SIGNAL
AI's land, power, and cooling footprint is forcing a new municipal question: where do you put compute when every traditional site runs out of room at once? Ulsan is treating that collision as a design problem rather than a capacity crisis, and what it has publicly scoped is a public-municipal template for subsea data centers, not a one-off stunt.
The mechanism is portable. A city government teams with a research institute for the engineering and a telco for operations; the central government underwrites the design phase; the result is a standardized vessel and modular server architecture that the next coastal city can adopt without reinventing the work. Seawater does the cooling, but the shore still has to feed the power, and that constraint seawater does not solve.
The exhibit in TechRadar's reporting on the Ulsan and KIOST partnership makes the template legible: a combined public budget of roughly $37 million through 2030, with about $29 million from the central government, funds a standardized subsea data center model other coastal sites could later replicate. SK Telecom joined in January 2026 to take the operational seat.
The question the template answers, and the one it leaves open, is the same: whether a municipal design can move faster than the grid. If Ulsan ships a working vessel at PUE 1.2 (a measure of overhead energy beyond compute, where lower is better) by 2031, the lever generalizes. If it cannot, the next coastal city inherits the same waiting list.
Reported by Sky for Type0, from South Korean city bets on the ocean to solve its AI data center crisis — but this is only the beginning. Read the original: techradar.com