| Title |
Structural Drivers of Capital-Area Concentration in Korean Data Centers and Policy Directions of the Power System Impact Assessment in the AI Data Center Transition |
| Authors |
Byungyun Bae ; Woo-jong Kim |
| DOI |
https://dx.doi.org/10.6106/KJCEM.2026.27.5.035 |
| Keywords |
Data Center Location; Seoul Capital Area (SCA) Concentration; Power System Impact Assessment (PSIA); AI Data Center; Exploratory Policy Analysis (EPA); Pipeline Gap; Construction Project Management; Dual-Track Decentralization Policy |
| Abstract |
This study analyzes the structural drivers of data center (DC) concentration in the Seoul Capital Area (SCA) and examines the policy implications of the Power System Impact Assessment (PSIA) during the transition toward AI data centers. Because PSIA is at an early implementation stage and post-implementation microdata remain limited, this study adopts an Exploratory Policy Analysis (EPA) approach based on secondary-data triangulation. Government regulations, industry reports, market outlooks, and prior studies are examined through a five-axis analytical framework consisting of infrastructure, market proximity, human resources, technology, and institutions. The results indicate that SCA concentration is not merely a business preference but a cumulative outcome of power and telecommunications infrastructure, customer proximity, specialized workforce availability, low-latency requirements, and administrative predictability. A pipeline gap is also observed as the SCA share increases from the planning stage to the active execution and expected completion stages. Sensitivity analysis confirms that the pipeline gap remains directionally robust, although its magnitude varies depending on the treatment of canceled projects. International cases from Ireland, the United States, Singapore, the Netherlands, and Japan provide indirect signals that grid-related regulations may produce a regulatory paradox if non-SCA regions lack viable implementation conditions. Finally, training and inference AI data centers require differentiated policy tracks because their locational logics differ. The paper proposes policy measures including power hosting capacity maps, one-stop permitting, non-SCA cluster evaluation, differentiated AI data center location strategies, and execution-centric performance indicators. The regulatory paradox is presented as an exploratory policy hypothesis, not as a rationale for weakening PSIA. |