• SK Telecom is building its future around AI
• It has already revamped its operations to position itself as a provider of AI as well as telecom services
• Now it has officially launched its GPU-as-a-service (GPUaaS) product, based (of course) on Nvidia hardware
• It is just one of a number of telcos with GPUaaS aspirations

SK Telecom (SKT) has launched its initial GPU-as-a-service (GPUaaS) to enterprise and government customers in South Korea from its AI datacentre in Gasan, on the outskirts of Seoul.

The service “allows businesses to choose the number and duration of GPUs [graphics processing units] depending on the scale or purpose of AI services,” and enables users to “configure customised packages, such as standalone servers, firewalls and dedicated lines,” noted SKT in this announcement. Prices are “flexibly determined depending on the contract period, the number of GPUs, and the form of prepaid billing,” it added.

SKT has been building up to the launch, a key pillar of its AI infrastructure superhighway strategy, for some months. During the second half of 2024, the Gasan datacentre was upgraded through the deployment of a cluster of Nvidia H100 Tensor Core GPUs in partnership with San Francisco-based AI infrastructure developer Lambda, with which SKT forged a partnership in August – see SK Telecom to deploy Nvidia GPU cluster with Lambda.

And a further upgrade is imminent: SKT plans to deploy Nvidia H200 Tensor Core technology in the first quarter of 2025. “By introducing the H200 in South Korea, SK Telecom plans to create an environment where domestic companies can develop AI technology and services more quickly and also to boost the expansion of GPUaaS customers,” the company noted.

Also critical to the launch of the service is SKT’s AI Cloud Manager, which “manages numerous GPU resources as if they were a single computer to maximise GPU performance and support reducing the learning time required for AI development.”

Kim Myong-gook, head of SK Telecom’s GPUaaS business, noted: “The introduction of GPUaaS marks the first instance of our AI datacentre business actively reaching out to customers. This is a crucial step in positioning SK Telecom as a major force in the AI infrastructure arena. Our goal is to boost our clients’ competitiveness by leveraging our GPU as a platform for their AI services,” added Kim.

According to local media reports, SKT says it has fielded enquiries from more than 100 potential customers, including large enterprises, universities and research institutions. 

SKT, which recently unveiled a new corporate structure that will enable it to better capitalise on growth opportunities related to AI and telecom services, isn’t alone in developing GPUaaS offerings, of course, as all of the major hyperscale datacentre operators have such services as part of their portfolios. 

Despite such mighty competition, a number of telcos are prepping and launching GPUaaS offerings, including: Telekom Malaysia; Indonesia’s Indosat Ooredoo Hutchison (IOH); Singapore’s Singtel and Malaysia’s Maxis in Asia Pacific;  Orange Business and Swisscom in Europe; and Ooredoo in the Middle East.

- Ray Le Maistre, Editorial Director, TelecomTV