KT is joining forces with AI chip company DeepX and transportation and commercial vehicle device company Sesol to target the on-device AI market based on neural processing units (NPUs).

KT said on Thursday it signed a three-way business agreement with DeepX and Sesol to cooperate on an "NPU-based on-device AI Internet of Things (AIoT) business".

The three companies will jointly develop a next-generation "AI edge box" equipped with a low-power, low-heat NPU. They plan to apply it to industrial sites requiring real-time video analysis, including electric vehicle (EV) charging stations, mobile closed-circuit (CC) TV and buses and commercial vehicles.

An AI edge box is a small computing device that carries an AI chip to analyse and process data in real time on site. By processing data locally instead of sending it to a central server or cloud, it can reduce communication delays.

Under the agreement, KT will be responsible for commercialising the AI edge box service and expanding the market, based on its network, control platform and business development capabilities. DeepX will provide its self-developed low-power NPU chip, DX-M1, and a software development kit (SDK). Sesol will handle development and manufacturing of the AI edge box applying the NPU, as well as field demonstrations.

The AI edge box being developed is designed to enable real-time AI video analysis while lowering power consumption and heat generation compared with graphics processing unit (GPU)-based equipment. The goal is to expand the scope of application to sites where it has been difficult to introduce existing equipment due to constraints on power supply or installation space.

KT plans to expand the on-device AIoT business centred on public institutions, local governments and the transportation and industrial safety sectors. It will also identify new AI edge services combining AI chips and communications networks.

Sung Je-hyun (성제현), executive managing director and head of the Capital Region Gangbuk Corporate Customer Headquarters at KT's enterprise unit, said, "We will target the next-generation AI video analysis market by combining AI chips and AI edge technology with KT's network and AI service capabilities." He added, "We will provide stable and efficient AI services across various industrial sites."