Solve the future +852 6264 5732 info@3unex.com
Nov 19, 2026 ·Market Insights

Arm China Unveils Cortex-M42 and Ethos-U55 AIoT Platform

Arm China's new AIoT prototype platform pairs Cortex-M42 with Ethos-U55, combining vector processing and neural acceleration for low-power edge AI devices.


Arm China introduced its Xingchen 100e AIoT prototype platform on November 18, pairing the Arm Cortex-M42 CPU with an Ethos-U55 neural processing unit. The FPGA-based platform demonstrates how CPU vector processing and dedicated neural acceleration can work together in low-power embedded systems.

The platform has run keyword detection, object recognition and automatic speech recognition workloads. It provides a reference for integrating AI functions into endpoint devices with constrained power and silicon budgets.

A Compact CPU for Intelligent Endpoints

Cortex-M42 uses the Armv8.1-M architecture and combines Helium vector processing with Arm Custom Instructions. Helium accelerates machine learning and digital signal processing, while custom instructions allow partners to tailor operations to particular workloads.

Its compact AHB-based design provides an upgrade route for existing Cortex-M4 and Cortex-M33 microcontroller designs. Target applications include wearable audio processing, sensor analysis, connected controllers and industrial condition monitoring.

Arm states that Cortex-M42 offers up to three times the vector compute capability and twice the vector power efficiency of Cortex-M33. The processor is included in the Arm Flexible Access and Arm Total Access portfolios.

Combining CPU and NPU Acceleration

Xingchen 100e adds Ethos-U55 acceleration to handle neural network workloads alongside the CPU. Arm China reports that its demonstrations achieved substantially higher inference performance than Cortex-M42 alone, with gains varying across the tested applications.

The platform follows the earlier Xingchen 300 design, which combines Cortex-M52 with Ethos-U55. Arm China plans a higher-performance successor for 2027.

Together, the processor and prototype platform extend the company's embedded AI options: Cortex-M42 adds vector processing to a compact MCU architecture, while the CPU–NPU combination supports more demanding inference workloads.

Have a sourcing requirement?

Our team replies to every inquiry within one business day.