AI Briefing
KO

[Trend Analysis] A Complete Breakdown of Floating Data Centers (FDC): Technical Structure and Infrastructure Potential of Marine Data Centers

·2026.03.12 14:32

Key point

Floating Data Centers (FDC), which leverage offshore infrastructure to overcome the power, land, and cooling limitations of land-based data centers, have entered the commercialization stage.

1 / 2

Details

With the rapid growth of cloud and AI models, land-based data centers face five major bottlenecks: power shortages, land depletion, cooling limitations, carbon regulations, and permitting issues. As a key alternative to solve these problems, Floating Data Centers (FDC), or marine data centers, are emerging.

FDCs are not simply offshore structures but hybrid infrastructure platforms that integrate power, land, cooling, and networking. They have now moved beyond the PoC stage and are on the verge of actual commercialization, broadly divided into three models.

  • Barge Type: A practical model moored near ports that utilizes land-based power
  • Offshore Platform: A model directly connected to offshore wind power generation to bypass power grid bottlenecks
  • Underwater/Submerged: A model that submerges modules at a depth of 30~100m to maximize cooling efficiency (Microsoft Project Natick is a representative example)

FDCs achieve high efficiency at the PUE 1.10~1.18 level through seawater cooling, and can reduce Opex by 25~35% compared to land-based facilities. They also have the advantage of dramatically shortening the power procurement period by directly sourcing power from offshore wind.

Going forward, FDCs are expected to establish themselves as core infrastructure for the AI era through various revenue models, including an Arbitrage model that targets power price differentials, dedicated high-density GPU Cluster zones, and Edge Gateways linked with submarine cables.

This summary was generated automatically by AI. Check the original for the author's claims and context. Copyright belongs to the original author.

Our guide explains how the AI works. Report summary errors, attribution issues, or removal requests via Contact.