More than 500 large-load applications have entered first-stage grid-impact review around Seoul, highlighting the growing importance of power availability in data center development.
South Korea has no shortage of interest in building data centers.
The harder question is how many projects can secure enough power, in the locations developers want, on the timetable their customers expect.
The pressure is most visible in the Seoul metropolitan area.
According to CBRE Korea, by March 2026 there had been 522 first-stage technical review applications under the Power System Impact Assessment process, representing a combined 33,592 MW of requested capacity.
Only 10 cases had received final power-supply approval, according to CBRE. AI-DC KOREA, drawing on underlying public data, puts their combined capacity at about 1,010 MW.
CBRE describes the 10 approvals as a 1.9% final approval rate.
That figure is striking, but it needs context.
It does not mean 98.1% of proposed data centers have been permanently rejected. The 522 applications sit at different stages of the review process, and the 33.6 GW figure represents requested capacity, not a construction pipeline.
AI-DC KOREA makes the same methodological distinction: the 10 approved cases represent a cumulative point-in-time outcome, not an individual project’s probability of eventual approval.
The more useful conclusion is that power access has become a significant constraint on data center development around Seoul.
Seoul still concentrates demand — and power
The Seoul metropolitan area remains the center of South Korea’s data center market.
AI-DC KOREA, using KEPCO data, calculates that at the end of 2023 Seoul, Incheon and Gyeonggi together accounted for 90 data center customer contracts and 1,403 MW of contracted power.
That represented about 60% of customer contracts and 70.6% of contracted power nationwide.
Those figures are not IT load. They refer to contracted electricity capacity.
The capital region remains attractive because of its concentration of customers, connectivity and digital infrastructure.
But that concentration also increases pressure on the power system as developers seek to add large new loads.
CBRE says new development is increasingly shifting toward what it calls a “power-following” structure, with locations being screened more heavily according to power availability.
That changes the site-selection equation.
Demand, connectivity and land remain important, but they do not by themselves make a site developable if sufficient electricity cannot be secured.
Grid assessment is becoming part of the development process
South Korea’s regulatory framework makes the relationship between large new loads and the power system explicit.
Under the Distributed Energy Promotion Act and its implementing rules, large electricity users above specified thresholds in designated areas are subject to a Power System Impact Assessment.
The assessment considers whether the system can maintain reliability after the proposed load is connected, the difficulty of constructing or reinforcing substations and transmission infrastructure, and measures that could reduce the project’s impact on the grid.
For large data centers, power planning therefore becomes part of the development process rather than simply a utility connection arranged near the end of construction.
Power feasibility can become a gating factor in whether a project can proceed.
“Power available” and “power available on time” are different questions
Technical feasibility does not necessarily mean electricity can be delivered on the timetable a developer needs.
AI-DC KOREA cites KEPCO’s standard construction period of roughly 6.5 to 7.5 years for 154 kV transmission and substation infrastructure.
That is a standard infrastructure construction period, not the average connection time for every Korean data center.
Individual projects can face very different schedules depending on existing grid conditions and the infrastructure required.
But the timeframe helps explain why connection timing matters.
For AI infrastructure, a project capable of receiving power in 2028 is not commercially equivalent to one that has to wait several additional years for transmission or substation work.
Developers therefore need to answer two separate questions:
Can the grid serve the proposed load?
And:
When can that power actually be delivered?
The distinction becomes more important as individual AI campuses require larger blocks of electricity.
Existing powered assets are becoming more valuable
The supply constraint is also visible in the operating market.
CBRE reports that vacancy in the Seoul metropolitan area remains below 5%.
Average colocation rents increased from about KRW 140,000 per kW in 2019 to approximately KRW 250,000 per kW in 2025, an increase of more than 70%.
Those increases cannot be attributed to power constraints alone. Demand growth, construction costs, financing conditions and other market factors also affect pricing.
But secured power can change the position of an asset.
An operating data center with both location and electricity already secured is different from a proposed facility that still needs to complete grid review or wait for major infrastructure work.
CBRE links that combination of secured power and location with a scarcity premium for existing assets.
The geography of new development may gradually widen
The Seoul metropolitan area will remain attractive for many workloads.
Customer proximity, network density and business demand continue to favor locations near the capital.
But large AI workloads introduce another consideration.
If securing a large block of power in one location requires substantially more grid reinforcement or a much longer delivery schedule than elsewhere, developers have an economic reason to examine alternative sites.
That does not mean data centers can simply move anywhere.
Latency, fiber connectivity, customer requirements, land and operating conditions still matter.
But some latency-tolerant training workloads may give developers more flexibility over physical location than applications requiring close proximity to end users.
South Korea could therefore see more development follow available power rather than remain concentrated exclusively around existing digital hubs.
That is a possibility, not yet a confirmed market shift.
The application numbers are a warning, not a pipeline forecast
The 522 applications should not be interpreted as 522 financed data centers.
The 33.6 GW requested should not be presented as committed development.
And CBRE’s 1.9% cumulative final-approval figure should not be treated as the probability that an individual application will fail.
Those distinctions are important.
But the administrative data still reveals a significant imbalance between the volume of large-load applications entering the process and the capacity that has so far reached final supply approval.
For AI infrastructure investors and developers, that makes power due diligence increasingly important before announced megawatts or completion dates are taken at face value.
The useful question is not simply how large a proposed data center will be.
It is whether the project has a credible path to power — and when that power can actually be delivered.
In South Korea, that is becoming an increasingly important part of determining whether a data center proposal can turn into operating infrastructure.
Sources
- CBRE Korea — data center investment press release — 2026-07-13.
- AI-DC KOREA Research — development, permitting and power census — September 2026 (as stated in the manuscript; exact date not confirmed).
- KEPCO — overhead transmission construction procedures — Publication date not specified.
- Korean National Law Information Center — linked enforcement provisions — Live legislative reference; applicable versions may change.

