Shibuya's IT Companies and the Lineage of Japan's Social Game Boom

What Actually Supported Japan's Internet Boom

Japan's internet boom from the late 1990s through the early 2000s is often described with the metaphor of "Japan's Silicon Valley." But when you look at the actual structure, both the source of the capital and the path to growth were quite different from the American venture ecosystem of the same period. Getting that straight first makes everything that follows much easier to understand.

1. The Capital Did Not Come from Independent VCs

What drove the American dot-com era was independent venture capital — firms that raise funds from pension funds, university endowments, and similar limited partners. Japanese venture capital, by contrast, had been dominated since the second venture boom of the 1980s by bank-affiliated and securities-firm-affiliated VCs: funds set up as subsidiaries of major financial institutions. JAFCO (which began within the Nomura Securities orbit) and Daiwa Securities' Nippon Investment Finance (NIF) are representative examples.

This difference determines investment behavior. The capital in an affiliated VC ultimately traces back to the parent financial institution's balance sheet or its clients' money, and combined with the fact that Japanese corporate finance was centered on indirect finance (bank lending) in the first place, investment decisions tended to be comparatively conservative. The American VC philosophy — premised on a power law, where "nine out of ten companies can fail as long as one returns 100x" — did not simply transplant into Japan and start working.

2. Operating Companies Doubled as Incubators

What played the large role instead, in Japan's first internet wave, was operating companies themselves. SoftBank is the emblematic case. In November 1995 it acquired 5% of the then-newly-founded Yahoo! Inc. for roughly ¥200 million, and in April 1996 an additional investment raised its stake to around 37%, making it the largest shareholder. That same year, 1996, it formed a joint venture with Yahoo! Inc. to establish Yahoo Japan Corporation and launched the domestic portal "Yahoo! JAPAN." This is a movement in which "investing" and "standing up a business through a joint venture" were fused into one — quite different in character from a VC acting as a pure financial investor.

Likewise, the general trading companies (sogo shosha) and major telecom and manufacturing firms were active in this period in taking stakes in internet companies and forming joint ventures. For a Japanese internet venture, having established large corporations that supplied not just capital but also customers, creditworthiness, staff, and distribution functioned to some degree as a substitute for the absent independent VC layer. Put the other way around: the conditions around a startup's independence were different from the American model.

3. Mothers as an Early Exit

The other decisive factor was the Tokyo Stock Exchange's Mothers market. On November 11, 1999, the TSE announced Mothers (Market of the High-growth and Emerging Stocks) as a new market for emerging companies. Trading began on December 22 of the same year, with Internet Research Institute and Liquid Audio Japan as the first two listings.

The essence of Mothers lay in the looseness of its listing criteria. A company founded only recently, or one still operating at a loss, could list as long as it could demonstrate growth potential. That was a fundamentally different design from the existing First and Second Section standards. The result was a standard path for Japanese internet ventures — "list on an emerging market within a few years of founding, then raise capital there and grow the business" — quite unlike the American pattern of stacking up repeated large private VC rounds while staying private. The existence of this path largely dictated the growth speed of the Shibuya companies discussed below.

4. The Business Models Skewed Toward Portals and E-commerce

The business areas chosen by Japan's first-wave internet companies were also narrower in range than in the United States. At the center sat portals (advertising) and e-commerce. Yahoo! JAPAN, Rakuten, auctions, internet ad agencies — concentrated into two templates: "gather an audience and sell ads," and "build a marketplace and take a commission." In the same period in the US, companies were emerging simultaneously across a much wider set of layers — infrastructure, enterprise software, search technology, payments, networking hardware — and that difference in breadth fed into later differences in industrial structure.

Bit Valley — The Emergence of Shibuya's IT Ventures

Against this backdrop, IT companies began physically clustering in Tokyo's Shibuya district. The name comes from the "Bitter Valley" concept proposed in 1998 by Kiyoshi Nishikawa, then head of NetAge. Shibuya's two characters translate literally as "bitter" and "valley"; crossing that with bit, the smallest unit of data, produced "Bit Valley." That the name was a nod to Silicon Valley is obvious enough.

Bit Valley was less a formal organization than a loose movement built around networking events and a mailing list among founders. At its 1999 peak, mixers drew hundreds to around a thousand attendees to venues in Shibuya. Out of this came the executives of companies that still matter today. Below we look at three of the most representative.

CyberAgent (founded 1998)

Founded by Susumu Fujita on March 18, 1998. Its initial business was internet advertising agency work: it owned no media of its own, and instead bought ad inventory on other companies' websites and resold it to advertisers — a labor-intensive business. The internet ad market at the time was in its infancy, a field you could break into on sales strength alone.

Then, in March 2000, just two years after founding, it listed on TSE Mothers. Fujita was 26 at the time, becoming the youngest president of a listed company in Japanese history. This is precisely the archetype of how "Mothers as an early exit," described above, shaped the growth path of Japanese internet ventures. The company then worked to shift from an agency (contract-work) model toward its own services, launching the blogging service Ameba Blog (Ameba) in 2004. That was the point at which it first became a company with its own media and its own platform.

Its serious entry into games came in the 2010s. In May 2011, Koichi Watanabe, formerly of Silicon Studio, founded Cygames with investment from CyberAgent. Just four months later, in September 2011, it shipped its first title, Rage of Bahamut. This is the company that would go on to make Uma Musume, discussed below.

DeNA (founded 1999)

Founded in March 1999 as the limited company DeNA by Tomoko Namba, formerly of McKinsey. That November it launched the internet auction site Bidders. In other words, its founding business was not games but e-commerce and auctions — fitting neatly into the "Japan's first wave skewed toward portals and e-commerce" tendency described above. Through models such as supplying auction functionality to partner sites, Bidders grew into the second-largest such service in Japan after Yahoo! Auctions.

The turning point was mobile. In 2004 it launched Mobaoku, an auction site for mobile phones, which grew into one of the largest mobile auction sites in Japan. It listed on TSE Mothers in February 2005. Then, in February 2006, it launched Mobage Town, a mobile social network with games attached. This spread explosively among junior-high and high-school students, reaching five million registered members roughly 15 months after launch. Here an e-commerce company shifted its center of gravity toward being a game platform company.

GMO Internet Group (origins in 1991)

The oldest of the three in origin, and the furthest from games. Its root is Voice Media, founded by Masatoshi Kumagai in May 1991, which initially distributed audio content over the Dial Q2 premium-rate service and sold related equipment. In November 1995 it was renamed InterQ, Inc. and pivoted into internet business, and in August 1999 it achieved what is described as the first JASDAQ listing by an independent Japanese internet venture. In April 2001 it was renamed Global Media Online (GMO).

What GMO chose was the infrastructure layer of the internet: domain registration, hosting, SSL certificates, and payments. "Discount Domain dot com," launched in 1999, is the predecessor of today's Onamae.com. In later years it expanded into financial territory as well — securities, FX, and crypto assets. The contrast is instructive: another product of the same Bit Valley, it moved not toward consumer-facing content but toward selling the substrate that other companies' internet businesses run on, showing that the Shibuya cluster was never converging on a single business model.

How the Gacha Revenue Model Came Together

In the late 2000s, Japan's mobile internet had a set of conditions that were unusual even by global standards. Feature phones (so-called garakei) were highly developed, and carrier billing — a mechanism that let even minors without credit cards make small payments — was in nationwide use. What arose on top of this was the browser-based social game.

The other pole, alongside DeNA's Mobage Town, was GREE. Its starting point was a social network Yoshikazu Tanaka stood up as a personal hobby project in February 2004; in May 2007 it released the mobile social game Tsuri-Sta, and it listed on TSE Mothers on December 17, 2008. DeNA, for its part, launched Kaito Royale in October 2009 to major success.

Structurally, the important development is that both companies became platforms in this period. In January 2010, DeNA opened Mobage Town to outside developers as an open platform, and on March 28, 2011 renamed the service to Mobage. That is, it went from being a company that made games to being a company that provided the place where other companies' games ran, plus the billing infrastructure underneath. You can read this as the "provide a marketplace and take a commission" model, honed in its 2000s e-commerce and auction business, being transplanted directly into games.

And what became established on top of that was gacha. The name comes from physical capsule-toy vending machines (gachagacha), and it refers to a mechanism in which items or character cards are dispensed at random for a fee. The decisive difference from a traditional buy-once game is that the price of the content is not fixed, and there is no ceiling on what an individual user can spend. Free-to-play lowered the barrier to entry, and a revenue structure in which a small number of heavy spenders carry most of the revenue was established here.

Three Cases — IP Combined with Gacha

The reason the gacha model has kept producing major hits for more than a decade is not simply that the model is powerful. In Japan's case, it is combined with the asset value of existing IP. The three examples below illustrate well how that combination works.

THE IDOLM@STER — Twenty Years That Started in an Arcade

The original THE IDOLM@STER was an arcade game from Namco (now Bandai Namco Entertainment) that began operation on July 26, 2005. Development had started in October 2001, and as an idol-production game with no real precedent at the time, it went through a long development period before reaching the market. In other words, the starting point was not gacha at all — it was an arcade cabinet.

That IP moved to mobile in the 2010s. On November 28, 2011, jointly developed by Bandai Namco and Cygames, THE IDOLM@STER Cinderella Girls (popularly "Mobamasu") launched on DeNA's Mobage. The game system was reworked into a card-battle format in which character cards are collected via gacha. It ran for eleven years, until its service ended on March 30, 2023.

From there the brand kept branching and expanding across platforms: Million Live! on GREE in 2013, SideM featuring male idols in 2014, and Shiny Colors on the browser platform "enza" in 2018. What deserves attention here is that an IP from a traditional game company (Bandai Namco) went mobile by way of a Shibuya IT company's platform (Mobage) and development capacity (Cygames). The division of labor between IP holder and platform operator was established in this period.

Uma Musume Pretty Derby — Five Difficult Years, Then a Huge Success

Cygames' Uma Musume Pretty Derby is a game in which the player raises characters that are anthropomorphized versions of real racehorses. The project was first revealed at AnimeJapan 2016 on March 26, 2016, with details announced at Cygames NEXT 2016 that August. The game was initially slated for release in winter 2018, but was postponed for quality reasons and delayed repeatedly thereafter. The eventual release came on February 24, 2021 — roughly five years after the first announcement. In the game industry, prolonged delays are usually read as a bad sign, and at the time there were indeed people asking whether it would ship at all.

The outcome was the opposite. From launch it ran away with the top spot in Japan's mobile game revenue rankings, and it was reported to have reached on the order of ¥100 billion in revenue during 2021. Cygames' results for the fiscal year ended September 2021 showed revenue of ¥222.1 billion (up roughly 101% year on year) and ordinary profit of ¥90.7 billion (up 273%), a sharp expansion driven by the title. Parent company CyberAgent's game segment also posted record revenue. As of May 2024, cumulative worldwide revenue was reported to have passed roughly ¥379 billion.

What this case shows is that the gacha model is not necessarily only a mechanism for building cheaply and quickly to recoup fast. Because live-service games are premised on sustained long-term spending, there was a structure in which five years of up-front craftsmanship could still be recouped. And the design decision to turn Japanese horse racing — an existing cultural asset, with real historical data — into characters is a hybrid IP strategy: a new IP that nonetheless borrows the accumulated weight of existing stories.

Pokémon Trading Card Game Pocket ("Pokepoke") — DeNA's Return

The most recent case is the app popularly known in Japan as "Pokepoke." It reworks the Pokémon Trading Card Game for smartphones, and released worldwide on iOS and Android on October 30, 2024 (following a soft launch in New Zealand on September 26, 2024).

The development credits go to Creatures Inc. and DeNA, with The Pokémon Company as publisher. The reappearance of DeNA here is the very theme of this article: a company that started as an auction site in 1999 and built a social game platform with Mobage Town in 2006 is, in 2024, handling development and operation of a mobile app for one of the largest IPs in the world.

The commercial results were extremely large as well. Reported figures include roughly two million downloads and about $2.7 million in revenue on the first day, $100 million within 17 days of launch, roughly $200 million in the first month, and about $400 million within ten weeks. As of February 2025 it was reported to have exceeded 100 million cumulative downloads and $500 million in cumulative revenue, substantially outpacing Pokémon GO's launch month. DeNA's profit was reported to have risen by an enormous margin year on year as a result.

At the same time, in digitizing a physical card game, a design that puts the experience of opening packs at its core means that card distribution is itself gacha. Reviews praised the collection experience and accessibility highly, while also criticizing the gacha elements and the limits (wait timers) on how often packs can be opened.

Connecting the Lineage — What Was Actually Happening

Viewed as a single thread, the following structure comes into focus.

First, Shibuya's internet ventures did not begin as game companies. CyberAgent began in advertising agency work, DeNA in auctions, GMO in audio information services and then internet infrastructure. None of these origins has anything to do with games. What they possessed was the technology to aggregate web traffic and the machinery to bill against it — not game development know-how.

Second, the feature phone — a distinctly Japanese environment — connected those two assets to games. Carrier billing as a payment layer, and a device carried at all times. Build a "place" on top of that, and the capabilities these companies already had could be used directly. DeNA opening Mobage as a platform in 2010 was an entirely natural choice: rather than making games itself, it would behave as a platform operator.

Third, existing IP flowed in. IP from a traditional game company such as THE IDOLM@STER, the cultural asset of horse racing, and a globally iconic property like the Pokémon Trading Card Game — all of them rode on the platforms and the development and live-operations capacity that the Shibuya companies had built. The combination of strong IP × the gacha revenue model has kept producing major hits for over a decade, from Mobamasu in 2011 to Uma Musume in 2021 to Pokepoke in 2024.

And finally, back to the capital structure we started with. Japan's internet ventures grew up not on large independent VC rounds but on corporate investment and an early listing via Mothers. This is less an "inferior" version of the Silicon Valley mechanism than a different mechanism that worked under different constraints. The result: Japan's game content market reached roughly ¥2.3961 trillion in 2024 (Famitsu Game Whitepaper 2025), its mobile segment remains one of the largest in the world, and Japanese mobile games have become one of the country's significant export categories.

In short, Japanese internet ventures that started from a capital structure unlike Silicon Valley's went on to grow an industry Silicon Valley never produced: character-IP-driven, live-operated gacha mobile gaming. The question of whether Shibuya's Bit Valley "managed to become Japan's Silicon Valley" was probably the wrong question to ask in the first place.

Summary

• Japan's late-1990s internet ventures were supported not by independent VCs but by bank- and securities-affiliated VCs and by operating companies such as SoftBank
• TSE Mothers, announced in November 1999, created a standard growth path — listing within a few years of founding — distinct from the US pattern
• Shibuya's "Bit Valley" (from Kiyoshi Nishikawa's 1998 "Bitter Valley" concept) produced CyberAgent (1998), DeNA (1999), and GMO (origins in 1991)
• None of the three started as a game company; their origins were advertising, e-commerce, and infrastructure respectively
• Feature-phone carrier billing made gacha spending possible, and Mobage (opened as a platform in 2010) and GREE captured the platform layer
• Complete gacha was judged problematic under the premiums and representations act by the Consumer Affairs Agency in May 2012; drop-rate disclosure and similar measures are now handled through industry self-regulation
• THE IDOLM@STER (2005 arcade → 2011 Mobamasu), Uma Musume (announced 2016 → released 2021), and Pokepoke (2024, developed by Creatures and DeNA) demonstrate the durability of strong IP combined with gacha
• DeNA threads through this entire lineage, from platform provider in the 2000s to developer-operator of a major global IP in the 2020s