KT: "Reduce Costs with Monthly GPU Subscriptions"…Partners with Chungbuk National University

Providing a High-Performance GPU-Based Computing Environment
Efficiently Enhancing the Speed and Competitiveness of AI Development

On December 4, KT announced that it will provide its high-performance Graphics Processing Unit (GPU) subscription service, 'K GPUaaS,' to the Industry-Academia Cooperation Foundation of Chungbuk National University’s Small but Strong Specialized Zone Support Center (Chungbuk Cheongju Small but Strong Specialized Zone), aiming to expand its market presence.

K GPUaaS is a service launched in September that allows users to access Nvidia's H100 GPUs, secured by KT, on a monthly subscription basis. By reducing the initial infrastructure setup costs and maintenance burdens, it enables more efficient improvement in the speed and competitiveness of artificial intelligence (AI) development.

The Chungbuk Cheongju Small but Strong Specialized Zone is a core technology research institution operated by Chungbuk National University. KT is providing a high-performance GPU-based computing environment optimized for AI development and training at this location through the application of K GPUaaS. In addition to professional consulting for GPU utilization, KT also supports resources needed for the entire process of AI development and commercialization, including open innovation programs.

Starting with this case, KT plans to expand the application of K GPUaaS to various companies and institutions across Korea. The company expects to improve GPU accessibility for companies and institutions that require high-performance computing for AI deep learning model training and inference, as well as data analysis.

K GPUaaS utilizes 'InfiniBand,' an ultra-high-speed networking technology, to provide seamless communication between GPU servers, enabling the implementation of large-scale distributed learning environments. Furthermore, with GPU virtualization and partitioning technology, a single GPU can be divided into multiple units, allowing resource allocation and adjustment according to workload, thereby increasing GPU utilization efficiency. All GPU infrastructure, data, and network are managed domestically, minimizing concerns about customer data being leaked overseas.

Yoo Seongbong, Head of the AX Business Division at KT Enterprise (Executive Vice President), stated, "We will continue to support more customers so they can use high-performance GPUs and distributed learning environments without cost concerns, thereby contributing to the revitalization of Korea's AI development ecosystem."