A wave-energy startup wants to put AI's data centers out at sea

AI's data centers are running out of room, not of power, but of place. Land is fought over, grid interconnects are queued for years, water for cooling is contested in drought-prone regions, and the social license to build a 1-gigawatt campus next to a town has become the new bottleneck. The response is a small industry of escape hatches: orbital compute, underground bunkers, small modular nuclear, and now, floating.

A small Edinburgh wave-energy company, Mocean Energy, is pitching one of the more literal versions. Its proposed "Blue Core" data center would sit offshore, draw power from the company's existing "Blue Star" wave-energy generator, and cool itself with seawater pulled through a closed loop. Managing director Cameron McNatt frames the pitch as removing three constraints at once: power, land, and community opposition. The company is raising a £5M (~$6.3M) pre-Series A round to fund the work alongside its existing offshore-power business, a small check by data-center standards.

The unit-level math is the part to watch. Each Blue Core module is rated at 500 kW to 1 MW, with several units combined into a "farm" linked by low-Earth orbit satellite connections. For comparison, a single hyperscale AI campus today typically runs in the hundreds of megawatts to gigawatts, and AI data centers are projected to consume 1,200 TWh a year by 2035, with AI-optimized compute demand set to quadruple. McNatt claims the closed-loop ocean cooling would consume roughly 5% of the energy budget, against an onshore cooling overhead of 15% to 35%. That is a company-stated design target, not field data, and one that has to survive salt, biofouling, and seasonal temperature swings.

Wave energy converts ocean surface motion into electricity; the technology is related to but distinct from tidal power, which exploits predictable gravitational cycles. Blue Star, the underlying wave-power platform, has finished 18 months of full-scale sea trials with offshore energy majors (the largest global oil and gas companies), and the company forecasts its first commercial Blue Star sales in 2028. A first Blue Core demonstrator is targeted to follow. The "social license" claim is doing real work here: by moving offshore, the project shifts from a community-permitting fight to a maritime-permitting fight, which is a different problem, not a solved one.

The gap between pitch and product is what makes this a useful read for any other floating, orbital, or underground data center announcement. Three tests have to pass before the category becomes a real option rather than a rendering. First, can a 500 kW–1 MW unit actually carry a modern AI training or inference workload, or does the architecture have to be rebuilt around a modularity hyperscalers have not yet adopted? Second, is a 5% ocean-cooling overhead achievable across sites and seasons, or is it a best-case number? Third, can a pre-Series A startup reach commercial unit economics before onshore hyperscale buildout absorbs the available demand, talent, and grid?

Mocean's announcement is, for now, a proposal and a lens. The interesting story is not whether this particular company wins; it is whether the constraint map it points to keeps pushing serious capital offshore in the next 24 months. If it does, expect more pitches in this shape. If it does not, this reads in five years as a clever early sketch of a problem that got solved on land.