KT puts Korean AI chips, language model into one enterprise server

KT has introduced a new enterprise server that integrates Korean-made AI technology into a single unit. The KT NPU LLM Station combines an artificial intelligence processor, language model, and operating platform into one appliance. This new offering allows companies to run generative AI exclusively within their own networks.

The system is built around Rebellions’ ATOM-MAX neural processing unit, a Korean-designed chip specifically for inference tasks. KT’s ATOM-MAX chip has shown higher performance and better power efficiency compared to traditional graphics processing units in independent testing. The server also incorporates KT’s proprietary Mi:dm K 2.5 Pro large language model, which specializes in complex reasoning and agent-based applications.

The KT NPU LLM Station comes with an application programming interface platform, enabling businesses to manage the hardware, model, and related services as a single, cohesive package. Unlike conventional data centers, the appliance is installed within a customer’s facility, allowing data and AI computations to be processed internally.

This design caters to industries with stringent security requirements and network-separation rules, ensuring that sensitive information remains within the confines of the company’s own network.