The Daily Byte
Breaking
Welcome to The Daily Byte — real stories from across Pakistan.
Saturday, 21 November 2026 · Pakistan

The Daily Byte

Stories from Pakistan that deserve to be heard
Minor World

Drogue floating AI data centres run on ocean heat

By The Daily Byte ·14 Nov 2026 ·4 min read
An engineer operates a large 3D printer in Drogue's Washington, D.C. laboratory.

A Washington startup has raised $20 million to put artificial intelligence data centres on ships and power them with heat from the ocean. Drogue, founded only about six months ago in Washington, D.C., plans to have a multi-megawatt prototype at sea in early 2028, running servers far from any onshore power grid.

The funding was disclosed exclusively to TechCrunch on 13 November 2026, and confirmed in a second report by GazeWire. Investors in the round include XYZ Venture Capital, Toyota Ventures, Tamarack Global, Dream, FDVC, Garuda Ventures, Flexcap, Boost VC and Slow Ventures. Drogue has not disclosed a valuation.

The idea is a response to a timing problem. Building a data centre on land means finding a site, securing planning permission and waiting for a grid connection, a process that can take eight to ten years. Demand for computing power to train and run AI systems has grown far faster than that. Founder and chief executive Sam Sipe, a former US Navy pilot who flew E-2D Hawkeye aircraft from aircraft carriers, put it bluntly: AI demand has increased tenfold in just three years, and the old timelines are, in his words, completely incompatible with it.

How ocean heat becomes electricity

Drogue's system uses ocean thermal energy conversion, known as OTEC, a technique first developed in the 1950s and tested in the 1980s before being set aside. It works like a heat pump run in reverse. Instead of using electricity to move heat, it uses a difference in temperature to generate electricity.

Warm water from the ocean surface is first used to cool the data centre hardware. The now-warmer water is then used to boil ammonia, a chemical that turns to gas at minus 28 degrees Fahrenheit. The ammonia gas drives a turbine that generates electricity, before passing through a condenser that turns it back into liquid. Cold water drawn from about 1,000 metres deep, at around 39 degrees Fahrenheit, cools the condenser, and the ammonia returns to the start of the sealed loop to begin again.

Surface waters are heated by the sun, while deep water stays much colder, so the temperature gap the system depends on is constantly renewed. As the planet warms, the oceans have been absorbing an additional 5.8 million terawatt-hours of energy every year, against about 180,000 terawatt-hours used by all of human civilisation in a year. Sipe's argument is that taking a small share of that heat would not disturb the balance.

Why put the servers at sea

There also had to be customers who could use the power while still at sea. Sending electricity to shore would need long, expensive cables, and manufacturing goods on board would need equally expensive shipping. Data centres fit the model because the servers are compact compared with other industrial equipment, and sending data off the platform is cheap.

The first prototype will be installed on a traditional oceangoing ship, which Drogue is in the process of buying. Later versions would be built on floating structures closer in design to offshore oil platforms. Each vessel would stay at sea for about a decade, according to Sipe.

What is proven and what is not

Drogue already holds an approval in principle, a preliminary sign-off, from the American Bureau of Shipping, which sets and checks safety standards for ships. That is a real milestone, but it is only the start. The company has not named any data centre customers, shared pricing, or published measured figures for how much power a vessel would produce once pumping and cooling are taken into account.

The history is also a warning. The basic idea was tested in the 1980s and abandoned, and Sipe himself says the earlier efforts failed because nobody could raise enough capital to build at a scale that worked. His case is that manufacturing techniques for pipes and heat exchangers have improved enough since then to change the economics, and that he spent time interviewing people who worked on those early projects to learn what went wrong.

For now, Drogue is a plan with funding and a regulator's early blessing, not a working power station. If the prototype due in early 2028 produces net power and runs AI workloads reliably at sea, it would offer data centre developers a way around grid queues that now stretch for years. If it does not, it will join a long list of ocean-energy ideas that looked better on paper than in the water.