Disaggregated Offshore Compute Outpaces Integrated Providers

The key battleground is not whether offshore compute works, it is who can get usable megawatts in the water fastest with the fewest custom dependencies. Samsung Heavy, Mousterian, MOL, Hitachi, HiCloud, and Nautilus all break the problem into familiar pieces, ship hulls, grid or wind power, liquid cooling, and standard server gear, which can shorten procurement, permitting, and maintenance compared with Panthalassa building energy, platform, and compute as one tightly coupled system.

• Samsung Heavy already has ABS approval in principle for a 50 MW floating data center design, and Mousterian gives it a U.S. path to market. That means a hyperscaler could buy capacity from a shipbuilder and a local development partner, instead of waiting for one company to prove a whole new offshore stack.
  1 eagle 2 prnewswire
• MOL and Hitachi are pushing an even simpler model, retrofit an existing vessel into a floating data center. HiCloud shows the same modular logic in another form, pairing subsea compute with offshore wind and seawater cooling, reaching 24 MW in Shanghai with a stated path to 500 MW.
  3 hitachi 4 datacenterdynamics
• Microsoft validated the reliability side of ocean compute with Project Natick, then stopped short of commercialization. That leaves buyers more likely to favor designs that keep service, power, and server components closer to standard data center workflows, which is where Nautilus and other disaggregated models are strongest.
  5 datacenterdynamics 6 sacra

The next phase of offshore compute will likely look less like one moonshot platform and more like a supply chain, with shipbuilders, power partners, cooling specialists, and GPU vendors each handling one layer. If that pattern holds, Panthalassa's advantage will depend on proving that integration delivers meaningfully lower cost per deployed megawatt, not just a cleaner architecture.