AI compute · Powered by the ocean

Compute that goes where the energy is.

UK-hosted GPU capacity for AI, running on a wave-powered offshore platform. No grid connection, no fresh water, no carbon from the power supply.

WavePower generated at sea
0Grid connections needed
8–15 °CSeawater cooling, year round
UKHosted and operated

The problem

AI is running out of power, not chips.

Data-centre electricity demand is heading for around 945 TWh by 2030, roughly double today. In the UK, new multi-megawatt grid connections take years, and industrial power is among the most expensive in Europe. The sea has the energy and the cooling. It needs infrastructure built for it.

Power is the constraint

AI capacity is gated by grid connections. Offshore energy is uncontested and sits next to every coastal city.

The ocean is the chiller

Cold seawater removes heat without chillers, cooling towers or fresh water, cutting the energy overhead of every rack.

No cable to shore for power

Every marine data centre so far has relied on a power cable from land. Generating on the platform removes that constraint.

The platform

One hull. Power, cooling and compute together.

A composite platform carries an array of wave floats driving direct-drive linear generators, with a battery buffer and seawater heat exchangers on the same hull. Data travels to shore by microwave link.

The platform, wave technology and power generation are designed, built and operated by our platform partner, H2Terminals Limited. Ocean Compute leases space and power on the platform and owns and operates the compute.

Rendering of several wave-powered compute platforms at sea at dusk, each carrying large composite spheres
Concept rendering: platforms at sea. Image: H2Terminals.
  1. 01 · WAVE

    Floats rise and fall

    4 m composite floats move with the swell, hung below the deck.

  2. 02 · GENERATOR

    Direct drive

    Linear generators on deck, dry and serviceable. No turbine or gearbox.

  3. 03 · BUFFER

    DC bus and battery

    Power is conditioned and stored on the hull to smooth the sea state.

  4. 04 · COMPUTE

    Compute on board

    AI computing runs on the platform, right beside the power source.

  5. 05 · COOLING

    Heat to the sea

    Titanium plate exchangers reject heat to seawater in a closed loop.

Ocean Compute

GPU capacity for AI training, hosted in the UK.

Ocean Compute owns and operates the compute on the platform: the GPU servers, networking, storage and the link to shore. Capacity is aimed first at AI training and batch workloads, which suit an offshore site and flexible scheduling.

Hardware

Latest-generation GPU clusters

Current-generation AI accelerators, built to validated data-centre designs.

Sustainability

Clean by design

Powered by the sea and cooled by it. No grid draw, no fresh water and no carbon from the power supply.

Connectivity

Microwave to shore

Diverse point-to-point microwave links on motion-stabilised antennas, with satellite backup for management traffic.

Power & cooling

Variable waves, dependable compute.

Wave output changes with the sea. We plan the compute around dependable power, not peak output.

  • Battery ride-through smooths the swell and covers short lulls.
  • Flexible workloads: training and batch jobs checkpoint and throttle, and this is agreed with customers up front.
  • Reserved hotel load keeps control systems, cooling pumps and communications powered at all times.
  • Sized for reliability: generation, storage and backup are planned together so customers get dependable service.
Bar chart of power usage effectiveness: global average 1.56, best hyperscale 1.10, sealed marine 1.07, Ocean Compute target 1.10
Power usage effectiveness, lower is better. Target, to be proven at commissioning. Image: H2Terminals.

Security & governance

A UK company, in control of its own compute.

Our halls, our keys

Ocean Compute controls access to its data halls. Platform staff enter only for defined maintenance, escorted and logged.

Encrypted end to end

All traffic to and from the platform is encrypted with keys held only by Ocean Compute.

Independent ownership

Ocean Compute Limited is a UK company, wholly owned by its founder and separate from its platform partner.

Export-control compliance

We follow UK and US export-control rules for advanced computing, including customer checks and end-use restrictions.

About us

Founded to put AI where the power is.

Portrait of Gerry Wilkinson

Gerry Wilkinson

Founder & Director

Gerry founded Ocean Compute to bring AI compute to one of the UK's largest untapped energy resources: the sea. He is the company's director and sole shareholder. He is also co-founder and CEO of H2Terminals, the offshore energy infrastructure company that develops the wave-powered platform.

Gerry is a director of Lowtherbridge, a hydrogen fleet transport service. He has helped raise Innovate UK funding for a zero-emission internal combustion engine running in a closed-loop system, and for an AI solution that automates tax returns for freelancers.

His career began in financial software, where he raised £1m and designed the world's first operational loss database, in collaboration with Cambridge Judge Business School. He went on to advise growth companies as a Regional Director at Boardroom Advisors from 2018 to 2023, and founded Vigorate to mentor and fund young founders.

Join us

We're building a world-class team.

We are looking for experienced people who want to build a new kind of AI infrastructure. If that is you, we would like to hear from you.

Independent directors

Non-executive directors with experience in data centres, infrastructure finance or technology governance.

Data centre operations

Leadership for GPU cluster operations, uptime and on-site engineering in a marine environment.

Network & security

Design and run high-capacity links to shore, end-to-end encryption and physical security.

Contact

Talk to us about capacity, partnership or joining the team.