The Substation Seat Is Only Sold Once — Busan's 'AI Hub' at a Crossroads: Rent Out Power Slots or Use Them as a Regional Industrial Spark Plug

Busan City and South Gyeongsang Province are calling for an 'AI hub.' They aim to attract data centers and make the Southeast region Korea's AI heart. Press releases count jobs and investment amounts. But there's one thing those numbers don't ask. What incoming data centers actually consume isn't land—it's power.

Land Is Abundant, Seats Next to Substations Are Precious

GPU racks barely use any land. They simply drink electricity wholesale. So the real variable in location isn't price per pyeong. It's the distance to Korea Electric Power Corporation (KEPCO) substations, power receiving capacity, and the grid connection queue. The Seoul metropolitan area's queue is already blocked. New connections are backlogged for years. The card that Busan and South Gyeongsang have put forward is actually a detour around that.

There's a point to pause here. If what we're selling isn't 'land' but 'empty slots in the power grid,' this is a business of renting out bottleneck assets that Busan holds. The side collecting the toll isn't the local government that provided the site. It's operators like SK and Naver who layer computation on top, and KEPCO, which controls the transmission grid. Busan paves the road. The fees are collected elsewhere.

Building or Extracting?

The key question is this: Is it on the building side or the extracting side? For data centers to 'create' value in the Southeast region, something local must be integrated on top of that power and cooling infrastructure. Things like port logistics AI, shipbuilding and machinery simulation demand, and research compute from Pusan National University and maritime clusters. Only then does the data center become not an endpoint but a shared input that the next 100 local companies can plug into.

The opposite scenario goes like this: Servers run using local power, greenbelt relaxation, and tax breaks, but all that computation drains out to Seoul and global cloud demand. Clean power is pre-emptied by tenants, while grid congestion costs and cooling water burdens are shouldered by the region. Profits flow outward, costs stay in the Southeast. That's where the symmetry breaks.

If you flip the incentive structure one layer back, the fork becomes clearer. Why do local governments rush to hand over slots all at once? Because attraction achievements to showcase within their term and investment amount announcements for ribbon-cutting ceremonies work as political incentives. Quick announcements are rewarded; slow industrial ignition is not.

The counterargument is fair: "Aren't data centers just ordinary real estate development, profitable enough from tax revenue and construction jobs alone?" Partly true. But that's one-time. Once you hand over a scarce asset like a power slot, regional industries can't ride on top of it—only the slot gets consumed. When the next AI company comes, there's no substation seat left to offer. That's the real cost.

Coordinates in the U.S.-China Game, and Tomorrow's Question

After CHIPS, the U.S. has bundled AI infrastructure itself as a security asset. It views power, semiconductors, and data centers as one package. Korea isn't a subcontractor in that chain—it sits in a position holding the physical bottleneck of the power grid. The crossroads is whether the Southeast will use it as a rental business or as a spark plug for industrial ignition.

If you're a Busan reporter, ask this tomorrow: Not the investment amount in the press release, but KEPCO's power receiving contract capacity and grid connection timeline. One signal is enough to test the hypothesis: Do the proportion of regional demand in the tenant contract and the ratio of local PPAs to KEPCO power contracts show meaningful numbers? Or is only the power draining out?