"Supporting smaller and faster edge AI development by combining low-power IoT and on-device AI"

Presenting edge AI-enhanced platform through new products such as nRF54L and nRF54LM20B

AI functions such as real-time sensor data analysis and anomaly detection implemented at MCU level

[Editor's Note] Nordic Semiconductor is presenting a platform that enhances edge AI capabilities through new products such as nRF54L and nRF54LM20B, centered on IoT SoCs based on low-power wireless technologies including BLE, Thread, Zigbee, and Matter. By combining Neuton AI, Axon NPU, nRF Connect SDK AI Add-on, and nRF Cloud and Edge AI Lab, the company has enabled AI functions such as real-time sensor data analysis, anomaly detection, and healthcare and industrial monitoring to be implemented at the MCU level. Through this integrated AI and IoT ecosystem, Nordic is supporting next-generation smart device development across various industries. Against this backdrop, Nordic Semiconductor will present at 'e4ds Tech Day 2026' on the topic of 'On-device AI completed with Axon, Neuton, and MCP,' planning to introduce edge AI development expertise and Nordic's AI development support solutions to domestic edge AI developers. Accordingly, this publication met with Jin Hyuk Yoo, Director at Nordic Semiconductor, who will be presenting at 'e4ds Tech Day 2026,' to discuss development considerations in edge AI and Nordic's solutions.

■ Please introduce Nordic Semiconductor.

Nordic Semiconductor is a Norwegian semiconductor company specializing in ultra-low-power wireless connectivity technology.

We supply system-on-chip (SoC) and modules, as well as development software that support various wireless protocols including Bluetooth Low Energy (BLE), Thread, Zigbee, Matter, Wi-Fi, cellular IoT (LTE-M/NB-IoT), and DECT NR+. Recently, we have also begun developing and manufacturing all components necessary for ultra-low-power wireless communication, including RF Front-end ICs and PMICs, in-house.

Our product portfolio spanning the nRF52, nRF53, nRF54, nRF70, and nRF91/92/93 series is being applied across a wide range of fields including wearables, smart home, healthcare, asset tracking, industrial sensors, and electronic shelf labels (ESL). With cumulative shipments of tens of billions of units, we have established ourselves as a trusted wireless solution in the industry.

We pursue a hardware strategy that maximizes efficiency in the balance between low power consumption and performance through our nPM series PMICs and nRF21 series RF Front-end ICs. Recently, through nRF Cloud and nRF Edge AI Lab, we are expanding our strategy from a hardware-focused company to an integrated platform company encompassing hardware, software, and cloud.

■ Nordic is known for having strengths in ultra-low-power wireless connectivity. As edge AI is added to the existing IoT market centered on BLE, Thread, Zigbee, and Matter, I'm curious how customer requirements are changing.

Until now, customers' top priority has been "sending data with the least power and most reliably to the cloud."

If we were only thinking of products focused on connectivity, now with edge AI added, the focus is shifting from simply sending data well to having the device itself make judgments before sending.

For example, instead of continuously transmitting raw sensor data wirelessly, there is increasing demand to filter out only anomalies or specific events within the device and transmit only meaningful results.

This approach reduces wireless transmission and reception frequency, extending battery life, reducing network traffic and cloud costs, and improving response speed.

Since sensitive data does not need to be transmitted outside the device, privacy requirements are naturally addressed.

As a result, customers now view wireless connectivity and edge AI not as separate challenges, but as problems to be solved together within a single low-power design objective.

■ Nordic is presenting edge AI not as a simple software function, but as a development flow that includes hardware, SDK, and cloud tools. If you were to explain Nordic's edge AI strategy in one sentence—

Our direction is to connect our proprietary ultra-small AI model (Neuton) and on-chip AI accelerator (Axon NPU) with nRF Connect SDK, Edge AI Lab, and nRF Cloud as one integrated system, so that anyone can apply on-device AI to ultra-low-power devices without requiring specialized data science expertise.

The key is that the entire process from model development through Edge AI Lab, firmware integration through nRF Connect SDK and AI Add-on, to firmware deployment and updates through nRF Cloud is seamlessly connected within the Nordic ecosystem.

What characterizes Nordic's strategy is that our hardware—including nRF54LM20B equipped with Axon AI operating on an NPU basis and all nRF Series capable of running Neuton AI—is organically integrated with software such as Edge AI Lab, nRF Cloud, and nRF Connect SDK.

■ What does Nordic consider the most representative edge AI use case?

We can point to anomaly detection and condition monitoring using time-series sensor data as the most representative use case.

This involves real-time analysis within the device of data from accelerometers, IMUs, and PPG (photoplethysmography) sensors, with applications in fall detection, equipment anomaly vibration detection, and biometric signal-based health monitoring.

These applications can achieve high accuracy with relatively light computation, making them well-suited to ultra-small Neuton models running on the CPU.

Beyond this, audio and image-based use cases are rapidly emerging.

Keyword-based voice command recognition, sound classification, and person detection and image recognition tasks require more computation, necessitating hardware acceleration like that provided by nRF54LM20B equipped with Axon NPU.

Examples include always-on voice commands in wearable devices, person and object detection in smart home devices, and abnormal sound detection in industrial environments.

■ I'm curious what Nordic product line is likely to be introduced in the e4ds Tech Day 2026 presentation. In particular, what significance does the nRF54L series have in edge AI?

For an edge AI-related session, we expect the nRF54L series, and specifically nRF54LM20B as the top-tier lineup of this series, to be the focus.

Additionally, the Nordic Edge AI Lab model generation tool and Edge AI Add-on for nRF Connect SDK will also be covered. Most importantly, we plan to primarily explain tips that can help developers using Nordic products.

The nRF54L series is the successor to the existing nRF52 series, featuring a 128MHz Arm Cortex-M33 processor that delivers approximately 2x the processing performance and 3x the processing efficiency compared to the nRF52840.

Thanks to this improved CPU performance, the greatest significance is that ultra-small Neuton models can be run on the MCU without burden, even without a separate AI accelerator.

When equipped with our proprietary ultra-small Neuton model, developers can incorporate on-device AI as a 'standard feature rather than an optional feature' while maintaining low-cost, low-power design.

Taking this one step further, the nRF54LM20B, the highest-performance model in the series, integrates our proprietary AI accelerator Axon NPU, enabling processing of TensorFlow Lite-based models up to 15 times faster than standalone MCU operation, and with 7-8 times higher performance and efficiency compared to competitive wireless NPUs.

In that the device can handle computationally intensive tasks such as audio and image recognition on-chip, nRF54LM20B represents the technical pinnacle of Nordic's edge AI strategy.

■ Finally, please share the core message Nordic aims to convey at e4ds Tech Day 2026, along with advice for edge AI developers.

Edge AI is no longer an expensive option, but rather a fundamental element that can naturally be integrated into ultra-low-power wireless IoT devices.

In this regard, Nordic is unveiling on-device AI demonstrations combining the nRF54L series, Neuton models, and Edge AI Lab, and providing various examples.

At this e4ds Tech Day, we will also focus on conveying concrete implementation methods that Korean developers can actually apply to product design, in line with this direction.

Edge AI is now a solution that is neither expensive nor difficult, and can operate with low power even on inexpensive MCU-based ICs, so we will help you add new ideas to the IoT products you envision and create new products.

Registration for 'e4ds Tech Day 2026,' where Director Yoo Jin Hyuk will present on the topic 'On-device AI completed with Axon, Neuton, and MCP,' can be done through the banner below.