Arm has announced the Cortex M42 processor, bringing advanced vector computing and on device machine learning to low power microcontrollers. The core pairs the Armv8.1 M architecture with Helium signal processing to deliver up to 3x the vector performance of the older Cortex M33. Silicon designers can access the new design immediately through Arm Flexible Access and Total Access licensing programs.
Arm designed the Cortex M42 for cost sensitive and space constrained hardware. The core uses the Armv8.1 M instruction set alongside Helium vector extensions and a compact AHB bus interface. This configuration creates a direct upgrade path for manufacturers currently running older Cortex M4 or Cortex M33 designs. Helium technology allows a single processor core to execute general system control code while accelerating complex math through SIMD execution across single data channels. Compared to the Cortex M33, the Cortex M42 delivers up to 3x higher vector compute capability and up to 2x higher vector energy efficiency.
The processor runs machine learning models locally on endpoint hardware without routing workloads to external cloud servers. Primary consumer deployments include voice processing and keyword spotting in wearable electronics. In industrial settings, the core monitors vibration sensors and tracks environmental data on factory hardware. Arm specifically built the silicon to operate close to robot motors and mechanical joints, analyzing electrical signals in real time to detect mechanical wear before equipment fails.
Arm added the Cortex M42 directly to its commercial licensing programs to speed up tape out timelines for device makers. The Flexible Access catalog currently includes 319 active commercial member companies with more than 500 completed tape out projects. The Total Access tier counts 60 corporate partners supporting over 600 custom silicon tape outs. This licensing framework gives developers access to the new microarchitecture for mass production across industrial, automotive, and wearable hardware segments.
