Rebellions 'Atom-Max' Integrated with KT LLM and API
"A Catalyst for AI Transformation in Security-Sensitive Sectors Like Defense, Manufacturing, and Finance"

On August 19, KT announced the launch of an enterprise sovereign AI appliance that integrates a domestically developed artificial intelligence (AI) semiconductor and a large language model (LLM) within a single server.


A KT official is inspecting the 'KT NPU LLM Station.' Provided by KT

A KT official is inspecting the 'KT NPU LLM Station.' Provided by KT

View original image


The 'KT NPU LLM Station' features the 'Atom-Max' neural processing unit (NPU) from Rebellions, KT’s proprietary LLM 'Midum K 2.5 Pro', and an operational API platform. Every component—including the AI semiconductor, model, and platform—uses domestic technology.


Atom-Max is a Korean NPU specialized in inference tasks. Its processing capability and power efficiency have been validated through official testing, outperforming graphic processing units (GPUs) of the same class. 'Midum K 2.5 Pro' is KT’s proprietary model with strong capabilities for complex inference and agent utilization.


KT cited robust security as the product’s key feature. Installed locally within an organization, all data and AI computations are processed solely in-house. KT anticipates this will act as a catalyst for AI transformation (AX) in regulated industries—such as public sector, defense, pharmaceuticals, manufacturing, and finance—which have been unable to adopt foreign generative AI due to network segregation and related regulations.


The product also reduces adoption barriers. As a fully integrated device—including the server, AI model, and platform—it can be used immediately for in-house document-based question answering (RAG) and other AI features from the day of installation, without the need for additional setup. It delivers better power efficiency than comparable GPU servers and provides industry-standard APIs for compatibility with existing AI services.


KT plans to add KT-developed AI agents over time—such as meeting minutes automation, coding assistance, and the business automation agent 'K-Claw' (tentative name). KT will also partner with specialized agent developers to expand into custom AI build-outs for clients. Furthermore, KT will broaden the application domain of the 'KT NPU LLM Station' as an edge data center for physical AI, aiming to implement a low-power, low-latency AI infrastructure.



Jinhyung Lee, Senior Vice President and Head of KT's AX Business Division, commented, "We have enabled sovereign AI to be used immediately at industrial sites," adding, "This will offer a practical choice for companies and institutions looking to embark on AI transformation while safeguarding data sovereignty."


This content was produced with the assistance of AI translation services.

© The Asia Business Daily. All rights reserved. Unauthorized AI training and use prohibited.

Today’s Briefing