What's actually in the tank? The autonomous vessel fuel race

Week 28 | June 2026

Fifty-five hulls and no crew aboard any of them. Last week we asked who runs a fleet that size from shore. This week the question gets more basic: what's actually in the tank?

If we strip away the autonomy stack, the comms architecture and the shore-based operators watching six screens each, every uncrewed vessel still answers to the same constraint that's governed ships for two centuries - fuel. Diesel runs out. Batteries run flat. And unlike a crewed vessel, nothing aboard can sail into port, tie up, and wait for a bunkering barge. An autonomous platform's endurance is its major value proposition. Cut the fuel line and you've built an expensive sensor that drifts.

Four technologies are now racing to solve this from completely different directions. Nuclear promises to make the question irrelevant for a decade at a time. Hydrogen is already doing it today, just not for very long. China has built a battery-swap architecture for commercial coastal shipping that happens to look exactly like forward-deployed military logistics. And a small Massachusetts company is selling the US Special Operations Command on heat that makes electricity without making noise.

None of these are competing products in the normal sense. They're competing answers to the same procurement question, and the honest assessment is that only one of them is already proven on long-endurance autonomous missions.

Start with what exists, because everything else in this piece is further out than the coverage might suggest.

Cellula Robotics, out of Burnaby, British Columbia, announced on 27 April that its Envoy AUV had completed a fully submerged mission of 1,257 miles (2,024 kilometres) over 385 hours. More than 4,000 manoeuvres, no surfacing and hydrogen fuel cells the entire way. The Envoy is 8.5 metres long, displaces roughly 3,700 kilograms, and runs on PEM fuel cells built by Infinity Fuel Cell in Connecticut. It can anchor itself to the seabed for persistent monitoring, then release and continue.

Six weeks later, Cellula and Defence Research and Development Canada confirmed the long-endurance programme had exceeded its performance specifications. That second announcement matters more than the headline-grabbing record. A company press release tells you what a vendor wants you to believe. A DRDC validation tells you a defence customer ran the numbers and they held up. Canada is now sitting on procurement-grade evidence for a hydrogen AUV with ASW and Arctic monitoring framing attached.

The UK has its own version of this story, and it comes with a name you already know. In August last year, ACUA Ocean's Pioneer, the same platform Neil and Mike Tinmouth's team has been pushing through certification, became the first vessel to achieve Lloyd's Register compliance under the Maritime and Coastguard Agency's Workboat Code Edition 3, Annex 2. Pioneer has run a continuous 24-hour offshore mission entirely on hydrogen - zero emissions, and fully remote. On Pioneer, nine cylinders of compressed hydrogen feed a 45-kilowatt Nuvera fuel cell at 58% thermal efficiency, well ahead of the 50-52% most fuel cells manage.

Put Envoy and Pioneer next to each other and the pattern is hard to miss. One proves long-endurance hydrogen propulsion underwater. The other proves it on the surface, certified, in UK waters, under exactly the kind of shore-side operating model covered last week. Neither is a concept. Both are in the water, doing missions, and generating the kind of evidence procurement officers can actually act on.

That's the bar everything else in this piece has to clear. So far, nothing else has.

Naval nuclear power isn't new and nobody's pretending otherwise. The USS Gerald R. Ford runs two A1B reactors generating roughly 1,400 megawatts thermal between them, and this summer the Navy plans to plug the carrier into Naval Station Norfolk and export power to the base, a resilience play in case the grid goes down during a crisis. Virginia-class submarines have run on nuclear power for decades, refuelling once if at all across a multi-decade service life. None of this is the gap. Crewed naval nuclear has been solved since the 1950s, and the Ford test is really a story about energy resilience at fixed installations, not about autonomy.

The actual gap is narrower and harder. Can you shrink a reactor small enough, and automate it completely enough, to put it on a hull with nobody aboard to operate it? That's a different engineering problem from anything the US Navy has built, and right now the honest answer is: the regulatory and design groundwork is largely in place, but not one vessel has been laid down.

Walk the stack and you can see why. On the reactor side, TerraPower's MCRE, Aalo Atomics' Critical Test Reactor, Deployable Energy's Unity, AMPERA's TRISO-fuelled design and KAERI's SMART100 are all in active development, with several racing toward criticality this year. Aalo is targeting 4 July for its Aalo-X demonstration unit at Idaho National Laboratory, though the company's own roadmap now says the reactor itself won't be complete until later in 2026, even if criticality lands on schedule. On the integration side, Lloyd's Register launched a UK Maritime Nuclear Consortium in January, bringing together Rolls-Royce SMR's civil subsidiary, Babcock International, Global Nuclear Security Partners, Stephenson Harwood and NorthStandard under one roof to actually design ships around these reactors.

The Approval in Principle activity has been genuinely busy, and this is where the coverage gets sloppy enough to be worth straightening out. HD Hyundai received an AiP in June this year, but it was Lloyd's Register issuing it at Posidonia for a car carrier fitted with a molten salt reactor, not ABS at Gastech for an LNG carrier as widely reported. The 100-megawatt output figure attached to that design doesn't trace back to a primary source, so treat it as unconfirmed. Separately, ABS issued its own AiP back in September last year, for a different HD Hyundai design, an LNG carrier. Two AiPs, two classification societies, two ship types - frequently conflated into one story.

Seatransport's design has had the more interesting trajectory and the brief undersold it. Lloyd's Register issued a formal Approval in Principle in February, moving the programme from concept to genuine design acceptability, and added a 90-metre variant alongside the original 73-metre hull. Samsung Heavy Industries, meanwhile, now holds three confirmed AiPs across different reactor configurations: a compact molten salt reactor design from 2023, an LNG-fuelled molten salt reactor from September last year, and a floating SMR design from December.

Then there's the Australian angle - and it matters. AMPERA Energy, developing a TRISO-fuelled reactor with CORE POWER and TerraPower, took a US$10 million investment from Scorpio Tankers on 2 April. Buried in that announcement is the fact AMPERA has agreed to source Australian thorium for its fuel. That's a second Australian thread into the global maritime nuclear stack, sitting alongside Seatransport's UK-Australian design work. Meanwhile Ocius, builders of the Bluebottle USV, landed an A$176 million Royal Australian Navy Program of Record in March, for 40 additional vessels over five years, expanding the RAN's autonomous fleet to 55. Bluebottle runs on wave and solar power, not nuclear, and it belongs in the conversation about energy independence for a different reason: it's proof that when an Australian autonomous platform reaches procurement scale, the answer it landed on wasn't a battery at all. It was harvesting energy from the ocean it's already sitting in.

Reactor designs, classification pathways, regulatory consortiums, capital commitments. Every layer of the stack you'd need to actually build and certify a nuclear-powered autonomous vessel is now visible in the public record. What's missing is the one thing that would make any of it real: a keel laid down. Every AiP issued in 2025 and 2026 is a design endorsement. None of them is a build contract.

It's worth pausing here to consider what a genuinely different fifth pathway looks like, if only by contrast. EXMAR's ANTWERPEN, an ammonia-fuelled vessel, represents a real alternative fuel chemistry rather than an electrification or storage architecture - and it's a useful reminder that the broader shipping industry is experimenting well beyond the four threads this piece is tracking. But ammonia propulsion solves an emissions problem for crewed commercial shipping. It doesn't solve the resupply problem for an autonomous platform with nobody aboard to handle a toxic, corrosive fuel.

On 15 April, China's Ningyuan Diankun entered commercial service on the Ningbo-Zhapu coastal route. The official framing was straightforward and, on its own terms, entirely accurate: the world's largest pure electric intelligent container vessel, 9,600 deadweight tonnes, ten containerised swappable battery packs holding roughly 19,600 kilowatt-hours, dual 875-kilowatt motors, CCS-certified autonomous navigation, 1,462 tonnes of annual carbon reduction. A genuine green shipping milestone.

Read the fine print and a second story sits underneath the first. The propulsion system was developed by CSSC's 704th Research Institute, a defence research institute, not a commercial engineering shop, and the same institute that developed propulsion for the Zhuhai Yun, China's unmanned system mother vessel built explicitly to support autonomous naval platforms. Nobody concealed this - It's in the official specifications - It's just that the shipping press covered the green credentials and nobody connected it to anything else.

This is the same architecture OTI has tracked since Week 21, just with an energy logistics layer now attached. The forward-deployment problem for any autonomous naval platform is fundamentally a question of how you resupply something with no crew aboard to manage refuelling, operating somewhere a tanker can't easily reach. Containerised, swappable battery packs, designed for crane-assisted exchange and already proven in commercial coastal service, are a plausible answer to exactly that problem. Whether the Ningyuan Diankun's specific architecture ever shows up on a naval hull is beside the point. The development work, de-risked and deployed commercially, is already done.

That's civil-military fusion as a design principle rather than a conspiracy. The commercial programme is the development bed. No separate military programme needs to exist in the public record, because the public record already contains everything a defence planner would need.

The fourth pathway is the quietest, literally. Thermophotovoltaic generation, commercialised by Massachusetts-based Mesodyne under the LightCell name, converts heat directly into electricity using a solid-state process with no moving parts. No turbine, no compressor, no acoustic signature beyond whatever heat source is driving it.

In May, USSOCOM awarded Mesodyne a US$4.25 million contract for Tactical Expeditionary Silent Power Generation, the largest defence contract LightCell has secured to date and a clear signal of where the near-term customer base actually sits. The technology, developed out of MIT and Argonne National Laboratory under physicist Veronika Stelmakh, is a credible answer for small platforms that need power without giving away their position acoustically.

What it isn't, yet, is proven at sea. There's no confirmed AUV or USV deployment in the public record. TPV is the pathway furthest from the water of the four covered here, genuinely promising for expeditionary and special operations applications, but a maritime customer hasn't yet surfaced. 

Run the four pathways side by side and our institutional gap thesis holds again, in a slightly different shape than usual.

Hydrogen is the only one of the four with vessels actually doing missions today: Cellula underwater, and ACUA on the surface. Both are certified, both are validated by defense or classification bodies that don't hand out credibility for free. Nuclear has built an almost complete regulatory and design stack, reactors, integrators, classification societies, capital, but not one keel laid. China's battery-swap architecture is sitting in plain sight inside a green shipping success story, doing the development work for a forward logistics problem nobody's officially calling a forward logistics problem. TPV is real money from a real defense customer chasing a genuinely useful capability that hasn't found a hull yet.

Technology isn't the problem. The hydrogen platforms prove the engineering works. The nuclear AiPs prove the regulatory pathway exists. What's missing across three of the four pathways is the same thing that's been missing in every Deep Dive this publication has run: someone willing to commit capital to a hull before the doctrine catches up, rather than waiting for the doctrine to justify the hull.

Defense planners watching this space don't need another reactor announcement. They need to ask which of the four pathways their own platforms can actually wait for, and which ones they need running now. For long-endurance autonomous missions today, that question already has an answer. It just isn't nuclear.

Ocean Rainforest's quiet acquisition of Alamarsa, eighteen months ago now, was a clearer signal than anyone treated it as at the time: marine biomaterials supply chain consolidation is already underway. The procurement side hasn't caught up. Next week we look at why, and at an uncomfortable detail buried in the defence sustainability story: Saab is still the only major defence company with SBTi-validated targets, and Lockheed Martin's shareholders are the ones pushing it to follow. 

Since you have been, thanks for reading.

Cheers,

Mick

Ocean Tech Intelligence provides informational analysis only. Nothing in this publication constitutes financial, investment, legal, or strategic advice. Readers act on this content at their own risk. For full details see our Disclaimer.