Aikido Prepares Offshore Data Center Pilot

The race to power AI has sparked audacious ideas, including lofting servers into orbit for round-the-clock solar. A growing camp argues the better bet is far closer to Earth. Offshore wind developer Aikido is preparing a submerged demonstration data center off Norway, placing compute inside the buoyant pods of a floating wind turbine. If it performs as modeled, the company aims to scale the concept off the UK with a multi-megawatt turbine feeding a double-digit-megawatt data center.

Why the Ocean Beats Orbit for Compute Workloads

Space offers abundant sunlight but little else that data centers need. Heat has nowhere to go in a vacuum, so orbital facilities must carry vast radiators and reject energy slowly. Bandwidth is costly, latency is high, and maintenance borders on impossible. The ocean, by contrast, delivers dense power, efficient cooling, easy fiber backhaul, and tugboat access when hardware needs a refresh.

Latency matters. A submerged facility a few dozen kilometers offshore can link into terrestrial fiber with single-digit millisecond round trips. Even low-orbit constellations struggle to match that across busy terrestrial routes, and geostationary links are an order of magnitude slower. For AI training and inference that shuttle petabytes, the water wins.

Inside the Floating Wind-Powered Offshore Data Center

Aikido’s pilot integrates compute racks directly into the sealed, pressurized pods of a floating turbine. The initial unit targets roughly 100 kilowatts of IT load to validate power quality, thermal performance, and marine hardening. The follow-on design pairs a 15–18 megawatt turbine with a 10–12 megawatt data hall, with onboard batteries smoothing wind variability and black-starting systems after outages.

The architecture borrows lessons from subsea energy: corrosion-resistant alloys, cathodic protection, redundant seals, and anti-biofouling coatings. Fiber and power umbilicals thread through swivel joints to handle motion. Modules are designed for swap-out during scheduled weather windows or can be towed to port for deeper maintenance—no rockets required.

There is precedent. Microsoft’s Project Natick tested a sealed subsea data center off Scotland and reported remarkably low failure rates, attributing reliability in part to an inert nitrogen atmosphere and stable temperatures. Aikido’s concept aims to combine that reliability with dedicated renewable power from the same platform.

Cooling and Efficiency Advantages at Sea

Seawater is an enormous heat sink. Liquid-to-liquid loops coupled to titanium plate exchangers can shed waste heat efficiently without evaporative cooling or massive chillers. That opens a path to power usage effectiveness near 1.1, compared with 1.3–1.6 for many land-based sites. Immersion cooling for GPUs further raises density while trimming fan energy.

Environmental safeguards are non-negotiable. Designs typically isolate heat transfer within closed loops, ensuring no fluids or biocides contact the ocean. Thermal plumes disperse quickly in high-energy waters, and impact assessments—overseen by agencies and classification societies such as DNV and ABS—evaluate noise, electromagnetic emissions, and marine ecology before deployment.

Grid Bottlenecks Make Offshore Attractive

The allure is not just cooling. Power access is the constraint throttling AI growth. The International Energy Agency estimates data centers could consume more than 1,000 TWh globally in the near term, driven by AI training and inference. In several regions, including Ireland, data centers are already on track to draw a striking share of national electricity—around one-third by some projections—triggering moratoriums and grid upgrade delays.

Meanwhile, the interconnection queue for new generation and large loads has swelled. Research from Lawrence Berkeley National Laboratory shows multi-terawatt backlogs and median wait times measured in years. Co-locating compute with offshore wind creates a dedicated microgrid that bypasses crowded substations. It also bluntly addresses NIMBY concerns: no diesel noise, no visible warehouses, no heat islands near neighborhoods.

What Hyperscalers Could Run at Sea Efficiently

A 10–12 megawatt offshore module can host thousands of AI accelerators, depending on chip power budgets and cooling strategy. That is enough for sustained inference, fine-tuning, or chunked training workloads that do not require a single, monolithic cluster. Latency-sensitive services can stay onshore while power-hungry batch jobs ride the wind. Fiber backhaul ensures data remains within national jurisdictions and existing privacy regimes.

Challenges That Still Need Solving Offshore

Marine engineering is unforgiving. Saltwater eats metals, waves induce fatigue, and connectors must tolerate motion without microcracking. Insurance, classification, and cybersecurity for mixed energy-IT platforms form a new compliance stack. Permitting spans seabed leasing and navigation safety, often under entities like the UK’s Crown Estate and national energy regulators. None of this is trivial, but the oil and gas sector has decades of playbooks for operating complex assets offshore that the cloud industry can adapt.

The Bigger Picture for Offshore Compute Growth

Offshore wind is maturing fast, with capacity factors that commonly exceed onshore performance and a growing pipeline of floating projects opening deep-water markets. Pairing that resource with modular, seawater-cooled compute is a pragmatic response to the AI power crunch—scalable, serviceable, and grounded in existing supply chains.

Space will keep its allure for science and specialized comms. For AI’s immediate needs, though, the smart move may be to skip orbit entirely and let data centers float, hum, and cool just beyond the horizon.