Home / Technology / Ulsan Dives Deep: Underwater AI Data Centers Tested
Ulsan Dives Deep: Underwater AI Data Centers Tested
23 Aug
Summary
- Ulsan is testing underwater data centers for AI workloads.
- The project has a KRW 51.1 billion budget through 2030.
- Seawater cooling aims to reduce reliance on mechanical systems.

The South Korean city of Ulsan is actively exploring the development of underwater data centers as a solution to the escalating demands of artificial intelligence computing. This initiative seeks to overcome limitations in land availability, power infrastructure, and cooling systems that challenge conventional data centers across the nation. Partnering with the Korea Institute of Ocean Science and Technology (KIOST), Ulsan is designing a standardized subsea data center model.
The collaborative effort, running until 2030, has a total budget of KRW 51.1 billion, with significant funding from the central government. Ulsan aims to finalize this standardized model by 2030, with plans for a commercial underwater data center complex starting in 2031. The project focuses on creating a replicable, carbon-neutral template for underwater computing.
Testing will occur at depths of approximately 20 meters, utilizing seawater averaging 13.3°C annually for cooling. The design emphasizes pressure-resistant vessels and ultra-high-efficiency hybrid cooling systems. Standardized modular designs for server units and power equipment are intended to facilitate future expansion, targeting an operational efficiency of PUE 1.2.
SK Telecom joined the project in January 2026 to manage GPU infrastructure and server operations. The broader team includes various research institutions and industry experts contributing to vessel engineering and software management for AI workloads. A long-term vision includes a potential KRW 500 billion investment for a facility supporting up to 100,000 servers, contingent on the success of the initial testing phase.