Samsung SDS has set a target of generating annual revenue in the 6 trillion won range from its data center business by expanding operations through 2031. As surging demand for artificial intelligence (AI) inference heightens the need for data centers, the company's strategy is to expand infrastructure to secure stable earnings.
According to industry sources on the 19th, Samsung SDS unveiled this data center business plan at an analyst day earlier this month. Samsung SDS currently operates data centers in Sangam, Suwon, and Chuncheon. Related revenue alone stands in the 1 trillion won range. The company aims to build additional data centers over the next five years to increase revenue about sixfold.
The plan is broadly divided into a first-phase project running from this year through 2029 and a second-phase project running through 2031. Both center on the Korea AI Computing Center in Haenam, scheduled to open in 2028, and the Gumi AI Data Center, targeted to open in 2029.
In the first phase, Samsung SDS is pursuing a plan to secure 40 megawatts (MW) at the National AI Computing Center and 60 MW at the Gumi AI Data Center. The second phase similarly centers on expanding capacity by 40 MW and 60 MW at the two data centers, respectively.
Samsung SDS is also actively expanding its DBO business, which handles the design, construction, and operation of data centers on an end-to-end basis. The company plans to secure data centers of more than 500 MW through the DBO business alone by 2031. If the plan is achieved, the DBO business alone could generate 1 trillion won in annual revenue. Samsung SDS previously selected the DBO business as a new venture last year, and this year said it is conducting a proof of concept (PoC) with a major domestic asset management firm.
The industry views this Samsung SDS strategy positively, given the characteristics of AI. According to consulting firm McKinsey, global AI training and inference workloads are expected to grow at an average annual rate of around 30% through 2030. In particular, AI inference workloads alone are projected to more than quadruple from 20.9 gigawatts (GW) last year to 93.3 GW by 2030. As AI inference workloads increase, the need for AI data centers equipped with graphics processing units (GPUs), power, and cooling devices inevitably grows as well. In particular, operating a data center with GPU as a service (GPUaaS) is expected to secure stable revenue through subscriptions from corporate clients.
This is also why other companies are stepping up to operate even modular data centers to preempt infrastructure demand driven by AI inference. Both LG CNS and Elice Group are pursuing plans to build modular data centers in Busan by the end of this year. "Companies with the technology and experience of operating data centers will materialize revenue growth as it coincides with the period of rising AI demand," an industry official said. "For Samsung SDS, the effects are expected to become apparent after 2028, when its data centers go into full operation."
