Home > Business > Industry

Samsung, OpenAI join hands to develop floating data centers

Published: 14 Oct. 2025, 19:05 Updated: 14 Oct. 2025, 19:49

Microsoft retrieves the data center used in its Natick Project from the ocean in 2020. [MICROSOFT]

Samsung is partnering with OpenAI to develop floating data centers on the ocean, seeking a scalable solution to the immense energy and cooling demands of AI infrastructure.

Samsung C&T and Samsung Heavy Industries are in talks with OpenAI to co-develop offshore data centers — also known as floating data centers — that can be deployed at sea, according to industry insiders on Tuesday.

The discussions followed OpenAI CEO Sam Altman’s visit to Korea on Oct. 1, during which he met with Samsung Electronics Chairman Lee Jae-yong.

The project aims to install data centers directly on the ocean, offering a way to bypass land-based space constraints while drawing cold seawater for direct cooling.

Samsung has not yet built a floating data center, but it expects to combine Samsung Heavy Industries’ experience in offshore plant development with Samsung C&T’s background in land-based data center construction to strengthen its capabilities.

“We plan to use proprietary technologies to develop floating data centers, offshore power facilities and integrated control centers,” a Samsung official said.

Korea has also been exploring underwater data centers as an alternative to the conventional land-based ones.

Since 2022, the Korea Institute of Ocean Science & Technology (Kiost) has been developing underwater server technology as part of a broader project to expand the use of undersea space. In October 2024, Kiost signed an MOU with GS E&C and Posco to cooperate on building an eco-friendly underwater data center complex off the coast of Ulsan. The plan includes installing 100,000 servers at a depth of 30 meters (98 feet) by 2030. Officials hope to complete a testbed by that year.

The drive to move data centers offshore reflects growing pressure on land-based infrastructure to keep up with surging AI workloads.

A Samsung Electronics logo and a computer motherboard appear in this illustration taken on Aug. 25. [REUTERS/YONHAP]

Korea has around 190 data centers as of 2024 and is expected to reach over 300 by 2030. But building large-scale centers on land comes with challenges: securing real estate, high construction and energy costs and opposition from local residents over noise, heat and limited job creation.

Underwater data centers, on the other hand, could reduce land and construction costs by an estimated 30 to 40 percent due to simpler structural requirements. Most importantly, they allow operators to use naturally cold seawater to cool servers more efficiently.

Land-based data centers typically spend about 40 percent of their total power on cooling. But facilities located 30 meters below sea level, where the average temperature of Korean coastal waters is 13.3 degrees Celsius (55.9 degrees Fahrenheit), could cut total energy use by up to 70 percent.

Other countries are also moving in this direction. China recently began operating its first commercial underwater data center off the coast of Hainan at a depth of 35 meters, according to the South China Morning Post. Microsoft tested a similar concept through its Project Natick in 2018, deploying an experimental underwater data center off the coast of Scotland’s Orkney Islands.

But technical challenges remain. Saltwater corrosion, marine life such as barnacles and jellyfish clogging pipes or attaching to heat exchangers and the risk of environmental pollution from coatings and materials all present obstacles.

“To address these issues, we are considering options like indirect cooling using closed-loop heat exchangers,” said Han Taek-hee, a researcher at Kiost. “As demand for AI data centers explodes globally, the importance of seawater-based cooling technologies will only grow.”

This article was originally written in Korean and translated by a bilingual reporter with the help of generative AI tools. It was then edited by a native English-speaking editor. All AI-assisted translations are reviewed and refined by our newsroom.
BY NA SANG-HYEON [[email protected]]