SK Telecom announced on the 5th that it has launched the GPU-as-a-Service (GPUaaS) 'Haein' (海印·Haein), based on the latest NVIDIA graphic processing unit (GPU) 'Blackwell B200.' GPUaaS refers to a service that provides a cloud infrastructure of GPUs designed for large-scale model training and inference.
The SK Telecom GPUaaS service, which started on the 1st, offers a single cluster composed of over 1,000 B200 units. SK Telecom established the latest B200 GPU cluster following the H100 within its AI data center (hereinafter referred to as AI DC). This is intended to help alleviate the domestic GPU shortage.
The company noted that this is the 'largest and highest-performance GPUaaS in the country' and said it is a more advanced form than the H100-based GPUaaS launched last December, emphasizing that it will significantly contribute to the expansion of infrastructure for the domestic AI industry in the future. It added that this was achieved through SK Telecom's competitive AI DC and close cooperation with global AI leaders, calling it a representative example of the strength and collaboration in its AI pyramid strategy.
The name 'Haein' (海印·Haein) given to this cluster by SK Telecom was inspired by 'Haeinsa,' where the Tripitaka Koreana is stored. This signifies that the SK Telecom B200 cluster aims to become an AI infrastructure that encompasses the digital Tripitaka Koreana, representing K-sovereignty.
The GPUaaS of this Haein cluster has been selected for the 'AI-computing resource utilization enhancement (GPU leasing support) project' of the Ministry of Science and ICT. It is expected to be actively used in developing a national AI foundation model with international competitiveness.
The GPU leasing support project was established to secure international competitiveness for the domestic AI industry and to advance the national AI ecosystem. SK Telecom plans to develop the augmented AI DC into a critical foundational infrastructure for the advancement of the domestic AI industry through this project. The company noted that the launch of the B200-based GPUaaS is another significant achievement following the opening of the Korea region of Lambda, a global AI cloud specialist, in December last year.
To launch this GPUaaS, SK Telecom has been pursuing strategic cooperation with Penguin Solutions, a leading provider of AI DC integration solutions in the U.S. Penguin Solutions is recognized as a top 10 certified partner of NVIDIA and has focused on designing, delivering, and managing high-performance computing (HPC) solutions for over 25 years, establishing a strong position in the global market.
In July last year, SK Telecom signed an investment plan worth $200 million with Penguin Solutions, its largest AI investment to date. The company has prepared a differentiated GPUaaS service while discussing specific mutual cooperation measures by forming a synergy task force. Additionally, through cooperation with its strategic partner, the global AI server manufacturer Supermicro, it was able to rapidly supply and deliver GPU servers in a short period.
SK Telecom's unique technological capabilities are utilized in various ways. By leveraging its own virtualization solution, 'Petasus AI Cloud,' the company maximized operating rates by immediately partitioning and reconfiguring the GPU cluster according to customer demand. It also provides an AIOps solution, 'AI Cloud Manager,' which has accumulated know-how in large-scale model development, offering users enhanced work efficiency and development convenience.
Kim Myung-guk, head of SK Telecom's GPUaaS Business Division, stated, 'The SK Telecom Haein cluster GPUaaS, prepared on the largest scale using the latest GPU B200, will be a catalyst for enhancing the AI competitiveness of customers and the nation.' He added, 'As an AI infrastructure provider, we will exert efforts to build a national AI highway.'