What Are the Conditions for Success of Data Centers? A Map of AI Data Centers in Korea – Bitplanet

The operational readiness status and breakeven conditions of four domestic AI data center projects.

[Block Media Bitplanet] What are the factors that determine the success or failure of a data center? Profitability is not guaranteed solely by securing power and demand. Various conditions such as rental rates, occupancy rates, and financing conditions have been compared and analyzed from multiple angles.

Major Data Centers in Korea as Seen on the Map

Four large projects under construction or in progress (Chapters 02-05), four large assets currently operational (Chapter 00), and two cases of power approval and financing (Chapters 01 and 06) are marked at their respective city and county locations. The power consumption of data centers follows the official announcements of each company or institution, and since the criteria (power supply capacity, business scale, etc.) differ, they are not aggregated. The locations of Kakao Ansan and Yongin Deokseong, which do not address power consumption in the text, are noted only.

EXECUTIVE SUMMARY

Key Conditions for Securing Power and Demand and Achieving Breakeven for Data Centers

Key Insight Profitability is not guaranteed solely by securing power and demand. The timing of achieving the breakeven point (BEP) varies depending on actual rental rates, occupancy rates, and financing conditions. This report compares and analyzes the operational readiness status of four data centers in Paju, Ulsan, Haenam, and Gumi, as well as breakeven conditions by business type based on publicly available data.

Key Findings From August 2024 to March 2026, a total of 522 applications were submitted for the first technical review of the power system impact assessment for data centers in the metropolitan area, with an application capacity of 33,592 megawatts (MW). However, only 10 projects (1,010 MW) received final supply approval. Among the four data centers analyzed, Paju AIDC (LG Uplus) is the only one that has completed both power supply confirmation and leasing contracts for one building before completion.

Key Monitoring Indicators Future monitoring should focus on the power system impact assessment announcements and the detailed criteria of the special law on AI data centers (AIDC), as well as trends in the main review approvals in the metropolitan area. Additionally, compliance with the completion target of June 2027 for Paju Building 1 and the operational goal of November 2027 for Ulsan Phase 1 (41 MW) are also key monitoring targets. Once actual leasing conditions and revenue/cost data after operation are disclosed, the breakeven point can be estimated more concretely.

AT A GLANCE

Comparison of Power, Demand, and Capital for Four Large Projects

The power acquisition, leasing demand, capital procurement status, and operational goals of Paju, Ulsan, Haenam, and Gumi were compared. This is not a comprehensive survey of all data centers in the country nor a representative sample. The items included in the business type and investment costs differ, so economic superiority cannot be determined solely by the scale of investment. The status of power acquisition is organized based on the official announcements of each project entity, and projects at different administrative procedural stages are not compared on the same level.

Source: [9][10][11][12][7][13][8][14][15][3]. The capacity of data centers is referenced from announcements by each company and the government, and is distinguished from IT load for server supply. The Ulsan schedule is based on the announcement as of June 2025. It is not possible to simply compare the unit price per MW as the items included in the investment amount differ by project.

Status of Four Major Data Centers Currently Operational

Currently operational major data centers include the company-exclusive 'Naver Gak Sejong' and 'Kakao Ansan', as well as rental facilities 'Hanam IDC' and 'Epoch Anyang'. Hanam IDC has a power supply capacity of 40 MW, and Epoch Anyang (40 MW) was completed in early 2025. Naver Gak Sejong, which opened in November 2023, is designed to reach a maximum power supply capacity of 270 MW upon completion of all six phases of expansion, with approximately one-sixth of the total space currently operational as of October 2025. Naver Cloud has announced plans for the second and third phases of expansion, but the actual construction status has not yet been confirmed. It is necessary to clearly distinguish between the operational and expansion status of existing assets and the supply plans of the four new projects.

POWER APPROVAL

Only 10 Power Supply Approvals in the Metropolitan Area

The supply of power for new data centers in the metropolitan area is very limited. Among the 522 applications submitted for the first technical review conducted from August 2024 to March 2026, 53.4% (279 cases) were deemed 'not supplyable'. Of the 24 cases that entered the main review, only 10 projects (including one in Seoul) passed the final approval. In contrast, in non-metropolitan areas, 187 out of 214 applications for the first technical review received a 'supplyable' notification, creating a stark contrast. The final approval count in the metropolitan area (10 cases) is only 1.9% of the cumulative application count (522 cases). However, since some cases are still under review, this ratio cannot be viewed as the final approval probability for individual projects.

As of the end of March 2026, out of the cumulative application capacity of 63,846 MW nationwide, 9,583 MW remains undecided regarding supply availability. Meanwhile, the GS AI Data Center in Donghae, Gangwon, passed the power supply review for Phase 1 (100 MW) in September 2026 and is set to begin construction in November.

The power system impact assessment is a procedure that examines the impact of large-scale power demand on the national power grid in advance. With the enforcement of the "Special Act on the Activation of Distributed Energy" on June 14, 2024, the power authorities have started collecting relevant statistics from August of the same year. The detailed operational standards were announced for administrative notice in July 2026, but as of September 15, the final announcement of the regulations has not been confirmed.

Meanwhile, the "Special Act on the Promotion of the AI Data Center Industry (AIDC Special Act)" will come into effect on March 10, 2027. This law includes exemptions from the power system impact assessment for the construction of new AIDCs in non-metropolitan areas, the expansion of existing facilities, and the conversion of existing data centers for AI use. The criteria for the target power capacity will be determined by presidential decree. The Ministry of Science and ICT held a discussion meeting on September 9, 2026, for the establishment of subordinate regulations. Therefore, the applicability of exemptions for the projects in Ulsan, Haenam, and Gumi should be closely examined based on the capacity criteria and detailed requirements for each project in the future.

Paju AIDC: Secured 200 MW Power and Completed One Building Contract

Business Type: Rental. The structure generates revenue by leasing data center space and infrastructure to clients. The rental rate and the capacity that generates revenue are key performance indicators.

Current Status and Facts: LG Uplus Paju AIDC is set to begin construction in May 2025 with a scale of 200MW. The goal for the completion of the first building is June 2027, with plans to sequentially operate a total of four buildings by 2028. The approved investment amount through the board of directors is 615.6 billion KRW for the first phase and 1.3489 trillion KRW for the second phase, totaling 1.9645 trillion KRW. The company has stated that it will fund the investment with its own cash without external borrowing. The supply of 200MW of power has been confirmed, and according to the company, the lease contract for the first building has already been fully completed before construction. As of June 5, 2026, the progress rate is approximately 20%.

Analysis and Evaluation: It is difficult to calculate the exact operating break-even point due to the confidentiality of specific contract capacities (MW), duration, rental unit prices, and whether pre-lease contracts for buildings 2 to 4 exist. Although the company mentions an upward trend in rental unit prices, the actual level of the signed rental fees has not been confirmed. Minimizing the time lag (Ramp-up period) from contract signing to actual revenue generation is essential to reduce initial financial burdens.

03 SK AIDC Ulsan: Securing Construction Volume Power... 1GW Expansion in Negotiation Stage

Business Type: Commercial (SK·AWS Collaboration). This is a project jointly promoted by SK Group and Amazon Web Services (AWS). The rental and revenue-sharing conditions are not disclosed, making it difficult to compare with typical rental-type businesses.

Current Status and Facts: SK Telecom, SK Eco Plant, AWS, and Ulsan Metropolitan City held a groundbreaking ceremony on August 29, 2025. SK Broadband and other affiliates are participating in the project, with SK Eco Plant responsible for infrastructure construction. Earlier, in June 2025, SK and AWS signed the main construction contract. The goal announced by the Ministry of Science and ICT at that time was to prioritize the operation of phase 1 (41MW) by November 2027 and full operation of 103MW by February 2029. The total project cost of 7 trillion KRW includes an investment of 4 billion USD from AWS.

According to media reports in September 2026, the power supply for the currently under-construction initial facilities has been completed. However, SK has not officially announced the exact capacity due to a non-disclosure agreement (NDA). SK Chairman Chey Tae-won stated on September 11 that negotiations for an additional expansion of approximately 900MW have made significant progress, but Korea Electric Power Corporation explained that no official administrative procedures for additional supply have been received. The discussion of expansion negotiations does not necessarily mean final approval for power supply.

Analysis and Evaluation: It cannot be definitively concluded that cooperation with AWS directly translates into confirmed rental income for SK. The detailed breakdown of the total investment amount (construction costs, equipment costs, etc.) is not clearly distinguished, making direct comparisons with the facility investment costs of other projects limited.

04 National AI Computing Center Haenam: Establishing 40MW Public AI Resources, Target Completion in 2028

Business Type: Public Infrastructure. This is a public-private partnership project providing AI computing resources to academia and research institutions. Revenue from usage fees and the achievement of public computing resource supply are the main evaluation criteria.

Current Status and Facts: The National AI Computing Center in Haenam Solar City will be established with a power capacity of 40MW and 15,000 AI semiconductors. The groundbreaking ceremony was completed on August 3, 2026, with a target completion in 2028.

Regarding project cost estimation, the Ministry of Science and ICT announced over 2.5 trillion KRW (2026-2030), while Haenam County announced 2.9 trillion KRW (announced in June 2026). The differences in the detailed items included in the announcements from the two institutions have not been clearly confirmed. Meanwhile, the special purpose corporation (SPC) 'CoArc', in which Samsung SDS is the largest shareholder, has submitted an electricity usage application to Korea Electric Power Corporation. From 2027, some computing capacity will be supplied early using existing data center resources from Samsung SDS.

Analysis and Evaluation: Haenam will provide usage rights and fee discounts to startups and research institutions. Therefore, when evaluating performance, it is necessary to consider not only rental income but also the achievement of the policy goal of supplying public AI infrastructure. Additionally, completing the electricity usage application does not guarantee that power supply approval has been confirmed.

05 Samsung SDS Gumi: 60MW for AI·GPU Services... Construction Status Unconfirmed

Business Type: Self-Operated Service. Samsung SDS directly operates the data center infrastructure and provides AI·GPU cloud services to clients. Service revenue and equipment investment costs determine the profit and loss structure.

Current Status and Facts: Samsung SDS has decided to construct a 60MW AI data center at the Gumi site and announced an investment of 427.3 billion KRW on January 2, 2026. This amount covers building and equipment construction costs, and additional investments may be made in the future as infrastructure expands, such as AI GPUs. Initially, construction was aimed to start in July 2026 and operate by March 2029, but as of September 15, the actual construction status and results of the power system impact assessment have not been confirmed.

Analysis and Evaluation: It is not possible to simply compare the building and basic facility investment amounts with other data centers in terms of unit price or cost advantage. To assess the profitability of the AI·GPU services that Samsung SDS will operate directly, a comprehensive review of future service revenue forecasts and the scale of additional equipment investments, such as expensive GPU servers, is required.

BREAK-EVEN

06 Comprehensive Analysis of Break-Even: Differences in Contract Conditions and Funding Structure

When comparing break-even points, it is essential to accurately understand the scope of included costs. The 'operating break-even' mentioned in this report refers to the level that covers pure operating expenses (OPEX) and interest costs on borrowings. This differs from the accounting break-even that reflects depreciation costs and should be distinguished from the payback period for fully recovering the initial capital invested.

Rental Fees and Contract Conditions. The rental fees for data centers in the metropolitan area have increased from around 140,000 KRW per kilowatt (kW) per month in 2019 to around 250,000 KRW per month in 2025. The proportion of confirmed occupancy and pre-leases among new supply volumes in the metropolitan area was high, with 99.7% in 2024 and 99.4% in 2025, and it was found that 16% and 23.5% of the planned volumes for completion in 2027 and 2028, respectively, had completed pre-lease contracts. However, since the remaining time until the completion date varies, it is difficult to conclude that market demand has slowed down solely based on the numerical differences in pre-lease rates.

In the domestic market, a contract structure that offers initial rental discounts and fixes the annual increase rate at about 2% is common. In contrast, global large contracts of 10MW or more are typically set for long-term periods of over 10 years, with annual increases of 3% or linked to the consumer price index (CPI).

For example, Hanam IDC charges rental fees per kW for the contracted usage capacity and separately settles electricity fees per kilowatt-hour (kWh).

Key Reference Cases: Gwacheon IDC and Hanam IDC

Gwacheon IDC: LS Securities estimated the average operating rate required for Gwacheon IDC to achieve break-even by 2025 to be about 25%. This estimate considers the increase in depreciation costs and interest burdens, and there may be differences from actual achievement. This operating rate cannot be uniformly applied to other data centers.

Hanam IDC: According to Macquarie Infrastructure's Q1 2026 IR materials, Hanam IDC has signed lease contracts for 99% of the target IT load (25.44MW). The company expects the point at which rental income occurs from the entire target IT load to be around mid-2027. Even if lease contracts are signed, there is a time lag before actual revenue flows in, and this point does not necessarily mean that break-even has been achieved.

Funding. Even if the power system is approved and tenants are secured, failure to secure funding can halt the project. A data center development project in the metropolitan area in 2026 signed a 20MW electricity usage contract with KEPCO but failed to pay the land balance on time, resulting in the loss of the deadline benefit (EOD) and an early repayment demand from creditors. In contrast, the Yongin Deokseong-ri AI data center (total project cost of approximately 1.4 trillion won), which did not disclose its anchor tenant (core tenant company), successfully secured 710 billion won in project financing (PF) on April 7, 2026. This illustrates why one should not hastily judge the possibility of funding based solely on tenant disclosure.

Project-Specific Verification Items

Paju AIDC: It is essential to ascertain the actual contract rental price, initial discount conditions, contract volume and duration, and the pre-leasing rate for buildings 2 to 4. Additionally, the timing of rental income collection after the completion of building 1, as well as the levels of operating costs and interest burden from borrowing, should also be verified.

SK AIDC Ulsan: It is necessary to check whether the phase 1 facilities currently under construction are operating normally. Confirm whether the contract structure between SK and AWS is a simple lease or a revenue-sharing model, and understand the ratio of capital investment and operating cost sharing. The application and approval status for an additional 900MW expansion and the final investment confirmation should also be examined.

National AI Computing Center Haenam: Before operation, it is crucial to verify the final approval for 40MW power supply and compliance with the completion schedule. After operation, it is necessary to review the level of public subsidies, the fee structure, the actual computational resource provision and utilization rates, and the cost burden due to fee subsidies.

Samsung SDS Gumi: The priority is to confirm whether construction has actually commenced and whether the power system impact assessment has been passed. Once these two conditions are verified, it is essential to check the scale of additional GPU and server equipment investments, the status of customer demand acquisition, and the revenue trends from AI and GPU services.

CONCLUSION

C Rental Rates, Occupancy Rates, and Profitability Gaps Due to Funding

The point at which a data center achieves its break-even point is determined by rental rates, occupancy rates, and funding conditions. Even for facilities of the same scale, the timing of income generation based on the actual rental rates agreed upon, the contract capacity, and the interest burden from financial borrowing can vary significantly. Even if a pre-lease contract is signed, if the actual move-in and rental income collection from the client company are delayed, the initial financial burden will inevitably increase.

Market Impact and Implications

Significance to the Market. For operators who have secured power, it is crucial to acquire customers and increase server utilization rates. The completion and operational target dates for the four data centers analyzed in this report are between 2027 and 2029. According to data compiled by CBRE in July 2026, the proportion of pre-leased data center capacity in the metropolitan area is only 16.0% for 2027 and 23.5% for 2028. As new data center supply gradually becomes visible, the speed at which uncontracted capacity is converted into lease contracts could determine the profitability of each operator.

However, this pre-leasing rate is based on the IT load of metropolitan area data centers and does not directly represent the actual demand for the four AIDCs. CBRE interpreted this as a movement by client companies to secure leased space in advance due to concerns about future power shortages. If the completion of new data centers is delayed due to power system connection delays or if remaining capacity is quickly depleted, there is a possibility that the overall market's 'supply shortage' will continue to drive market trends, rather than competition among operators for customer acquisition.

Operator Verification Conditions. Operators must secure flexibility to phase and adjust capital investment (CAPEX) and funding execution according to the actual move-in and utilization schedules of tenant customers. Especially for operators utilizing external borrowings, it is essential to have an emergency fund plan to withstand the fixed operating costs and financial interest burdens that arise when completion or operation is delayed. In reality, as seen in the case of Hanam IDC, even with a high rental contract rate, there is a time lag before the total contracted rent is received. Therefore, operators must accurately calculate the time lag from the contract signing to the actual revenue generation and reflect it in their cash flow plans.

Future Verification Indicators. In the future, it will be necessary to track the normal implementation of target completion and power supply schedules, as well as the pre-leasing rate trends of quantities belonging to the same completion year over time. After the center is operational, it is necessary to diversify key management indicators according to the business model. For rental-type data centers, it is crucial to strictly differentiate and track contracted capacity, actual server operating capacity, and the capacity for which actual rent is charged. Additionally, integrating the analysis of applicable rental rates with borrowing interest rates and repayment schedules will allow for accurate assessment of the impact of increased contracted capacity on actual revenue and financial costs. For public infrastructure types, the allocation of public AI computational resources by institution and company, as well as the actual utilization rates of computational resources, should be considered key indicators. For self-operated service models, monitoring the operating rates of high-cost servers and equipment and the resulting trends in cloud service revenues should be a focus.

REFERENCES

R Sources

[1] 'Metropolitan Area Data Centers' are flooding in... but the power grid is 'overwhelmed' → Is the countryside an alternative? · Money Today · 2026-05-23 https://www.mt.co.kr/tech/2026/05/23/2026052210211399740

[2] The power approval rate for metropolitan area data centers is only '1.9%' (CBRE Korea "Korea Data Center Investment" report introduced on 2026-07-12) · KHARN · 2026-07-14 https://www.kharn.kr/news/article.html?no=31267

[3] LG U+ "5 trillion won orders for AIDC by 2030... proposing AI infrastructure standards" (1st) · LG Uplus Newsroom · 2026-06-08 https://news.lguplus.com/21965

[4] Large-scale power demand sites are pre-verified... promoting the establishment of operational standards for power system impact assessments · Energy Daily · 2026-07-09 https://www.energydaily.co.kr/news/articleView.html?idxno=201248

[5] Special Act on the Promotion of the Artificial Intelligence Data Center Industry (Law No. 21759, enacted on 2026-06-09, effective on 2027-03-10) (1st) · National Law Information Center · 2026-06-09 announced https://www.law.go.kr/lsInfoP.do?lsiSeq=286707&viewCls=lsRvsDocInfoR

[6] "Designating data center special zones and power permits"... The Ministry of Science and Technology has begun discussions on subordinate laws for the AIDC Special Act · AI Times · 2026-09-09 https://www.aitimes.com/news/articleView.html?idxno=215050

[7] The Ministry of Science and ICT listens to voices on-site for AI G3 with the launch of 'Ulsan AI DC' · Aju Economy · 2025-06-20 https://www.ajunews.com/view/20250620162359703

[8] LG Uplus invests an additional 1.3489 trillion won in Paju AIDC · Smart Biz · 2026-07-29 https://www.smartbizn.com/news/articleView.html?idxno=149817

[9] [On-site] "As a core country in the AI supply chain"... Government and private cooperation 'National AI Computing Center' begins construction · ZDNet Korea · 2026-08-03 https://zdnet.co.kr/view/?no=20260803162041

[10] Construction of the National AI Computing Center in Haenam Solasido started in July 'smoothly' · Newsis · 2026-06-25 https://www.newsis.com/view/NISX20260624_0003681878

[11] Groundbreaking Ceremony for 'SK AI Data Center Ulsan' Held (Phase 1) · SK Telecom Newsroom · 2025-08-29 https://news.sktelecom.com/214708

[12] 103MW Power Secured... Solution for '1GW Wall' at SK Ulsan AIDC · Cookie News · 2026-09-15 https://www.kukinews.com/article/view/kuk202609140213

[13] Samsung SDS, Gyeongsangbuk-do, and Gumi City Sign MOU for Construction of Gumi AI Data Center (Phase 1) · Samsung SDS Newsroom · 2026-01-08 https://www.samsungsds.com/kr/news/gumi-260108.html

[14] LGU+ Increases Profitability with 2 Trillion Won Investment in Paju AI DC · NewsPim · 2026-08-06 https://www.newspim.com/news/view/20260806001016

[15] LGU+ Prepares for the 'Era of Vera Rubin'... Reveals Next-Generation Infrastructure Strategy at Paju AIDC · Digital Daily · 2026-06-07 https://www.ddaily.co.kr/page/view/2026060709273226967

[16] Samsung SDS Promotes Construction of 60MW Gumi AI Data Center... Operational by 2029 · ZDNet Korea · 2026-01-08 https://zdnet.co.kr/view/?no=20260108111816

[17] Kakao Opens First 'Own Data Center'... "Stability is the Best in the Country" · NewsPim · 2024-06-11 https://www.newspim.com/news/view/20240611000836

[18] Naver's 'Gak Sejong' Data Center to Support for 10 Years... Opens the Hyper Scale Era · Bloter · 2023-11-06 https://www.bloter.net/news/articleView.html?idxno=608037

[19] Hanam Data Center (Asset Overview: Completion Date 2023-11-28, Power Supply 40MW·IT Load 25.44MW, Tenant LG CNS) (Phase 1) · Macquarie Korea Infrastructure Investment Co. · Issuance Date Unknown https://www.mkif.com/ko/assets/hanam-data-center.html

[20] "If This Goes On, Everything Will Be Taken Away" Overseas Big Players Snatch Up Korean Prime Assets · Edaily Marketin · 2025-02-18 https://marketin.edaily.co.kr/News/Read?newsId=01144726642071832

[21] Naver Cloud to Start Construction of Gak Sejong Phase 2 and 3 'As Early as the End of the Year' · The Bell · 2025-10-27 https://www.thebell.co.kr/front/newsview.asp?key=202510270910271020102172

[22] Local AIDC Power Evaluation 'Bottleneck'... Waiting for 10GW · Seoul Economic Daily · 2026-08-03 https://www.sedaily.com/article/20075225

[23] Donghae AI Data Center Power Supply Review Passed... Construction Expected to Start in November · Gangwon Ilbo · 2026-09-07 https://www.kado.net/news/articleView.html?idxno=2070895

[24] Special Law to Activate Distributed Energy to Take Effect from June 14 (Phase 1) · Ministry of Trade, Industry and Energy · 2024-06-13 https://www.motir.go.kr/kor/article/ATCL3f49a5a8c/169176/view?displayAuthor=&endDtD=&mno=&pageIndex=1&rowPageC=0&schClear=on&searchCategory=0&searchCondition=1&searchKeyword=&startDtD=

[25] National Assembly Enacts 'Special Law for Promotion of AI Data Center Industry'... Designates National Strategic SOC and Special Zones · Daily Korea · 2026-05-07 https://www.hankooki.com/news/articleView.html?idxno=332202

[26] "SK Ulsan AIDC 900MW Expansion Making Significant Progress" · Financial News · 2026-09-13 https://www.fnnews.com/news/202609131805554245

[27] Macquarie Infrastructure (088980) Adds Digital Assets by Acquiring Hanam Data Center · Samsung Securities Research · 2024-07-31 https://www.samsungpop.com/common.do?cmd=down&contentType=application%2Fpdf&inlineYn=Y&saveKey=research.pdf&fileName=2010%2F2024073109544577K_02_03.pdf

[28] It's Time for Change, Finding Beneficiaries of Cloud Transition (Estimated Break-Even for Gwacheon IDC, p.48) (Phase 1) · LS Securities Research · 2025-07-08 https://msg.ls-sec.co.kr/eum/K_20250708_31922_15.pdf

[29] Q1 2026 IR Materials (Status of Hanam Data Center Operations and Construction, p.13) (Phase 1) · Macquarie Korea Infrastructure Investment Co. · 2026-04 https://www.mkif.com/assets/mkif/ko-kr/investor-centre/public-filings-and-reports/2026/2604-ir-general-presentation-1Q-2026-kor.pdf

[30] AI Changing Real Estate Finance Trends... EOD Nan Data Center Also Finds New Owner · Invest Chosun · 2026-07-22 https://www.investchosun.com/site/data/html_dir/2026/07/21/2026072180105.html

[31] Korean Data Center Investment: Diagnosing Scarcity Premium and Exit Possibilities Created by Supply Constraints (pp.20, 24, 31) (Phase 1) · CBRE Korea · 2026-07-12 https://www.cbrekorea.com/insights/reports/한국-데이터센터-투자

Limitations of A Data and Calculation Basis

① Limitations of Analysis and Data

• Scope of Verification: Four construction and promotion projects presented in public announcements with a scale exceeding 40MW were compared. The definition of capacity varies by project. The analysis reference date is September 15, 2026. The status of construction in Gumi has not been confirmed. The Gwacheon IDC and Hanam IDC were used as reference cases to examine break-even estimates and rental income generation schedules.
• Selection Criteria: Projects were selected from Paju, Ulsan, Haenam, and Gumi, where project entities, investment plans, and operational or completion goals could be verified from public documents. Differences among rental-type, commercial-type cooperative projects, public infrastructure-type, and self-operated/service-type were compared in the four cases. This does not represent a comprehensive investigation of all domestic projects meeting capacity criteria or a representative sample.
• Cases Excluded from Detailed Comparison: Detailed comparisons by business model were limited to the four projects mentioned above. GS Donghae was introduced as a reference case that passed the power supply review. Each of Sejong, Kakao Ansan, Hanam IDC, and Epoch Anyang, which are already operational, were treated as reference cases separate from new construction and promotion projects. Yongin Deoksung was introduced only as a funding case.
• Scope of Market Data: Rental rates, pre-lease rates, and contract practices were sourced from KHARN's secondary citations and cross-referenced with the original CBRE Korea documents (pp.20, 24, 31). The pre-lease rate is based on the IT load (MW) of supply assets in the metropolitan area and does not represent the entire domestic AIDC. The kW basis for rental rates and whether electricity costs are included were not specified in the original documents.
• Discrepancies Between Data: The project cost for the National AI Computing Center was announced by the Ministry of Science and ICT as over 2.5 trillion KRW, while Haenam County announced it as 2.9 trillion KRW, resulting in a discrepancy.
• Limitations of Case Application: Approximately 25% of the Gwacheon IDC is the average operating rate estimated by LS Securities for cost and revenue projections in July 2025. This is not an actual achievement or an industry common standard. The 99% of Hanam IDC is the contract rate, and mid-2027 is when the company expects the total contracted rental income to begin. These two cases cannot be directly compared due to differences in indicators and cost ranges.
• Scope of Investment Cost Estimation: Haenam includes AI semiconductor construction, while Gumi focuses on buildings and basic facilities. The total investment amount for Ulsan also cannot be directly compared due to the inability to distinguish detailed included items. The 1.9645 trillion KRW for Paju is the sum of the approved amounts for phases 1 and 2. It could not be confirmed whether the separately presented land acquisition amount of 105.3 billion KRW is included in this total. Therefore, the land acquisition amount was not added, nor was this total indicated as including land.
• Limitations of Procedure Status: The status of the power system impact assessment notification, the commencement of Samsung SDS in Gumi, and the commencement of the second and third expansions in Sejong could not be confirmed as of September 15, 2026.
• Nature of Sell-Side Data: The data from Samsung Securities and LS Securities are research reports analyzing the respective stocks. They were used to verify the basis for contract structures and break-even estimates, and investment opinions and target prices were not cited.

② Basis for Major Figures Calculation

The ratios and totals calculated directly were distinguished from estimates and forecasts provided by sources.

③ Methodology

• Data Priority: Official data from companies and regulatory agencies → Original data and aggregated database → Specialized media → Syndication media → Community data in that order.
• Data Classification: Only when the original document is directly verified is it classified as primary data. If the original document cannot be obtained, it is indicated as a secondary citation, and the fact of not securing the original document is disclosed.
• Reference Date and Inquiry Date: Values confirmed at a specific point in time are marked with a reference date, while real-time fluctuating dashboard figures are marked with an inquiry date.
• Handling Conflicting Figures: When different figures are confirmed, both are presented together. If a reference value must be chosen, the public disclosure of the estimation range, clarity of the inquiry time, and the possibility of separating detailed items are prioritized over the media's recognition, and the basis for selection is specified in the text.
• Self-Calculation: The formulas for directly calculated ratios and totals and the basis for estimates and forecasts provided by sources are presented in Appendix ②.
• Legal and Regulatory Interpretation: Regulatory agency data is cited only within the scope of explaining general structural classifications and applicability, and is not broadly interpreted as official classifications or approvals for specific products. Explanations of product structures and regulatory interpretations are separated into distinct paragraphs.
• Disclosure of Conflicts of Interest: All reports include the same disclosure statement in a fixed position at the top of the document.

Published by Bitplanet Research Lab · Written by Kim Tae-won · Reviewed by Kim Soo-young

Disclaimer This material is an industrial analysis for informational purposes and is not an investment solicitation. It does not recommend the trading of specific stocks and does not present price forecasts or target prices. The author is not a lawyer or investment advisor, and all figures are based on public sources (as of the time of writing, subject to change thereafter).

[Block Media Kim Yoon-hye]

Comments 0