SKT Opens Gasan AI Data Center… Official Launch of ‘GPUaaS’

Launch of ‘SKT GPUaaS’ based on NVIDIA GPU H100… Prepared for over a year with Lambda, “SK Telecom announced on the 13th that it launched its AI cloud service ‘SKT GPUaaS’ after completing the pilot operation of the Gasan AI Data Center.”,

On December 23, at the SK Broadband Gasan IDC, SKT CEO Yoo Young-sang is inspecting the preparations for GPUaaS [Photo=SKT],

“Since announcing its ‘AI Infrastructure Super Highway’ strategy at the ‘SK AI Summit 2024’, SK Telecom has established an ‘AI DC Division’ as a separate organization in its year-end restructuring and launched GPUaaS, accelerating its efforts to advance its AI infrastructure business.”,

“The AI Infrastructure Super Highway strategy involves three main components: AI Data Center GPUaaS and Edge AI, aimed at building AI infrastructure nationwide.”,

“The newly opened Gasan AIDC provides optimal data colocation environments for high-density GPU server operations, boasting a power density per rack of 44kW, roughly nine times the domestic data center average of 4.8kW.”,

“‘SKT GPUaaS’ is a subscription-based AI cloud prepared over the past year with the American company Lambda. To achieve this, Lambda has established a Korean region in the Gasan AIDC, marking its first region in the Asia-Pacific area.”,

“In February, SK Telecom made a strategic investment in Lambda, a global GPU cloud company, as its first global move to propel its AI DC business.”,

“Founded in 2012 by AI engineers, Lambda provides AI cloud services using GPUaaS to global tech giants. Leveraging its collaboration with Lambda, SK Telecom plans to actively penetrate the AI cloud market on the back of stable GPU procurement.”,

“This newly launched GPUaaS is based on the NVIDIA GPU H100, the highest-performing GPU currently introduced in Korea.”,

“Corporate subscribers can conduct stable AI training and inference using dedicated GPUs, and security is significantly enhanced when using firewalls and dedicated networks. Furthermore, even with rapidly increasing AI workloads, additional GPUs can be easily incorporated into standalone servers, alleviating the burden of additional facility investments.”,

“By using GPUaaS in conjunction with SK Telecom’s ‘AI Cloud Manager’, GPU resources can be managed more efficiently. The ‘AI Cloud Manager’ optimizes GPU performance by managing numerous GPU resources as if they are a single computer, thereby reducing the training time required for AI development.”,

“The pricing for SKT GPUaaS is flexibly determined based on contract duration, number of GPUs, and prepayment models to greatly enhance subscriber convenience. For instance, a subscriber wishing to use 32 GPUs for 24 months can have a tailored pricing plan. Short-term services for 1 month or 2 months are also available.”,

“Additionally, SK Telecom plans to introduce the latest GPU, H200, within the first quarter. By being the first in Korea to adopt the H200, they intend to accelerate the development of AI technologies and services by domestic companies, while also expanding GPUaaS subscriptions.”,

“Kim Myung-guk, head of the GPUaaS business at SK Telecom, stated, ‘The launch of GPUaaS is the first instance of AI Data Center business reaching out to subscribers, marking a significant milestone for SK Telecom to establish itself as a key player in the AI infrastructure sector,’ adding that they aim to build a GPU Farm that boosts the nation’s AI competitiveness.”