AI Compute Breakthrough, Kraaken Ship Delivers 60 MW With

Optimal Transit has unveiled a second configuration of its Kraaken data-center vessel — a zero-emission ship designed to do three jobs at once.

It can host 60 megawatts of AI compute, power 32,000 homes, and produce clean water for 150,000 people, all at the same time. No fuel. No grid connection.

That positions the Kraaken as a mobile, self-sustaining AI infrastructure platform.

The targets are disaster relief zones, remote industrial sites, and off-grid deployments — the places land-based data centers simply can’t reach.

Key Takeaways

Optimal Transit published full technical details of the Kraaken’s second configuration on August 11
The vessel can host 60 megawatts of AI compute with no fuel and no grid connection
The same ship simultaneously produces clean water for up to 150,000 people
Land-based data centers on cooling spend roughly 30% to 40% of total energy consumption

Kraaken AI Compute At 60 MW Aboard A Ship With No Grid Dependency

The company published the full technical details in a Business Wire release on August 11 this year.

The Kraaken’s second configuration pushes the vessel’s AI compute hosting capacity to 60 MW.

For context, a single modern GPU cluster running large language model training at scale typically consumes between 20 MW and 30 MW of continuous power. The Kraaken’s 60 MW envelope is therefore large enough to run two full-scale frontier training clusters simultaneously, or dozens of inference clusters serving real-time AI workloads.

The vessel generates its own power through zero-emission energy systems, a design choice that eliminates the need for a grid connection or diesel fuel.

Optimal Transit has not publicly detailed the specific generation technology in this configuration, though maritime zero-emission platforms of this scale typically rely on combinations of hydrogen fuel cells, advanced nuclear microreactors, or large-scale renewable energy with onboard storage.

Beyond AI compute, the same ship simultaneously desalinates and distributes clean water for up to 150,000 people and generates electricity sufficient for 32,000 homes. That triple-output design is the defining engineering bet behind the Kraaken concept: stranded power capacity aboard a maritime vessel can serve multiple critical functions at once, making the economics of the platform more defensible than a single-purpose floating data center.

From Disaster Relief To Contested AI Infrastructure Markets

The Kraaken’s stated target markets are disaster zones and remote deployments, areas where land-based power infrastructure has failed or never existed.

That positioning matters strategically. AI compute demand is growing faster than power grid expansion in most markets, including the United States, where data center operators have spent much of this year on waiting lists for grid interconnection.

A self-powered, mobile compute platform sidesteps that bottleneck entirely.

A hurricane-damaged coastal region, a mining operation in a remote basin, or a military forward base all share the same constraint: abundant demand for AI inference and no reliable power grid. The Kraaken’s architecture is a direct answer to that constraint.

Data center operators have signed deals for dedicated nuclear capacity, onsite gas generation, and even geothermal taps to escape grid queues.

The Kraaken’s self-powered 60 MW envelope addresses the same problem from a different angle, trading permanent land rights for the ability to sail to a new port when demand shifts.

Also Read: Amazon Data Center Pollution Permit Shatters All Records With Devastating 33 Million TON CO2 Ceiling

How Floating Data Centers Work, And Why 60 MW Of AI Compute Is Meaningful

A floating data center is, at its core, a server hall built into the hull or superstructure of a ship rather than a conventional building. The engineering challenges are different from land-based construction: vibration management, salt-air corrosion, cooling in an ocean thermal environment, and power distribution across a moving hull all require non-standard solutions.

Ocean water provides a near-unlimited heat sink at stable low temperatures, a structural advantage the Kraaken’s maritime design exploits directly.

Data centers on land spend a significant share of their power budget on cooling, often 30% to 40% of total energy consumption. A ship drawing cooling water directly from the sea can dramatically reduce that overhead, improving the effective AI compute density achievable per megawatt of input power.

At 60 MW of hosted AI compute capacity, the Kraaken is large by most standards.

The typical hyperscale data center campus operates between 100 MW and 500 MW, but those are fixed, land-locked facilities requiring years of permitting and construction. A vessel that can sail to a port, connect to a local network uplink, and begin serving 60 MW of AI compute workloads within days of arrival represents a different category of infrastructure: fast, flexible, and independent of local grid conditions.

The First Configuration Set The Baseline

Optimal Transit has unveiled the second configuration of the Kraaken, increasing compute headroom beyond what the first configuration established earlier this year.

The second configuration published on August 11 this year clarifies the humanitarian capacity figures, including the 150,000-person clean water figure and the 32,000-home power figure.

When a floating data center can simultaneously serve as disaster relief infrastructure, it dramatically lowers the political and logistical friction of deploying it. A government or aid organization that needs clean water production after a storm also gains AI compute capacity as a byproduct, changing the procurement calculus entirely.

The land-based AI data center buildout has driven power costs and permitting timelines to multi-year highs across the U.S. and Europe.

Against that backdrop, a deployable 60 MW AI compute platform that generates its own power and bypasses grid interconnection queues is not a curiosity. It is an increasingly rational infrastructure choice for operators who cannot wait for conventional capacity.