Telecom Giants Pour Trillions into GPU Clouds and Data Centers in Race to Dominate AI Infrastructure Market

The three major telecom carriers have jumped into the race to capture the AI infrastructure market. To overcome the slowing growth of their existing telecommunications businesses, they are adopting B2B infrastructure services—which combine AI computing resources with operational services—as a new growth engine. Their strategy involves massive investments in 'GPU clouds,' which lend AI computing power to corporate clients, and data centers.

The backdrop for the three carriers' concrete moves into the AI infrastructure business is the slowing growth of their core mobile operations. As of last December, 5G subscribers exceeded 38 million, but ARPU (Average Revenue Per User) has stagnated or declined, making it difficult to boost profitability solely through subscriber growth. The carriers are leveraging their nationwide networks, enterprise lines, and data center operation experience to expand their AI infrastructure businesses for corporate clients.

Each company's strategy to capture the AI infrastructure market is taking a distinct form. SK Telecom, based on its partnership with NVIDIA, has announced plans to build a gigawatt-scale AI cloud and operate an 'AI factory,' and is currently pursuing pilot projects in the manufacturing and defense sectors. KT has allocated 6 trillion won out of its 18 trillion won investment blueprint to AI infrastructure, centering on a 1GW-class AI data center (AIDC) and a 'Token Factory' management system. LG Uplus is concretizing a build-to-suit (BTS) data center business centered around its AIDC in Paju, targeting 5 trillion won in cumulative AIDC orders by 2030.

For the AI infrastructure business to translate into actual operation and revenue, it must be supported by power procurement, GPU acquisition, and anchor contracts (long-term agreements with major clients). Consequently, rather than short-term performance, the key competitive factors are emerging as actual capacity utilization rates and operational capabilities, such as securing data center capacity and industrial clients.