Infrastructure Architecture

Floating vs. Subsea Data Centers: Why Seabase Chose Subsea

Floating and subsea data centers both use water to address land and cooling constraints. They create very different operating systems. Seabase selected a modular subsea architecture because it better supports stable high-density compute near major coastal markets with a smaller surface footprint.

· 11 min read

← All Articles

The ocean-compute category includes several distinct infrastructure models.

Some companies place data centers on barges or floating platforms. Others place sealed compute vessels on the seabed. Additional proposals combine floating power generation with onboard computing or locate vessels farther offshore where land and grid constraints are less significant.

These approaches are often grouped together as offshore or marine data centers. Technically and operationally, they are very different.

Floating infrastructure remains a surface marine asset. It moves with waves, depends on moorings or stationkeeping, occupies visible water space, and remains directly exposed to wind, salt spray, humidity, storms, vessel traffic, and surface operating conditions.

Subsea infrastructure introduces pressure, retrieval, connector, corrosion, and intervention challenges. In return, it can operate below the most dynamic surface environment, use depth as a thermal and site-selection variable, and place large amounts of infrastructure near coastal demand without creating another visible industrial campus.

Seabase evaluated both concepts.

For the markets and operating model Seabase is pursuing, modular subsea infrastructure provides the stronger foundation. For a broader classification of approaches in this category, see the ocean compute company landscape. For how Seabase separates long-life infrastructure from retrievable compute modules, see Beyond Project Natick.

Floating Data Centers Are a Legitimate Model

Floating data centers should not be dismissed.

Nautilus Data Technologies operates a floating data center in Stockton, California, using a barge-based structure and water-cooled heat rejection. The company lists 6.5 megawatts of critical IT capacity, rack densities up to 55 kilowatts, and a stated PUE of 1.15 at the site. (Nautilus Data Technologies)

That model demonstrates several real advantages.

A floating facility can:

These advantages can be compelling in protected rivers, ports, and industrial waterfronts.

The issue is not whether floating data centers can work.

The issue is whether they provide the best architecture for a globally distributed network of high-density AI capacity positioned near major coastal population centers.

For Seabase, they do not.

A Floating Platform Remains in the Surface Environment

The ocean surface is the most dynamic part of the marine environment.

Floating structures respond to waves, wind, currents, changing water levels, vessel wakes, and storms. Their mooring systems, hulls, risers, power cables, and communications interfaces must accommodate repeated movement.

Offshore engineering standards therefore require analysis of motion, acceleration, structural stress, fatigue, corrosion, mooring behavior, and extreme environmental loads. The International Maritime Organization specifically identifies unit motion, wind moments, acceleration, and fatigue as design considerations for mobile offshore structures. (IMO)

Research into floating offshore systems also documents the coupled effects of wave loading, corrosion, material fatigue, and motion on moorings, cables, risers, and supporting structures. (MDPI)

These are solvable engineering problems. Offshore industries have managed floating structures for decades.

They are still additional problems that a fixed subsea platform does not experience in the same way.

A seabed-mounted system does not experience the same continuous six-degree-of-freedom motion as a floating platform, although it must still be designed for currents, vibration, seabed conditions, cable dynamics, and environmental loading. Its normal operating position can be fundamentally more stable.

For high-density computing, stability matters.

Servers, liquid-cooling systems, electrical equipment, optical systems, and rack-scale networks are designed around controlled operating environments. Motion can be engineered around, but it adds requirements that do not improve computing performance.

Motion Affects More Than the Hull

It would be incomplete to evaluate platform motion only as a structural concern.

Motion can affect:

A floating data center can use dampening, reinforced racks, flexible connections, isolation systems, and carefully selected operating locations.

Every added mitigation becomes part of the infrastructure cost, qualification program, and maintenance system.

In a protected river or sheltered harbor, the motion envelope may be manageable. Farther offshore, it becomes more demanding.

Seabase's objective is not to build a data center that tolerates vessel movement. It is to build a stable modular platform that uses the marine environment without inheriting unnecessary surface-vessel behavior.

Surface Access Is an Advantage and a Constraint

The strongest argument for floating infrastructure is access.

Technicians can enter the facility, replace components, inspect racks, and perform routine work without retrieving a submerged module. This makes a floating facility more similar to a conventional terrestrial data center.

Seabase does not deny that advantage.

The tradeoff is that preserving continuous human access keeps the facility in the same environment that creates many of the other constraints:

An occupied facility must support people as well as computers.

That means space for circulation, emergency access, lighting, life safety, maintenance, environmental conditioning, and working areas. A subsea compute module does not need to function as an occupied building while operating.

Seabase trades permanent human access for planned module retrieval.

The company's architecture treats retrieval as a normal operating procedure, allowing individual compute modules to be brought to a controlled surface facility for maintenance or hardware replacement while the shared infrastructure remains installed. That service model is developed further in Beyond Project Natick.

That is a different service model, not a claim that service is unnecessary.

Saltwater Exposure Does Not Disappear Above the Waterline

A floating platform avoids external hydrostatic pressure on its server spaces, but it does not avoid the marine environment.

Surface marine assets experience salt-laden air, humidity, spray, condensation, corrosion, solar heating, and weather variation. Those conditions must be managed through enclosure design, coatings, environmental control, filtration, inspection, and maintenance.

The server halls themselves can be sealed and conditioned. The wider platform, cooling equipment, hull, moorings, external cables, deck systems, and support equipment remain exposed.

A subsea module also requires corrosion protection and materials qualification. Its internal environment can, however, be sealed and controlled without supporting occupied access during operation.

Microsoft's Project Natick reported that servers inside its sealed nitrogen environment experienced approximately one-eighth the failure rate of comparable land-based servers during the experiment. Microsoft attributed the result partly to the controlled atmosphere and reduced human interaction. (Microsoft)

That result does not prove every subsea system will achieve the same reliability. It demonstrates that removing humidity, oxygen, and routine human disturbance can create a favorable internal environment for computing equipment.

Floating Infrastructure Remains Visible Infrastructure

One of Seabase's primary objectives is to place significant compute capacity close to dense coastal markets without building another large visible industrial campus.

A floating data center reduces land use, but it remains visible.

A barge, ship, or offshore platform occupies the surface. Depending on its scale and location, it may affect waterfront views, navigational routes, port operations, harbor planning, recreational use, tourism, community perception, surface noise, and lighting.

This may be acceptable within an existing industrial port. It becomes more difficult near residential waterfronts, tourism-driven markets, constrained harbors, or places where surface water is already heavily used.

Subsea infrastructure changes the public interface. The primary compute equipment is below the surface. Shore infrastructure, power connections, vessels, landing systems, and maintenance operations remain visible, but the operating compute platform does not occupy the skyline or require a permanently visible floating industrial structure.

That distinction directly supports Seabase's community-footprint strategy.

Surface Infrastructure Is Exposed Infrastructure

Data centers are increasingly recognized as strategic physical infrastructure. Their economic importance makes physical resilience relevant to operators, customers, and host jurisdictions.

In March 2026, drone strikes damaged Amazon Web Services facilities in the United Arab Emirates and Bahrain. AWS reported structural damage, interruptions to power delivery, and additional water damage caused by fire-suppression activity. Customers experienced degraded availability, and recovery was expected to take an extended period because the damage affected the physical facilities rather than only their software systems. (Reuters, Investing.com)

The lesson is not limited to one region or one conflict. A modern data center may contain billions of dollars of equipment and multiple layers of redundancy, yet still occupy a known surface location exposed to weather, accidents, sabotage, and aerial or surface attack.

Floating infrastructure preserves many of those same characteristics. A barge or offshore platform may be easier to isolate from casual public access than a terrestrial campus, but it remains visible above the water, accessible by surface vessels, dependent on mooring or stationkeeping systems, and concentrated within one identifiable marine asset.

Subsea deployment does not eliminate physical risk. It reduces ordinary surface visibility and generally requires more specialized access, equipment, and site knowledge to reach the primary compute assets. A subsea deployment introduces reduced visibility, depth, restricted marine access, and specialized intervention requirements between an external threat and the compute modules.

Shore-side power, fiber, landing facilities, and operating information remain physical assets and must be protected appropriately. The relevant distinction is not immunity. It is the amount of effort, access, equipment, and knowledge required to reach the primary compute asset.

Resilience Through Placement and Distribution

Seabase should not depend on reduced visibility alone.

Consequence reduction comes from architecture and geography: modular failure domains, distributed regional deployments, protected power and fiber routes, defined retrieval procedures, and planned workload recovery and regional placement through Nori. Seabase also evaluates locations near established port, energy, communications, and marine operating environments where access, monitoring, maintenance, and infrastructure protection can be coordinated with qualified local partners.

Integration with controlled port, utility, communications, and industrial operating environments can support those procedures without treating unrelated public systems as the protective envelope for private compute assets.

Modularity Limits the Consequence of Damage

Physical resilience also depends on architecture.

A billion-dollar surface campus or large floating facility can concentrate substantial computing capacity within one identifiable asset. Damage to power, cooling, or the structure can affect a large portion of that capacity simultaneously.

Seabase's modular approach creates smaller infrastructure failure domains. Individual compute modules are intended to be independently monitored, isolated, retrieved, and replaced. Capacity can also be distributed across multiple foundations and regions rather than concentrated entirely within one building, barge, or vessel.

This does not eliminate correlated risks. Several modules may still share shore power, fiber, or regional infrastructure. Those dependencies must be addressed through project-specific redundancy and recovery planning. Modularity nevertheless reduces the need to treat the entire deployment as one indivisible target.

Surface data centers, floating platforms, and subsea systems all require physical security. None is immune from deliberate attack. The difference is exposure. Floating and terrestrial facilities remain visible, accessible surface assets. Subsea infrastructure places depth, restricted marine access, specialized intervention requirements, and modular failure boundaries between an external threat and the primary compute equipment. For Seabase, that reduced exposure is another reason to build below the surface rather than merely moving a conventional data center onto it.

Depth Creates a Thermal Option Floating Systems Cannot Use

A floating platform generally rejects heat using water available near the surface.

Surface water temperature can change with:

A subsea system can use depth as part of site selection.

Depending on the location, moving deeper may provide colder or more thermally stable water while the infrastructure remains only a limited horizontal distance from shore.

This is important because horizontal distance and depth are separate variables.

A project does not necessarily need to move far offshore to reach a different thermal environment. It may remain close to metropolitan demand and coastal fiber while moving downward into more favorable conditions.

Depth is not automatically beneficial.

Greater depth increases pressure, structural requirements, cable considerations, deployment complexity, and retrieval requirements. The appropriate operating depth must be selected through site-specific engineering.

The advantage is optionality.

Floating infrastructure normally operates within the surface environment and requires additional intake infrastructure to access deeper water. Subsea infrastructure can place the heat-rejection system directly within a three-dimensional site envelope.

Floating Far Offshore Weakens the Network Case

Some floating-data-center proposals focus on placing large amounts of compute far offshore, potentially alongside wind, nuclear, gas, or other dedicated generation.

This may solve one part of the infrastructure problem: power.

It does not automatically solve networking.

A remote floating cluster still needs a high-capacity connection to users, data sources, cloud regions, and other compute infrastructure.

That requires either:

Power generation aboard a ship or platform does not make the resulting compute geographically close to customers.

If the platform depends primarily on satellites, its useful capacity is bounded by the latency, availability, and bandwidth of the satellite network. If it depends on submarine fiber, the project must still develop and protect that fiber system.

A remote cluster may generate inexpensive computation while creating a more expensive path to deliver it.

The cheapest megawatt is not necessarily the cheapest useful AI output. The latency tax remains a geographic cost, regardless of how the servers are cooled.

Seabase Is Optimizing for Metropolitan Adjacency

Seabase is not attempting to place compute in the middle of the ocean.

The company is developing infrastructure for nearshore deployment around major coastal markets.

The target geography combines:

A floating platform and a subsea platform can both connect to these systems.

The subsea platform can do so without occupying valuable surface space or remaining continuously visible.

That creates a distinctive site-selection advantage.

Seabase can evaluate marine areas near a market even when conventional land and surface-water development are constrained. For the demand-side geography of that strategy, see coastal metros and AI infrastructure.

Scaling a Floating Platform Means Scaling the Vessel System

Floating systems can be modular, but every floating increment requires some combination of:

A larger floating data center may gain efficiencies from a larger shared hull. It also creates a larger individual asset and a more concentrated operational failure domain.

Seabase separates long-life shared infrastructure from smaller retrievable compute modules. That allows capacity to be installed in stages without requiring every new block of compute to bring an entirely new vessel and mooring system.

The foundation can remain in place while compute modules are added, removed, serviced, or refreshed.

This is especially important for AI hardware because the compute lifecycle is much shorter than the expected life of marine infrastructure.

Subsea Infrastructure Has Real Disadvantages

Seabase should state these clearly.

Subsea systems must address:

These are not minor issues.

A poorly designed subsea platform could become difficult to service, expensive to retrieve, or locked to one generation of hardware.

That is why Seabase's modular architecture matters.

The company is not proposing permanently sealed infrastructure that remains inaccessible for its entire service life. It is designing around retrievable compute modules, shared long-life infrastructure, staged deployment, failure isolation, and planned intervention.

The goal is not to avoid marine complexity.

It is to select the marine complexity that produces the strongest long-term compute platform.

When Floating May Be the Better Choice

Floating infrastructure may be preferable when:

Those are legitimate operating cases.

Seabase is solving a different problem.

Why Seabase Chose Subsea

Seabase selected modular subsea infrastructure because it better aligns with the company's primary objectives:

  1. Place compute close to major coastal demand. Nearshore deployment can position infrastructure within a limited distance of millions of users and major enterprise markets.
  2. Reduce the visible and terrestrial footprint. The primary compute infrastructure operates below the surface rather than becoming another building, barge, or platform in the public environment.
  3. Use depth as a site and thermal variable. The platform can evaluate colder or more stable water without moving proportionally farther from shore.
  4. Avoid continuous surface motion. A fixed seabed platform does not inherit the same wave-driven operating behavior as a floating vessel.
  5. Remain adjacent to fiber and port ecosystems. Nearshore infrastructure can connect with cable routes, internet exchanges, industrial power, and marine operations.
  6. Separate long-life infrastructure from changing hardware. Shared infrastructure can support multiple generations of independently retrievable compute modules.
  7. Expand capacity in stages. Modules can be added as power and customer demand develop.
  8. Build a persistent geographic platform. The value of a qualified position near demand, power, fiber, and marine access can compound across many hardware generations.
  9. Reduce exposure to surface threats. Subsea deployment does not eliminate physical-security risk, but it makes the primary compute equipment less visible and generally more difficult to access through conventional surface methods. Distributed modules and regional infrastructure can further limit the consequences of damage to any single asset.

Different Tools for Different Markets

Floating data centers and subsea data centers should not be treated as interchangeable versions of the same product.

Floating systems preserve surface access and can work well in protected industrial waterways.

Subsea systems provide a stable, low-visibility operating environment with access to depth and a different relationship with coastal geography.

Seabase selected subsea infrastructure because its goal is not merely to cool servers with water.

The goal is to build a durable regional AI platform near the people, enterprises, networks, and energy systems that will use it.

For that objective, being below the surface is not an incidental engineering choice.

It is the geographic and operating advantage.

For a neutral map of companies and approaches across this category, see the ocean compute company landscape. To discuss modular subsea infrastructure, reserved capacity, or regional deployment requirements, contact Seabase.

Next step

Discuss modular subsea infrastructure

Contact Seabase