Nordic Semiconductor has announced that it is expanding its nRF54L series of chips, adding key new capabilities that are designed to enable "ultra-low-power edge artificial intelligence.
Specifically, the announcement, which comes as Nordic is attending the 'Embedded World' industry event in Nurembourg, details that the nRF54L series's new nRF54LM20B System-on-Chip introduces an integrated Neural Processing Unit (NPU) and associated developer tools.
Nordic said the nRF54LM20B’s large-memory NPU accelerates TensorFlow Lite–class models up to 15 times faster than execution on the Arm Cortex CPU, and delivers up to 7 times higher performance and 8 times better energy efficiency compared with the closest competing edge AI solution, making high-rate sensor, audio and event-driven workloads feasible on battery power.
"This new generation of edge AI-enabled capabilities fundamentally transforms what small, battery-powered devices can perceive and interpret in the real world," says Øyvind Strøm, Nordic's EVP for its Short-Range division.
Related reading
- Meta commits to up to 6GW of AMD chips in a multi-year deal that sent AMD shares up 9%
- NVIDIA recruits five cybersecurity firms to protect critical infrastructure using AI run on its own chips
- Meta and NVIDIA agree multiyear partnership to build hyperscale AI infrastructure
Nordic noted that the latest product also adds model-generation and optimization features to simplify on-device AI development.
Developers can upload custom datasets and specify architectures to produce classification or regression models. The company said the online Nordic Edge AI Lab will produce models for either accelerated inference on the NPU or efficient execution on the CPU, and highlighted wake-word creation from a single text input to avoid costly voice dataset collection.
The recap
Nordic expands nRF54L series with NPU-enabled nRF54LM20B SoC
NPU runs TensorFlow Lite–class models up to 15 times faster
Volume production of nRF54LM20B expected to begin in Q2 2026