Samsung Heavy Industries targets 2028 for floating datacenter
Class society design approvals and a year of staged shipyard deals put offshore compute on a real engineering track, though unit economics remain unproven.
Class society design approvals and a year of staged shipyard deals put offshore compute on a real engineering track, though unit economics remain unproven.
Samsung Heavy Industries has put a 2028 commercial-launch target on its floating datacenter program. The proposed design puts modular compute on a hull, draws condenser cooling from seawater, and backhauls to shore over submarine fiber. The date is the company's own. ABS and Lloyd's Register have already started signing off on the design.
The clearest signal is the American Bureau of Shipping's Approval in Principle, or AIP, for SHI's floating data center design. ABS is one of the world's main classification societies, the bodies that sign off on hull designs before a commercial ship can be built. AIP is what such a society issues when it has reviewed a concept and found it sound enough to advance to detailed design. ABS issued that certificate for SHI's design this year, putting the project on the same engineering review track as a new tanker or bulk carrier.
Lloyd's Register is moving in parallel. The classification society, SHI, and Capital Clean Energy Carriers Corp confirmed in late June that they are pushing the design forward together. Capital Clean Energy brings the marine fuel and power-generation side; SHI brings the hull and integration; Lloyd's brings the regulatory sign-off. That three-way structure is closer to a real offshore compute program than to a marketing sketch.
The Register reported on July 6 that SHI is publicly targeting a 2028 commercial launch, and Korean trade coverage carried the same target the same day. The date is SHI's own roadmap, not an independently verified delivery schedule with a named customer behind it. Until SHI names a buyer or a hull, 2028 should be read as the year the company wants the first vessel in the water, not a launch event with bookings.
The staging is what gives the timeline weight. SHI entered the floating datacenter conversation in April 2026, signed deals at the Posidonia maritime trade show in June, and laid out the lead-floating-data-center ambition in the same window. That is a deliberate cadence: concept entry, shipyard pipeline, class-society approvals, public launch year. Companies do not run that sequence for a poster.
The mechanism driving the move offshore is land-side. AI training and inference campuses are running into multi-year power-queue waits, water-use scrutiny from local regulators, cooling limits at high-density rack deployments, and permitting fights over new substations. A hull parked offshore can pull condenser cooling from seawater, anchor near a coastal substation, and skip the land permitting cycle a greenfield campus would face. Whether that is cheaper than a land build at today's power and capex prices is the open question. SHI has not published unit economics. Neither has any other shipyard in the floating-datacenter race.
Parent context sits one level up. Samsung Electronics announced a strategic partnership with OpenAI to accelerate global AI infrastructure, putting the shipbuilder cousin inside the same corporate family as a hyperscaler-aligned memory and foundry supplier. That is adjacency. It is not evidence that SHI has a launch customer. The OpenAI partnership is a Samsung Electronics story. The floating datacenter is an SHI story.
The 2028 target is a number SHI put on its own roadmap. The next question is whether ABS or Lloyd's Register moves from AIP to detailed design approval, and whether SHI names a hyperscaler, carrier, or sovereign cloud as the first hull's operator.