DEEP DIVE: Japan Built the Whole Thing Quietly

Week 44 | October 2026

On 30 January 2026, a 134-metre container ship called GENBU entered regular service on its run along Japan's Pacific coast between Kobe and Tokyo, calling at Osaka, Nagoya, Shimizu and Yokohama. Two days earlier the Ministry of Land, Infrastructure, Transport and Tourism (MLIT) had certified it as an autonomous ship. ClassNK, Japan's own classification society, had signed off two days before that.

Every one of the 53 companies in the consortium was Japanese. A Japanese foundation paid for most of the program, and a Japanese ministry put its stamp on the result.

That same month, on the defence side of the house, two smaller things happened. On 8 January the Acquisition, Technology and Logistics Agency (ATLA) signed a ¥312.4 million contract with Sumisho Aero Systems, Sumitomo's defence trading arm, to demonstrate a small multipurpose uncrewed surface vessel. The ministry's own call for proposals describes the type as a small boat built to find, track and attack enemy ships, and the tender lists the Maritime Staff Office as the delivery point. Neither ATLA nor Sumisho has said whose boat it is. Twelve days later the Japan Maritime Self-Defense Force announced it had taken delivery of a domestically developed small UUV for underwater defence. It didn't say who built it or how many it bought (Newsweek Japan counted at least six in the photo).

Three weeks. Two very different autonomy programs, and no ally named in either.

January 2026: Japan's commercial and military autonomy milestones, three weeks apart.

When I put this piece on the calendar back in July, the working line was that Japan had "built the whole thing quietly". I've spent the past week testing that line against the primary documents, and it's about half right. The commercial layer is real and certified, though more supervised than the word "autonomous" suggests. The military layer is funded and contracted, and mostly still at the demonstration stage. Plenty of its first pieces are imported. And the two layers, which ran on entirely separate tracks until very recently, have just started to touch.

That last part is the story.

I covered MEGURI2040 in Week 13, so here's the short version. The Nippon Foundation launched the program in 2020 with a goal of having half of Japan's coastal fleet operating autonomously by 2040. That's the foundation's number. The government hasn't adopted it. Stages 1 and 2 each ran to roughly ¥10 billion, and according to NYK the foundation put up more than 80% of the money. Philanthropy, in other words, paid for most of Japan's commercial autonomy.

Stage 2 produced four ships, and by 27 March 2026 all four had passed MLIT inspection and were in commercial service. First through was the ferry Olympia Dream Seto (65 metres, 500 passengers), certified on 5 December 2025 for the Shin-Okayama to Shodoshima run across the crowded Seto Inland Sea. GENBU followed in January. HOKUREN MARU No.2, an 11,413 GT RORO that mostly carries raw milk out of Hokkaido, passed on 9 February and now works the roughly 1,600 km round trip between Kushiro and Hitachi.

The last was MIKAGE, on 25 March. It's the least glamorous of the four and probably the most important. MIKAGE is a 749 GT coastal box ship, which MOL describes as the most widely used coastal container ship type in Japan, and it's a retrofit. If autonomy can be bolted onto the commonest hull in the domestic fleet, the 2040 target starts to look less like a press-release number.

Behind the ships sit two shore-based Fleet Operation Centers. The permanent one is inside Furuno's headquarters in Nishinomiya. The other is a seven-metre trailer, built mainly by Japan Radio, that can be towed somewhere safer if an earthquake or a blackout takes out the first. It's a very Japanese piece of design, and I mean that as a compliment.

Fifty-three companies sit in the DFFAS+ consortium that did the engineering, led by Japan Marine Science. Hold on to that list. It comes back later.

Now the honest bit. Every MEGURI2040 release describes these ships as achieving "Level 4-equivalent" autonomy, and MOL's own footnote admits the term is borrowed from the car industry because the International Maritime Organization hasn't settled definitions of its own. ClassNK's notation for GENBU is more candid. It's called AUTO-Nav2(All), and the "2" stands for "conditional autonomy (system use monitored by human)". The "(All)" means the system covers every phase of the voyage, berth to berth.

MLIT went a step further and wrote the human into the rules. Its 2025 amendments, in force from 1 June that year, class ships with an automated operation system as "special ships". The inspection guidance that followed on 30 June requires a defined split of roles between crew and system, a crew able to override the system at will, and a log covering at least one voyage or 48 hours, kept for six months.

I think this is the cleverest thing Japan did. It certified what the technology can actually do today (supervised autonomy on fixed coastal routes), put that into regulation, and then sent the ships to work. The IMO adopted its MASS Code at MSC 111 in May, with mandatory application not due until 2032; I covered the voluntary phase in Week 37. Japan had commercial certificates on four ships before the global rulebook was even finished.

The military side runs through a program called SHIELD: Synchronized, Hybrid, Integrated and Enhanced Littoral Defense. The Ministry of Defense's reasoning is straightforward. Cheap unmanned systems can now be bought quickly and in large numbers, and a layered mix of UAVs, USVs and UUVs gives Japan an asymmetric answer to an invasion force built around expensive crewed ships. The budget documents list ten asset types. Most of them fly, from small modular drones up to dedicated ship-attack UAVs, and the maritime pieces are one small multipurpose USV and one small multipurpose UUV. A separate effort is meant to control the whole lot from a single system.

Here I need to correct myself. In Week 21 I wrote that the FY2026 budget allocated ¥128.7 billion to SHIELD, with deployment targeted for FY2027. That figure was the request. The MoD asked for ¥128.7 billion in August 2025, the government's December draft cut it to ¥100.1 billion, and the budget the Diet passed on 7 April 2026 kept the lower number. The stated aim is for SHIELD to be built (構築) within FY2027, which ends on 31 March 2028.

I also wrote that the budget "explicitly designates domestic AI-enabled autonomous systems as a defence priority". It doesn't say that. Unmanned-asset defence capability is one of seven priority areas, worth about ¥277 billion in FY2026, and AI use is funded through its own separate line. As you'll see shortly, two of the first SHIELD purchases aren't domestic at all. 

So that's the plan on paper. The contracts tell a messier story. The largest SHIELD contract I could find went to a Japanese trading house selling an Australian drone. Three of the six contracts I've tracked went to companies in the same corporate groups as members of the consortium that built the autonomous ships. And tucked inside the FY2027 budget request is a study into putting autonomous-navigation technology on the civilian ships the military relies on for sea-lift.