SK Telecom (CEO Yoo Young-sang) is operating a cluster equipped with over 1,000 NVIDIA B200 GPUs, making a significant contribution to enhancing domestic AI competitiveness.

SKT announced on the 6th that it has launched the largest single cluster in Korea, 'Haein', based on the latest NVIDIA B200 GPUs, and has launched GPUaaS (GPU-as-a-Service), a high-performance AI infrastructure service.

This cluster is expected to contribute to expanding the infrastructure of the domestic AI industry and strengthening global competitiveness by being equipped with over 1,000 B200 GPUs.

SKT's GPUaaS is a GPU cloud infrastructure service for large-scale AI model learning and inference, and its performance and stability have been significantly improved compared to the H100-based GPUaaS launched last year.

The 'Haein' cluster was named after Haeinsa Temple, where the Tripitaka Koreana is stored, and symbolizes the K-sovereign AI infrastructure of the digital age.

This GPUaaS was selected as a supplier for the Ministry of Science and ICT's 'Strengthening AI-Computing Resource Utilization Base (GPU Rental Support) Project' and is a member of the National AI Foundation.It is used as a core infrastructure for Dell development.

This project is a government-led initiative aimed at enhancing the domestic AI ecosystem and securing global competitiveness. SKT will take the lead in resolving the GPU shortage issue and advancing the AI industry, focusing on the Gasan AI Data Center (AI DC).

SKT has promoted cluster construction and flawless operation through strategic cooperation with Penguin Solutions, a US-based AI DC integrated solution specialist.

Penguin Solutions is a global company with over 25 years of experience designing and operating HPC solutions and is an NVIDIA Certified Top 10 Partner.

SKT signed a $200 million investment agreement with Penguin in July of last year and formed a synergy task force to prepare for the differentiation of its GPUaaS service.

Additionally, through collaboration with global AI server manufacturer Supermicro, we were able to secure GPU servers in a short period of time and deliver them on time.

SKT's proprietary technology, 'Petasus AI Cloud,' maximizes operating rates by flexibly partitioning and reconfiguring GPU clusters according to customer demand, and enhances development efficiency and convenience through the AIOps-based 'AI Cloud Manager.'

The launch of this B200-based GPUaaS is another tangible achievement for SKT, following the opening of a Korean region for global AI cloud company Lambda last year, once again proving its leadership in the domestic AI infrastructure market. SKT plans to continuously expand its GPUaaS business in the future and strengthen infrastructure support to help the domestic AI industry become competitive on the global stage.

Kim Myeong-guk, head of SKT’s GPUaaS business division, said, “The Haein cluster based on B200 will become a key infrastructure that will enhance Korea’s AI competitiveness,” and “SKT will take the lead in building a national AI highway and provide world-class GPU cloud services.”