DMM Quit Robots. Now It’s Putting ¥5 Billion Into a Return.

DMM is buying another ticket to a business it already left. On September 9, 2026, Algomatic Dynamics announced ¥5 billion in financing from the Japanese group for robotics and physical AI. Chairman Keishi Kameyama supplied an unusually useful detail: DMM’s previous robot venture had lasted roughly six years before it withdrew. The new money arrives with a memory.

That is a fitting introduction to DMM.com Group, whose businesses stretch from anime streaming and online games to currency trading, English lessons and an Okinawa aquarium. The company’s product list can read like several browser windows accidentally pasted together. Yet the robotics return poses a coherent question: when does a failed idea become worth another attempt?

Kameyama’s answer in the announcement was specific. Robotics technology had advanced, and meeting the new venture’s leader, Yuki Nanri, convinced him to try again. That explains the decision; it does not establish that the new business will work. For a group accustomed to opening doors, the interesting discipline is deciding which ones merit reopening.

A department store with several kinds of checkout

Kameyama’s commercial roots run through video rental and adult entertainment. DMM’s online video and shopping operations began in 1998; the company now called DMM.com LLC was incorporated in November 1999. This history matters because its expertise grew around getting people to discover content, pay for it and return. The catalogue could change while parts of the machinery remained useful.

Today, the scale is substantial. DMM reports ¥428.7 billion in group revenue for the year ended February 2026. Its 60.12 million DMM.com service members measure registrations, rather than paid streaming subscriptions or monthly activity. Those distinctions are easy to lose when a company sells so many things. A member is an opportunity to serve someone, not proof that every venture has found a customer.

The most legible part of the collection is entertainment. DMM TV supplies anime and other video; DMM Books sells reading; DMM GAMES distributes and publishes games. Merchandise gives affection for a character somewhere else to go. These businesses can meet the same fan at different moments, with different products. The appeal of the arrangement is easier to understand from a fan’s week than from an organization chart.

DMM Premium’s advertised web price is ¥550 a month, including tax; registration through Apple’s or Google’s app store costs ¥650. Eligible streaming comes with the subscription, while some content requires separate payment. The membership also promotes benefits elsewhere in DMM and bundles with DAZN and pixiv. For an anime viewer, the practical question is whether the desired shows and benefits justify another monthly bill.

Other customers arrive with no interest in anime. DMM Eikaiwa connects learners with tutors for one-to-one English lessons online. DMM.make accepts orders for 3D printing, letting a designer obtain a physical part without buying a printer. DMM Online Clinic provides a route to remote consultations through medical partners. Each reduces a different inconvenience: scheduling a lesson, accessing manufacturing equipment or arranging an appointment.

The business model changes with the purchase. Digital subscriptions collect recurring payments. Books and merchandise depend on transactions. Printing sells production services. DMM FX provides margin trading through the securities business, with economics tied to trading activity and spreads. An aquarium sells admission to a physical place. Calling all of this an internet platform is convenient, but it conceals the work.

DMM consequently has several sets of competitors. Viewers can choose U-NEXT, Netflix or Prime Video; readers can buy through Kindle or Kobo; English learners can consider RareJob or NativeCamp. These are substitutes within particular markets, not a single league table. DMM’s distinguishing feature is the breadth of the operating group, especially its ability to carry entertainment into several formats. That breadth does not guarantee the best catalogue, lesson or price.

The promotion that customers understood too well

In March 2021, DMM renamed its ebook service DMM Books and offered first-time buyers 70% off as many as 100 books. Customers spotted the possibilities. Entire manga series suddenly looked affordable. Expensive reference books became attractive too: a later sales ranking featured pose collections, illustration guides and other books useful to artists. The promotion rewarded people who filled the basket.

The eventual reported loss was ¥6 billion. Coupon distribution ended early, in April, as demand exceeded expectations. The business lesson is almost embarrassingly concrete: limiting the number of items does not necessarily limit the subsidy to an affordable amount. A hundred inexpensive books and a hundred expensive ones are very different promises. Readers had every reason to notice the difference.

The technical account is just as revealing. DMM engineer Hiroshi Ono described older book-delivery servers coming under pressure first. Buyers were collecting backlists, shifting the load away from the usual pattern around newer releases. Later, purchasing 100 books could trigger 100 requests for user information. A generous basket was also a surprisingly demanding basket.

Ending the offer early brought another rush. In Ono’s account, engineers limited requests to protect the shared membership infrastructure, then pursued changes including cloud-based book delivery, additional monitoring and revised connection settings. The incident gave operational substance to an otherwise vague word: scale. The number of shoppers mattered, but so did what each shopper did.

A smaller retailer can borrow the diagnosis without borrowing the loss. Put a currency limit on the discount, calculate the largest permitted order and test that order through checkout. Test the announcement that the sale is ending, too. Scarcity is a marketing tool with a workload attached. The engineering review and the campaign budget belong in the same meeting.

Failure, with a balance sheet attached

Kameyama has been unusually plain about the survival rate. In a 2022 conversation at IVS in Okinawa, he estimated that around ten ventures begin and roughly two remain. He also described businesses stopped after six months and others allowed to lose money for years. It was an operator’s rough account, not an audited probability to plug into a spreadsheet.

“About ten start up, and perhaps two survive. Eight disappear.”

Keishi Kameyama · IVS conversation, 2022
Translated from Japanese

That tolerance comes with a warning from the same interview. Kameyama cautioned that startups might lose access to funding after one or two misses. He also said investments now needed board approval rather than simply his personal decision. The picture is more interesting than a founder waving through every idea: room to experiment, alongside a need to persuade other people.

The aquarium illustrates why patience sometimes needs a different clock. DMM Kariyushi Aquarium opened in Okinawa in May 2020, during the pandemic. Its exhibits combine animals with digital presentation, but the visitors still have to arrive in person. Feed, care and a building remain physical obligations. A group can transfer money between ambitions more easily than it can transfer the demand for online entertainment into aquarium attendance.

A theft that ended a business

DMM Bitcoin supplies a harder boundary. In May 2024, the exchange lost 4,502.9 bitcoin, valued at approximately ¥48.2 billion when the loss was disclosed. Japanese reporting subsequently described ¥55 billion raised through group support and other financing to replace the missing assets. The loss and the financing are different figures, measuring different things.

In December 2024, the FBI and Japanese authorities attributed the theft to North Korean actors. Their account began with a fake recruiter approaching an employee of wallet software provider Ginco. A malicious coding test compromised that employee; stolen session information later enabled access to communications. Authorities said the attackers likely used that access to manipulate a legitimate DMM transaction request.

That sequence matters because the initial compromise sat in a supplier relationship. A familiar consumer brand cannot, by itself, secure the chain of people and systems behind a financial transaction. Capital can help repair a loss after it occurs. It cannot make the original controls adequate retroactively.

The exchange’s service ended on March 8, 2025, with accounts and customer assets transferred to SBI VC Trade. This was the end of DMM Bitcoin’s service, not the end of DMM.com Group. It also shows why covering a customer shortfall and rescuing a business are separate achievements. An operation can fulfill the former obligation and still leave the market.

What has to be different this time

Against that history, the robotics return deserves a precise reading. The ¥5 billion is financing for Algomatic Dynamics from DMM, rather than an outside fundraising round for the entire conglomerate. The new venture says the money will support hardware and computing resources. Its proposed work includes collecting motion data, supplying AI hands and developing character robots for entertainment.

A domestic release of a multi-finger AI hand platform is planned for later in 2026. The announcement is a development plan, not evidence of widespread deployment. Its commercial test will come in the ordinary particulars of a customer’s site: the object to pick up, the task to repeat, the maintenance needed and the price of getting it wrong.

This is where DMM’s broad experience could help, although the connection remains an inference. A group with entertainment properties and manufacturing services has places to explore both character machines and practical production tasks. It still needs customers whose problems are worth the expense. Owning an interesting assortment of businesses is not the same as proving that a robot belongs in any of them.

The useful thing to copy is the requirement for a changed reason. Better technology and a newly trusted operator are explicit grounds for reconsideration. A smaller company would also need an affordable test and a stopping point. The comeback will be judged when the hand leaves the demonstration table and somebody expects it to do a paid day’s work.