SK Telecom to Build Gigawatt-Scale AI Data Centers, Launch GPU Service

Seoul, South Korea—SK Telecom (NYSE: SKM) announced plans to construct artificial intelligence data centers (AIDCs) requiring more than 100 megawatts in multiple regions across South Korea, intending to expand to gigawatt-scale capacity.

The company will establish an AIDC testbed in Pangyo, South Korea, in December 2024. 

The testbed will integrate three types of liquid cooling technologies: direct liquid cooling, immersion cooling, and precision liquid cooling. 

The facility will incorporate SK hynix's High Bandwidth Memory semiconductors and GPU virtualization solutions.

SK Telecom will launch its GPU-as-a-Service in December 2024 through a partnership with U.S.-based Lambda, deploying NVIDIA H100 Tensor Core GPUs. 

The company plans to upgrade to NVIDIA H200 Tensor Core GPUs in March 2025.

The Gasan data center will be converted into an AIDC to address domestic GPU shortages. 

The facility will provide cloud-based GPU services, enabling companies to develop AI services without direct GPU purchases.

SK Telecom is conducting proof-of-concept projects in six sectors: healthcare, AI robots, AI CCTV, and three other undisclosed areas. 

These projects will utilize Edge AI technology, which combines mobile communication networks with AI computing.

The company plans to power its data centers through renewable energy sources, including hydrogen, solar, and wind power. 

The facilities will connect to global markets through submarine cables.

"The upcoming 6G will evolve into a next-generation AI infrastructure where communication and AI are integrated," SK Telecom CEO Ryu Young-sang said at the SK AI Summit, which ran from November 4 to 5.

The initiative includes the development of customized servers and AIDC-utilizing communication infrastructure through global partnerships.

SK Telecom aims to lead global standardization efforts for Edge AI technology while pursuing entrance into international markets.