KT Launches GPU Cloud Service 'Hyperscale AI Computing'

[Asia Economy Reporter Eunmo Koo] KT announced on the 10th that it has launched a cloud-based Graphics Processing Unit (GPU) infrastructure service called 'Hyperscale Artificial Intelligence (AI) Computing.'

'Hyperscale AI Computing' is a pay-as-you-go service based on actual usage that provides high-cost GPU infrastructure through dynamic allocation. AI service specialists or AI developers can allocate and use GPU resources as much as they want at the desired time and return them afterward.

Hyperscale AI Computing supports smooth scaling when large-scale GPU resources are needed. It clusters GPUs beyond the maximum number that can physically operate on a single server for computational use. Even if all GPUs prepared at a specific time are fully utilized and resources are temporarily insufficient, requests are automatically queued and processed sequentially, ensuring excellent service stability.

It offers model programming compatibility. Hyperscale AI Computing allows the use of a multi-GPU environment without redesigning existing development sources. Additionally, resources needed at each development stage can be continuously expanded or reduced. Development can continue without service interruption, which was previously required to change allocated resources.

Meanwhile, KT and More are planning to build a large-scale GPU farm early next year following this product launch. In 2023, they plan to produce a dedicated AI semiconductor chip to promote the localization of GPU technology.

Yoon Dong-sik, Head of KT Cloud & IDC Business Promotion Office, said, "With the launch of Hyperscale AI Computing, we expect to change the AI service development market and cultural paradigm in South Korea," adding, "We will continue to strengthen cooperation with AI specialized companies to create an ecosystem where South Korean AI developers can actively operate."

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