Drogue, a Washington startup founded six months ago, told TechCrunch this week that it has raised $20 million to build data centers that run at sea and generate their own electricity from the ocean's temperature difference. Investors include XYZ Venture Capital, Toyota Ventures, Tamarack Global, Dream, FDVC, Garuda Ventures, Flexcap, Boost VC and Slow Ventures.
Founder and chief executive Sam Sipe, a former US Navy E-2D Hawkeye pilot, said a multi-megawatt prototype will be installed on a conventional oceangoing ship and sent out in early 2028. Later versions would sit on floating structures closer to offshore oil platforms than to ships.
The pitch is schedule, not novelty
Sipe's argument is about how long it takes to get power. "The data center industry is a real estate game, and so they're playing on 8- to 10-year timelines, including interoperability with utilities and [independent system operators]," he said. "AI demand has increased 10x in just the last three years. It's completely incompatible."
A vessel that arrives with its own generation skips the interconnection queue. Each ship would stay at sea for about a decade, Sipe said, with servers swapped and data shipped over cheap satellite or microwave links rather than through cables to shore. That last detail is why the workload fits: data is light, and manufacturing or mining at sea would require shipping materials both ways.
How the loop works
The technology is ocean thermal energy conversion, an idea first pursued in the 1950s and tested in the 1980s before being dropped. It runs a heat engine on the difference between warm surface water and water pulled from about 1,000 meters down, which averages roughly 4°C.
In Drogue's design the surface water cools the servers first, which warms it further. That water then boils ammonia, which turns to gas at minus 28°F. The gas drives a turbine, passes into a condenser, and is chilled back to liquid by deep water running over the condenser's fins. The ammonia circulates in a sealed loop and the warmed seawater is returned to the ocean.
The 1980s attempts stalled on capital rather than physics. Sipe said he tracked down engineers from those programs and asked what went wrong. "Everyone was like, dude, this stuff works. It's just a matter of being able to raise enough capital to build something big enough, because there's a sweet spot," he said. Better pipe and heat-exchanger manufacturing is the other change since then.
The energy budget is not the constraint. As the planet warms, the oceans absorb an additional 5.8 million terawatt-hours a year, against roughly 180,000 terawatt-hours consumed by all of human civilization. Sipe's point is that pulling heat out of the surface layer cannot close a temperature gap that size.
Permits, ships and the comparison that will follow
Drogue has "approval in principle" from the American Bureau of Shipping, a preliminary sign-off on the design, and is now buying its first ship. That purchase is the near-term milestone: without a hull there is nothing to test at sea, and a marine engine room is a harder place to run servers than a warehouse in Virginia.
The obvious comparison is orbital data centers, which have attracted far more attention and far more capital. Both ideas chase cheap power and put the load next to it. The difference is transport. A ship can leave a yard; a rocket has to reach orbit. That makes Drogue cheaper to test and much cheaper to abandon if the economics fail.
It also inherits problems the warehouse version does not have. Salt water, biofouling, storms and maintenance at sea all raise costs, and nothing in the plan addresses what happens when a vessel needs a technician who is not aboard. The 2028 prototype is the first chance to price those risks.
Data center builders are already paying premiums for speed elsewhere. Crusoe sold an unfinished Wyoming campus to Google, and Anthropic signed a $34 billion power deal for capacity that already exists. Drogue is betting some buyers will pay to skip the grid entirely rather than wait in it.
Source: TechCrunch