20GW Planned for 2035 Through Mega Project
Verified Demand Stands at Only 2.5GW
"Rigorous Validation Needed... Difficult to Recover Investments if Projects Halted"

The government announced in June that it would invest 550 trillion won through a mega project to build AI data centers totaling 18.4GW. However, only a very small number of locations have actually secured move-in contracts. Industry experts warn that, as procedures such as grid connections and the construction of power generation facilities are being carried out before occupancy contracts have been finalized, rigorous validation is necessary in advance because it would be difficult to recover investments if the project is suspended midway.


According to the report "Between Calm and Passion: On the Uncertainty of AI Data Center Demand" published by Next group on September 2, if the government’s mega project and the domestic AI data center plans announced by major conglomerates are realized, the scale of AI data centers in Korea would reach approximately 14.8GW by 2030 and about 20GW by 2035.


However, projects with confirmed long-term customers or verified demand stand at only 2.3GW for 2030 and 2.5GW for 2035. Most of the announced AI data center plans are premised on attracting overseas hyperscalers and AI companies to Korea, but the majority remain either without confirmed core long-term customers or at the stage of non-binding memorandums of understanding (MOUs).

Where Are the Tenant Companies for the 550 Trillion Won AI Data Center Investments? View original image

The report notes that even under an optimistic scenario, in which Korea establishes a strategic AI partnership with the U.S. and attracts AI computing demand on the scale of the United Arab Emirates, domestic demand for AI data centers would reach only 7.3GW by 2030 and 12.5GW by 2035—still falling short of planned capacity.


The report explains, "Major conglomerates possessing actual demand in Korea are planning to meet their needs by building their own dedicated AI data centers, and the immediate demand from Korea’s public, academic, and research sectors will also be supplied through the National AI Computing Center for the time being, leaving domestic new demand limited. As such, a significant portion of the projects must inevitably rely on attracting overseas hyperscalers, but Korea’s competitiveness as a location for AI data centers has yet to be proven."


Furthermore, there are concerns that the outlook for supply-demand gaps in U.S. data centers—which has been cited as a basis for large-scale expansion in Korea—does not reflect the rapidly changing AI infrastructure investment plans since the initial announcement.


According to Cushman & Wakefield’s 2026 Global Data Center Market Comparison report, Seoul (including the metropolitan area) ranked 10th among the leading markets, while Busan ranked 7th among the lower-tier markets.  JLL’s 2026 Global Data Center Outlook report also rated Seoul lower than Mumbai, India, and Sydney, Australia, in both "average grid connection queue period" and "average construction cost."


The report further states, "U.S. AI companies' overseas investments are concentrated in countries such as the UAE, India, and Canada, where abundant power, land, regional demand, and security conditions are combined. In contrast, the agreements that OpenAI has signed in Korea with the Ministry of Science and ICT, Samsung, and SK Telecom are still in the stage of cooperation and review based on MOUs."


The problem arises when infrastructure investments such as grid enhancement are made preemptively, despite such uncertain demand. Kim Sugang, the senior researcher who authored the report, pointed out, "If capacities for which leasing contracts or financing arrangements are not yet secured take precedence in securing grid connections, land, or power generation facilities and the project is then suspended, it will be difficult to recover the prior investments and support provided to infrastructure."



Research Fellow Kim suggested, "It is necessary to verify in advance the actual demand and local economic impact of large-scale AI infrastructure investments, and establish clear guidelines for the scope of public support, phased implementation conditions, and principles for cost sharing and facility utilization in the event of project cancellation, depending on the specificity of demand."


This content was produced with the assistance of AI translation services.

© The Asia Business Daily. All rights reserved. Unauthorized AI training and use prohibited.

Today’s Briefing