Innovating Data Infrastructure in Ulsan
As the demand for artificial intelligence computing continues to surge in South Korea, the city of Ulsan is pioneering a unique solution to address the limitations of land availability and energy-intensive cooling systems. By shifting infrastructure to the seabed, officials aim to create a sustainable model for the country's digital future.
Project Scope and Collaboration
The city has joined forces with the Korea Institute of Ocean Science and Technology (KIOST) to engineer a standardized, modular subsea data center. This initiative is backed by a budget of KRW 51.1 billion (approximately $37 million), with the majority of funding provided by the central government. The partnership is slated to run until 2030, with Ulsan aiming to finalize a commercial-ready complex by 2031.
Leveraging Ocean Cooling
A primary advantage of the underwater project is the ability to bypass traditional mechanical refrigeration. Situated roughly 20 meters off the Ulsan coast, the facilities will utilize the natural ambient temperature of seawater—which averages 13.3°C—to maintain optimal server performance. Engineers are focused on achieving high efficiency, targeting a Power Usage Effectiveness (PUE) rating of approximately 1.2.
«The goal is a working, carbon-neutral template for underwater computing that other sites could later replicate,» report project planners.
Technological Integration and Future Scale
In early 2026, SK Telecom joined the project to oversee the management of GPU-based infrastructure and server operations. Beyond the initial testing phase, there is a long-term conceptual plan that envisions a much larger deployment, potentially involving an investment of KRW 500 billion ($360 million) to support up to 100,000 servers. Whether this expansion becomes a reality will be determined by the success and stability of the initial underwater trials.
Ultimately, this venture represents a strategic pivot for South Korea. By moving away from crowded urban centers and leveraging the ocean, developers hope to provide a scalable, energy-efficient answer to the massive computational needs of the modern AI era.