopt AI in an affordable, scalable, and well-governed way.
