Nordic Semiconductor has unveiled a next-generation edge AI solution that enables high-performance on-device AI in ultra-low-power IoT devices, signaling a paradigm shift in the large-scale IoT market.

Nordic showcased a live demo of its new ultra-low-power edge AI solution at CES 2026 Venetian Expo booth (#52039).

Nordic says it has “dramatically simplified the implementation of edge AI and maximised power efficiency” with its new nRF54L Series SoCs, ultra-compact Neuton AI models, and development platform, the Nordic Edge AI Lab.

The centerpiece of this announcement is the nRF54LM20B SoC, which integrates the Axon NPU.

Based on technology from Atlazo, which Nordic acquired in 2023, this NPU delivers up to seven times faster performance and eight times higher energy efficiency than existing solutions for key edge AI tasks such as sound classification, keyword spotting, and image detection.

This means that real-time AI processing will become possible even in ultra-small battery-based IoT devices.

The nRF54LM20B has 2MB NVM, 512KB RAM, 128MHz Arm Cortex-M33, and RISC-V coprocessor. It has a large capacity and high performance configuration, and also supports the latest wireless technologies such as Bluetooth LE, channel sounding, and Matter over Thread.

Nordic describes it as “a next-generation platform that can implement edge AI as a basic function in large-scale IoT devices.”

Edge AI models have also been significantly lightweight.

The Neuton model provided by Nordic is an ultra-small AI model of less than 5KB, operates with only a CPU, and is up to 10 times smaller and faster than existing CPU-based models.

It can be automatically generated for various purposes such as anomaly detection, activity recognition, and biosignal analysis, significantly reducing the development burden.

Because all processing is done locally, it also has strengths in terms of privacy protection.

Nordic Edge AI Lab, a development platform, supports the easy creation and deployment of these models.

In fact, one global supply chain company has leveraged this platform to implement a smart tracking feature that detects events during the logistics process, such as shock and shaking, within the device itself.

The generated AI model was immediately deployed over-the-air (OTA) across all devices via Nordic’s nRF Cloud Lifecycle Service.

“Edge AI is no longer an option, it’s a necessity,” said Nordic Semiconductor CEO Vegard Uhlan. “Local AI is the only solution for billions of IoT devices to make decisions in milliseconds, maximize battery life, and protect personal data.”

Öyvind Strom, Senior Vice President, also said, “With the Axon NPU, Neuton models, and Edge AI Lab, advanced on-device AI is now within the reach of every embedded developer.”“It provides scalable performance from wearables to industrial sensing,” he said.

Nordic is sampling the nRF54LM20B SoC to select customers, with full production scheduled to begin in Q2 2026. Edge AI Lab and Neuton models are already available for the nRF54 Series and cellular IoT modules.