Games Confidence: Med by 2029
Japan's mobile game market stays mature: store spending shrinks, and revenue rests on spend per user and on sales outside the stores
Store-measured spending in Japan keeps falling while playtime holds. In-app game spending over the twelve months to July 2026 was just over USD 10bn, down 5.1%, on 585 million downloads, down 5.4%, with 16.7bn hours played, roughly stable since 2024. MCF's measure, which includes external billing, grew 8% in 2025, so part of the fall is a move off the stores rather than lost demand. In whale-monetised titles the top 5% of payers generate 81% of revenue by month 10, a figure for that title type, not the whole market.
›Evidence and background
Evidence weight: six independent sources in five layers point this way. Sensor Tower's store data (games, trending-games, apps) shows the decline, and MCF's all-route count (games) shows the move off the stores. Newzoo (games) puts first-half 2026 mobile downloads down 25% and cost per install up 30% to USD 0.56. Moloco (games, trending-games, product-craft) documents payer concentration, and Teikoku Databank (jp-macro) counts 10 smartphone-game bankruptcies in January-July 2026 against three in all of 2025. The Leisure White Paper (consumption) reports leisure time and leisure spending both turning down in 2025. The main counter-reading is Newzoo's forecast of 6.1% global market growth in 2026, which it attributes to existing players spending more. Store data cannot separate lower spending from spending moved off-store, and the one all-route measure rose 8% in 2025, so confidence is medium. Base rate: Teikoku Databank counts 29 smartphone-game failures over the past five years, and 2026 is on pace to approach the annual record of 12 set in 2015. Global casual-game downloads nearly halved from 11.2bn in the first half of 2021 to 6.34bn. Exposure: Digital Entertainment had FY2026 revenue of JPY 83.889bn, and in Q1 FY2027 it earned JPY 10.881bn of segment profit against JPY 1.258bn in Sports, about 80% of the total. Monster Strike's revenue is about 99% Japanese, so the direction of Japanese spending moves the group's largest profit source directly. Japan's top-grossing order is also shifting: eFootball led January-July 2026 with Monster Strike second, and 4X strategy grew 9.4% in Japan.
Why (PESTLE): Falling downloads and rising install costs, a shrinking store channel, revenue concentrated in heavy spenders, and time lost to video and AI apps. Against them: off-store sales that store trackers miss, IP collaborations that lift payer rates, and forecasts that see growth from existing players spending more.
Implication for MIXI: Digital Entertainment supplied about 80% of segment profit in Q1 FY2027, and Monster Strike's revenue is about 99% Japanese. Segment revenue rose 21.5% to JPY 19.53bn on IP collaborations while MAU kept falling, so growth rests on spend per user.
›Evidence and background
Weekly ranks swing with events: Monster Strike moved from first in Japan's weekly revenue chart in July to seventh for 10-16 September, when revenue fell 40% on the week during an anniversary free-pull campaign, so one week does not show the trend. In 2024 the title earned USD 570.7M, a post-peak level. Spending through its web shop does not appear in store rankings, so Sensor Tower's second place counts store spending only. Reliance on a small group of heavy spenders overlaps with what regulators cite on compulsive spending. MIXI's CFO said IP collaborations raised ARPU with both payer rate and ARPPU improving, so the Q1 rise was not a narrowing to fewer payers.
Open Resolves 2029 — Confirmed if Japan's in-store mobile game spending on public trackers stays at or below its 2026 level in each year to 2029. Defeated if it grows by more than 5% a year in two consecutive years, or if Japanese game downloads return above their 2025 level. Where an all-route measure such as MCF's grows, record the gap beside the verdict. Also defeated if an all-route measure such as MCF's Japan market grows by more than 5% a year in two consecutive years, since the store decline would then reflect a channel shift rather than a mature market.
Games Confidence: High by 2030
Game supply grows faster than demand, so new titles rarely break out and discovery, not production, sets the cost of a launch
Supply rises while installs fall. Counting all apps, not only games, monthly Google Play submissions went from 46,136 in March 2025 to 111,047 in July 2026, and App Store submissions more than doubled over the same 16 months. Of about 190,000 mobile games released in 2025, only 1.3% passed 500,000 first-year downloads. Global game downloads fell 12% in the first half of 2026 to 24bn. Catalogues gather with buyers of proven live titles: Tripledot bought Unity's Supersonic for USD 40M after buying AppLovin's games business for USD 800M, though both sellers cited their own focus.
›Evidence and background
Evidence weight: five independent sources in four layers describe the same shape: 42matters submission data (trending-games), Moloco's launch analysis (games, trending-games, product-craft), Sensor Tower and AppMagic download series (trending-games), Newzoo's install-cost figures (games) and Steam disclosure data (ai). The submission counts cover all apps, not only games, and the source does not identify the cause of the rise, so a causal link to generative tools remains an unverified hypothesis. On Steam, titles disclosing generative AI rose from 10.9% of releases in 2024 to 30.8% in 2026 but took only 10-27% of estimated sales. By 25 May 2026, 52 live-service games had shut. Base rate: casual-game downloads nearly halved from 11.2bn in the first half of 2021 to 6.34bn. Western match-3 spending was flat at USD 2.5bn in the first half of 2026, and only 1 of 120 new titles cleared USD 100,000 a month. The counter-evidence sits at the top: merge-2 spending rose 95% on 20% more downloads, and hybrid-casual Pixel Flow took USD 198.4M in its first year. Exposure: Digital Entertainment's new titles and STRIKE WORLD carry the discovery cost, while Monster Strike's installed base, in its 13th year, is an asset a new entrant cannot rebuild cheaply. MIXI publishes no launch cost or new-title figures, so the size of this exposure cannot be measured from outside.
Why (PESTLE): Tools lowering build cost (the contribution of generative AI is unverified), more solo and small-team releases, and stores that rank by performance. Against them: a few breakout genres such as merge-2 and hybrid-casual, platform programmes that surface new titles, and buyers willing to pay for proven live games.
Implication for MIXI: Monster Strike is in its 13th year and had grossed USD 11.4bn on mobile by its 11th, a base new entrants cannot buy cheaply. STRIKE WORLD, the overseas entry that began full operation in India on 16 April 2026, has published no download, revenue or retention figures.
›Evidence and background
Through LVLZero Tokyo, MIXI supports indie developers going overseas without taking equity or publishing rights, which puts it on the supply side of the same glut. The company reports that generative AI cut Monster Strike video production from 21 days to 3, a company figure; cheaper production does not change the discovery constraint.
Open Resolves 2030 — Defeated if, before 2030, the share of new mobile games passing 500,000 first-year downloads rises above 3% on a comparable tracker, or global mobile game downloads return above their 2025 level.
Apps Confidence: High by 2029
Store fees become a table split by jurisdiction, route and install date, and while statutory routes go unused, transactions move to publishers' own web shops
A commission stops being one number. Apple's single EU terms took effect on 1 October 2026: 26% with In-App Purchase, 20% with alternative payment, 15% on link-outs and 5% for off-store distribution. Google's new Japan fees, live since 30 September, split by install date: for new installs, 10% on the first USD 1M a year and on subscriptions and 20% above it; for existing installs, 25% above USD 1M; plus 5% with Play billing. In Japan the statutory routes are barely used: an eight-body survey found alternative stores at 5.9% and in-app alternative payment at 8.8%.
›Evidence and background
Evidence weight: eight layers (apps, revenue-models, games, trending-games, daily-apps, marketing, product-craft, mixi) carry mutually independent sources: the platforms' own documents (Apple's and Google's terms and rollout tables), industry surveys (the eight bodies, MCF), and regulators and courts (the European Commission, the JFTC, the UK Competition Appeal Tribunal, the US Supreme Court). D2C web shops are a parallel route worth about 15% of mobile in-app spending, about USD 17bn a year, and DoubleDown disclosed 52.4% of its Q2 social-casino revenue through D2C. On 24 September MCF argued that Apple's and Google's terms breach Article 8 of the smartphone act; no JFTC response has been published. Apple's terms still bar digital content sold only on the web. Base rate: within five years, statute- or court-based route rates came into force under the EU's DMA, Japan's smartphone act (in full effect from 18 December 2025), Brazil's CADE agreement and a US injunction, and the UK and Australia added settlements. The level is not settled: the UK tribunal's 17.5% and 10% benchmark is on appeal, and US link-out rate-setting has stalled. Counter-evidence on the realised effect is mixed: a third-party estimate puts Apple's US commission revenue down 18% since the start of 2026, while Apple's Services revenue grew 12% year on year, but Services covers far more than the App Store and gives no split of the regulatory effect. Exposure: Monster Strike's revenue is about 99% Japanese, so its effective fee follows Japan's terms (26% through Apple In-App Purchase, which is a 21% commission plus 5% processing; 21% with alternative payment; 15% on link-outs; 5% for alternative marketplaces) and its web-shop share. FamilyAlbum Premium follows the subscription rates.
Why (PESTLE): Three regimes on separate calendars (Japan's act, the EU's DMA and the US injunction), web shops that keep part of the store fee, and UK and US proceedings that keep fee levels in play. Against them: statutory routes barely used, web-only assortment still barred, and procedural and age-protection duties that move in-house with the payment.
Implication for MIXI: Without in-app steering, use of the Monster Strike Web Shop reached just under 50% in October-December 2025. Web-shop spending does not appear in store totals, and MIXI discloses no split of billing by route. FamilyAlbum has over 40% of users overseas, where route rates change on different dates by region.
›Evidence and background
The Web Shop sold a Jujutsu Kaisen collaboration gacha on the web ahead of the app, a timing difference rather than an exclusive, which is the lever Apple's terms leave open. External billing moves payment processing, refunds, tax handling and limits on minors' spending in-house, so lower fees and higher operating and compliance cost arrive together. Google's Japan cut is smallest for long-running titles whose users installed before 30 September, so Monster Strike's effective fee moves more with its Web Shop share than with the new table.
Open Resolves 2029 — In at least two of Japan, the EU and the US, a majority of large publishers run an off-store payment route AND several listed game companies disclose a D2C or web-shop share on a recurring basis, verifiable from company disclosure and industry surveys. Take-up of alternative app stores staying below 10% does not by itself defeat the prediction.
Consumers Confidence: High by 2030
Rules on chance mechanics and on retention and purchase design arrive through ratings, consumer law and courts before any gacha statute, and Japan's route runs through interface rules
The unit of regulation moves from what is sold to how purchase and return are designed. PEGI has rated paid random items 16+ by default since June 2026, and Brazil barred loot-box sales to under-18s from March 2026. On 17 September the European Commission proposed the EU KIDS Act, whose game duties cover addictive design and economic transactions; whether its bans reach login streaks is read two ways. In Japan the route is consumer law: on 10 September the Consumer Affairs Agency's study group finalised an interim report backing rules on misleading and aggressive interfaces, and 38 oripa buyers sued five sellers under the Premiums Act.
›Evidence and background
Evidence weight: seven layers (revenue-models, games, trending-games, marketing, consumption, emerging-experiences, eu-macro) carry independent sources: a ratings body (PEGI), legislation (Brazil's law, the Commission's proposal, Korea's odds-disclosure rules), a regulator (the Consumer Affairs Agency) and courts and claimants (the oripa suit). In Korea Nexon drew a record KRW 11.6bn penalty over drop-rate disclosure in 2024. High-spending customers face the same gaze: Australia's reform limits spend-linked pay for VIP managers, and New South Wales proposes banning VIP programmes. Base rate: Japan's only direct rule on paid random items is the 2012 ruling against complete gacha, and industry self-regulation has governed since. Abroad, at least three binding changes arrived in five years: Korea's statutory odds disclosure (in force March 2024), Brazil's ban on sales to under-18s and PEGI's new default rating. In the US the FTC's January 2025 order against HoYoverse (USD 20M) already bars loot-box sales to under-16s without parental consent. Counter-evidence: final adoption of the KIDS Act before 2028 is unlikely, and the Digital Fairness Act, expected to address loot boxes, had no formal proposal on 23 September. The Netherlands' Council of State held in March 2022 that loot boxes in EA's FIFA were not a game of chance under Dutch gambling law. Exposure: Digital Entertainment supplies about 80% of group segment profit and gacha is its core monetization. If the Consumer Affairs Agency lists unlawful interface patterns in subordinate rules, standard techniques such as limited-time offers and remaining-stock displays could fall within them. The same high-spender logic reaches the Sports segment's betting.
Why (PESTLE): Ratings, consumer law, court orders and settlements advancing together, design duties for minors, and harm evidence that regulators cite. Against them: final EU texts years away, disputed readings of what counts as compulsive design, and Japan's reliance on self-regulation for gacha.
Implication for MIXI: MIXI holds gacha monetization (Monster Strike, STRIKE WORLD) and betting (PointsBet, TIPSTAR, Chariloto) in one company, so odds disclosure, age checks, deposit limits and self-exclusion apply from both sides. Log-in rewards, limited-time offers and spend-linked rebates such as the 13th-anniversary Orb Back Challenge are designs whose treatment the KIDS Act readings and Japan's dark-pattern work leave open.
›Evidence and background
Monster Strike's 13th-anniversary campaign gave one free ten-pull a day for 13 days, a retention and return mechanic of the kind the KIDS Act debate addresses for minors; how the final text treats games is open. Japan has no statute aimed at paid random items, so the gap between its domestic standard and overseas statutory requirements widens while the Consumer Affairs Agency defines interface patterns. The ESG indices MIXI has joined do not score product-level chance mechanics. Because the draft's game duties cover economic transactions, gacha design aimed at minors in Europe could be set by child-safety law as well as by gambling and consumer law.
Open Resolves 2030 — Confirmed if, between October 2026 and 2030, legally binding requirements on paid random items or on retention or purchase-interface design (statute, subordinate rule, court order, enforceable undertaking or settlement term) newly apply to major consumer games in at least two of the EU, Japan and the US; the FTC's January 2025 HoYoverse order, already in force, does not count. Defeated if Japan's dark-pattern work ends without binding rules and both the KIDS Act and the Digital Fairness Act lapse.
Betting Confidence: High by 2030
In Australia and Canada, betting regulation bites through acquisition limits, taxes and conduct duties, so betting revenue and betting profit move apart
Regulators act on how customers are acquired and kept, and on tax, more than on whether betting is allowed. Australia's advertising law starts on 1 January 2027: a live-sport advertising ban from 15 minutes before play, direct marketing barred for 14 days after sign-up, three months after self-exclusion ends and indefinitely for at-risk customers, and no pay tied to customers' betting. New South Wales proposes a ban on VIP programmes. Canada opens province by province, with Alberta live since 13 July. Revenue and profit split: Atlantic City's first-half 2026 gaming revenue rose 7.3% while gross operating profit fell 15.5%.
›Evidence and background
Evidence weight: seven layers (betting, au-macro, sports-business, revenue-models, mixi, marketing, consumption) carry independent sources: Australia's parliament and communications regulator (ACMA), the New South Wales government, iGaming Ontario and Ontario's regulator, UK tax policy and a Lords committee, a Brazilian Senate committee, and industry disclosure. ACMA fined Dabble about AUD 1.07m over BetStop failures and fined Tabcorp AUD 2.7m. State point-of-consumption taxes run at 15-25%. In Australia, ASIC warns that prediction markets are unlicensed and ACMA blocked Polymarket, so they cannot enter as unregulated competitors. In Canada, securities regulators treat sports event contracts as outside securities law, leaving them to provincial gaming rules and the Criminal Code. Base rate: advertising rules have moved in one direction. Italy keeps a near-total ban, the Premier League's removal of gambling brands from shirt fronts applies from 2026-27, the Netherlands proposed a near-total online advertising ban in June, and Brazil went from an advertising and sponsorship bill approved in committee on 3 September to an outright ban by provisional measure on 25 September, the prohibition path covered in the wildcards. Tax moved the same way: the UK's remote gaming duty rose from 21% to 40% in April 2026. Counter-evidence: Great Britain's Gambling Commission invited proposals in June to cut unnecessary burdens, and Finland opens its market in July 2027. Evidence on ad bans is split: Denmark's channelisation fell from 90% to 70%, and Italy's ban coexists with a EUR 22bn illegal market. Exposure: the Sports segment had FY2026 revenue of JPY 65.848bn (+63.8%) and segment profit of JPY 5.088bn; in Q1 FY2027 it earned JPY 1.26bn on JPY 21.67bn of revenue. PointsBet says taxes and regulatory fees take about 50% of its revenue, a company claim.
Why (PESTLE): Harm evidence driving limits on advertising, inducements and VIP schemes, higher gambling taxes, self-exclusion and AML enforcement, and a Canadian market that opens province by province. Against them: burden reviews that stress channelling players to licensed operators, split evidence on whether ad bans work, and newly opening markets.
Implication for MIXI: MIXI holds 66.43% of PointsBet. In the nine months to March 2026 its Australian net win fell 4% and statutory EBITDA 30%, while Canadian iGaming net win rose 28%. Sports is the largest segment by revenue, yet in Q1 FY2027 it earned JPY 1.26bn of profit on JPY 21.67bn of revenue.
›Evidence and background
Canadian growth is casino-weighted: in August 2026 casino supplied 85% of Ontario market revenue, and PointsBet's Canadian sports handle fell 39% over nine months. In February Ontario's regulator proposed a five-day suspension of PointsBet Canada's registration over failures to detect and report suspicious bets, the province's first. Australian operators diverge: Entain's online revenue rose 13% in the first half of 2026, while the leader Sportsbet, with about 45% share, fell for a third year in 2025 and PointsBet fell 4% in the nine months to March 2026. The periods differ, so the figures show neither a market-wide decline nor a fall caused by company factors alone. betr, a listed competitor, holds about 27.7% and voted against every resolution at the August meeting.
Open Resolves 2030 — Confirmed if, through 2030, the core of Australia's 2027 advertising law (the live-sport advertising ban and the limits on inducements) stays in force or tightens, and no across-the-board cut in gambling tax rates occurs in the UK, Australia or Canada. Defeated if Australia repeals or materially relaxes the core advertising restrictions, verifiable from statute and regulator releases.
Betting Confidence: High by 2031
Japan keeps private sports betting illegal, and legal wagering keeps growing through online sales of the public competitions under tighter advertising rules
Japan remains the only G7 country without legal private sports betting, and sales move online inside the public competitions. Keirin ticket sales reached JPY 1,548.8bn in fiscal 2025, up 16.6% and a thirteenth straight rise, and JRA horse racing grew for a 14th year, with net voting the consistent driver. The growth has limits: a 2026 advertising guideline caps points rebates at 1% for ordinary races, and the amended addiction law bans online casinos and their promotion. Illegal offshore sports betting is estimated at about JPY 6.5tn a year.
›Evidence and background
Evidence weight: six layers (betting, revenue-models, sports-business, mixi, marketing, jp-macro) carry independent sources: the racing bodies' sales statistics, netkeirin's tallies, the Council for Sports Ecosystem Promotion's estimates, legal summaries from ICLG and The Japan Times, METI, Japan Sports Agency and Development Bank of Japan papers, and Jiji Press enforcement data. In fiscal 2025 boat racing set a record for a fourth year and JRA for a 14th. Base rate: sports-betting legalization has been raised repeatedly in LDP study groups and disowned as formal policy. The only widening in five years was adding basketball to toto in 2021, and what NPB's owners agreed to study on 16 July is a non-prediction baseball lottery. Keirin ticket sales rose about 2.45 times in ten years from JPY 630.8bn in fiscal 2015. Counter-evidence: sales are total stakes, so addiction and consumer-protection questions grow with them. An estimated 3.37m people in Japan have used illegal online casinos, and 196 of the 317 people arrested in 2025 illegal-casino cases were individual users. Exposure: the Sports segment's domestic wagering (TIPSTAR, Chariloto, velodrome operations) earns fees tied to net-voting turnover within a takeout of about 25%. TIPSTAR revenue rose 12.8% in Q1 FY2027 against about 3% growth in the keirin market, slowing from 55.8% in FY2026. Chariloto rose 71.4% in FY2026. PIST6 is not a MIXI business: MIXI sold its stake in the operating company in November 2022, and from August 2023 the operator JPF also sold tickets on its own official site, so TIPSTAR was one channel among others. PIST6 meetings have been suspended since September 2025, and Chiba City has pushed the restart to fiscal 2030. The public monopoly protects organisers, not sales channels, so growth at TIPSTAR and Chariloto also depends on competition between channels.
Why (PESTLE): Wider net voting, a public system whose legitimacy rests on returning proceeds to public purposes, and enforcement against illegal markets. Against them: addiction measures and advertising rules, the cap on points rebates, a crowd of AI tipping services that makes differentiation hard, and scrutiny that rises with total stakes.
Implication for MIXI: TIPSTAR and Chariloto earn commissions on net-voting turnover, and TIPSTAR revenue rose 12.8% in Q1 FY2027 against about 3% growth in the keirin market. MIXI also owns netkeiba, with about 17 million monthly users. PIST6 is not a MIXI business: MIXI sold its stake in 2022, and the operator also sells on its own site.
›Evidence and background
The rebate cap means rebates no longer set one betting site apart from another, which bears directly on how TIPSTAR and Chariloto design acquisition spend. Because netkeiba is a subsidiary, the surface any advertising rule reaches is wider than the betting apps. Social design for watching and betting with friends lowers the threshold to take part, so how growth among new audiences is reconciled with harm prevention is itself assessed.
Open Resolves 2031 — Defeated if Japan enacts legislation permitting private or fixed-odds sports betting by 2031. Also defeated if keirin ticket sales fall below fiscal 2025's JPY 1,548.8bn in two consecutive fiscal years. Adding a non-prediction baseball lottery to toto does not count as legalization.
Sports Confidence: Med by 2030
Japanese club sport shifts to arena-led and league-threshold economics, and broadcast distributions become less certain than ticket and venue income
Club revenue shifts toward venues clubs can operate. B.LEAGUE's top tier opened on 22 September 2026 with entry thresholds of JPY 1.2bn in revenue, 4,000 average attendance and a 5,000-seat arena, and Japan opened at least four new arenas in 2025 alone. The J.League moved to an autumn-spring calendar on 7 August, and its opener drew a record 63,960. Broadcast money is less certain: DAZN holds the J.League's domestic rights to 2033, but the deal has moved to profit-sharing. Betting on these games stays illegal in Japan.
›Evidence and background
Evidence weight: four layers (sports-business, mixi, revenue-models, bcp) carry independent sources: B.LEAGUE and J.League releases, Sporta Japan, SportsPro, The Japan Times, Mitsui Fudosan and the Asahi Shimbun. League figures are the leagues' own. Base rate: B.LEAGUE says attendance grew about 2.5 times and business scale nearly five times in the ten years since its 2016 launch, and it aims to be the world's second league after the NBA. The J.League's original domestic rights deal was a fixed ten-year agreement for 2017-2026 worth about JPY 210bn; the extension to 2033 moved to profit-sharing. Counter-evidence: more arena supply means harder competition for crowds and sponsors, and utilisation and operating cost decide returns. The autumn-spring calendar tests midsummer openings and winter operations, and the Aichi-Nagoya Asian Games competed for venues and demand in the same region at the September opening. Why the US rights framing was dropped: US figures such as the NFL's request for 50% more from CBS or a 13% fall in Week 1 audiences do not reach MIXI's revenue. The one link is DAZN, which has taken US NBA local rights for 11 teams, with exit clauses after this season. Exposure: Chiba Jets and FC Tokyo sit inside the Sports segment (FY2026 revenue JPY 65.848bn), which does not disclose club-level figures. Match days in the capital region have to assume stopped transport, holding fans and staff who cannot get home (up to 8.4 million stranded in a Tokyo quake) and deciding who may re-enter after evacuation; in Kumamoto seven tenant employees died in an explosion after being let back in.
Why (PESTLE): Arena investment driven by league entry thresholds, multipurpose venues used year-round, spectating as an experience, and a Europe-aligned autumn-spring calendar. Against them: competition from added arena supply, capex and utilisation burdens, broadcast distributions now on profit-sharing, and operation in summer heat and winter.
Implication for MIXI: Chiba Jets posted JPY 5.17bn of revenue in the 2024-25 season, the first B.LEAGUE club above JPY 5bn, and ticket revenue rose about 2.4 times after its move to the 10,652-seat LaLa arena TOKYO-BAY. FC Tokyo is in the first autumn-spring season.
›Evidence and background
Chiba Jets' ticket revenue rose from JPY 0.64bn to JPY 1.56bn, and pre-tax profit was JPY 0.50bn (+31.4%). FC Tokyo announced a merger with Nadeshiko League side Sfida Setagaya FC in June 2026, its first club-level exposure to women's football. MIXI promotes cross-traffic between spectating and betting, but public data showing it is limited. Illegal betting on Japanese sports is estimated above JPY 1tn, so match-fixing and athlete-harassment risk for FC Tokyo and Chiba Jets can arise outside any licensing regime.
Open Resolves 2030 — Confirmed if B.LEAGUE PREMIER keeps its revenue, attendance and arena entry thresholds at or above 2026 levels through 2030. Defeated if the thresholds are withdrawn or the J.League returns to a spring-autumn calendar. If J.League broadcast distributions are disclosed and fall below the pre-2026 fixed-fee level, that confirms the broadcast half.
Family Confidence: High by 2031
Japan's births keep falling, so family services grow through overseas users, price and paid features rather than through new domestic families
The domestic pool of new families keeps shrinking. Japan's 2025 births were 671,236, a tenth straight record low, with a fertility rate of 1.14, and the 670,000 level the national research institute had projected for 2040 arrived 15 years early. Nothing upstream points to a turn: January-July 2026 preliminary data show births up 0.5% but marriages down 6.2%, and in a government survey of about 100,000 people about 35% of unmarried respondents said they do not intend to marry. Growth in family apps shifts to overseas users and to goods and paid features.
›Evidence and background
Evidence weight: four layers (family, jp-macro, apps, mixi) carry independent sources that all point the same way: MHLW vital statistics, the Statistics Bureau, the Children and Families Agency's survey, and The Japan Times and Japan Today reporting the national institute's projection. Category moves agree: Tinybeans held about 95,000 paid subscribers flat while growing e-commerce revenue 646%, and Life360 grew advertising revenue 315%. Base rate: births and the fertility rate set record lows for ten straight years. Annual declines of more than 5% in 2022-2024 eased to 2.2% in 2025, but the direction did not change. The under-15 population was 13.29 million on 1 April 2026, down 350,000 on the year. Counter-evidence: marriages rose 2.6% in January-March 2026, but later preliminary data did not show it persisting. Exposure: the Lifestyle segment had FY2026 revenue of JPY 17.159bn, about 10% of group sales, with FamilyAlbum itself about 9%. FamilyAlbum reached over 65% of Japanese households with children by December 2024, so domestic growth is bound by births.
Why (PESTLE): Ten years of falling births, fewer marriages and weaker intent to marry, and high domestic penetration. Against them: growth in new families overseas, and revenue per user from paid plans, prints, GPS, advertising and goods.
Implication for MIXI: FamilyAlbum reached over 65% of Japanese households with children by December 2024, so its domestic growth depends on births. Over 40% of its 30 million users are overseas, and Lifestyle grew 24.5% in Q1 FY2027 on paid plans, prints, GPS, advertising and minimo.
›Evidence and background
Overseas, North America is about 20% of users and the largest market. US store-measured spending fell 3% in Q2 2026, but fee-free link-outs are named as the main cause, so store data cannot show whether demand there is growing. The category is consolidating: Tinybeans absorbed Qeepsake. Finer paid entry points, such as a JPY 380 sticker set and a USD 2.99 individual plan, raise revenue per user while the domestic pool shrinks. Its regulatory exposure runs through children's-data rules rather than age gates: FamilyAlbum is invitation-only and not named in ban lists, but the amended COPPA covers biometric identifiers and India's DPDP duties apply in full from May 2027. Lifestyle segment profit in Q1 FY2027 was JPY 0.26bn, about 2% of the roughly JPY 13.3bn segment total.
Open Resolves 2031 — Confirmed if Japan's births in 2030 are below the 2025 level of 671,236, verifiable from MHLW vital statistics. A sustained rise in births above that level defeats the prediction.
Consumers Confidence: Med by 2029
In Japan disposable time binds before disposable income, spending shifts inside the same wallet toward non-game apps and free video, and wage gains have yet to reach leisure
Competition becomes a contest for all low-ticket entertainment, not game against game. Japan's real wages rose 2.4% in July 2026, a seventh positive month, while real household spending fell 3.6%, an eighth decline, and the Leisure White Paper's advance edition says leisure time and leisure spending both turned down in 2025. Worldwide, non-game apps out-earn games: of USD 43.6bn of in-app spending in Q2 2026, games took USD 19.2bn. Surveys disagree on how fan spending spreads: Intage finds the share with an 'oshi' down to 33.6% while their game spending rose 2.6 times, but CDG counts participants up from about 13.84M to 19.4M with average spend falling.
›Evidence and background
Evidence weight: five layers (consumption, jp-macro, apps, creator-economy, games) carry independent sources: MHLW's Monthly Labour Survey, the Statistics Bureau's household survey, the Japan Productivity Center's Leisure White Paper, Sensor Tower, an Intage survey and Nielsen. In the US, streaming took 49.0% of TV viewing in July, with YouTube at a record 14.2%, and in China the free, ad-supported Hongguo Short Drama reached 168 million daily users. Amusement services pass on only 17.7% of cost increases. Base rate: household spending has fallen for eight straight months, and July's 3.6% drop was the steepest in about two and a half years. Japanese hours of game play have been roughly flat since 2024, so time is not growing. Counter-evidence: real wages have risen for seven months, the minimum wage rises to JPY 1,177 from October, and bills to cut the food consumption tax to 1% for two years from April 2027 (a revenue cost of JPY 4.3tn a year) go to a Diet where the ruling bloc lacks an upper-house majority. Tokyo's September core CPI rose to 2.7% and power and gas relief ended with September usage, which puts the real-wage streak at risk. If wage gains reach leisure, the prediction fails. Exposure: Digital Entertainment depends on Japan's low-ticket entertainment wallet and supplies about 80% of group segment profit. TIPSTAR and ticket revenue compete for the same wallet and hours.
Why (PESTLE): Competition for allocation inside one wallet (non-game apps, free video, microdrama and betting), falling Japanese household spending and leisure time, and spending concentrated in a few committed consumers. Against them: seven months of real wage gains, a higher minimum wage and a possible food-tax cut.
Implication for MIXI: In Q1 FY2027 IP collaborations lifted payer rate and ARPPU, and Digital Entertainment grew 21.5% to JPY 19.53bn while Monster Strike's MAU kept falling. TIPSTAR grew revenue 12.8% in Q1 FY2027 against about 3% growth in the keirin market.
›Evidence and background
What to watch is which turns first: the leisure line in household surveys, or payer rate and ARPPU inside individual titles. Deeper spending per user on a shrinking user base concentrates revenue in committed spenders, which overlaps with the social questions of compulsive spending and household strain that regulators now cite.
Open Resolves 2029 — Games' share of global in-app spending stays below its Q2 2026 level (USD 19.2bn of USD 43.6bn) AND Japan's real consumption expenditure in the household survey (two-or-more-person households) does not grow by more than 2% a year in two consecutive years, verifiable from public trackers and the household survey.
AI Confidence: Med by 2029
Top-tier list prices fall while the average inference price has levelled, cost per task turns on cache and context, and detectability, audits and disclosure settle in as fixed costs
Price cuts now start at the top. On 22 September 2026 Anthropic priced Claude Opus 5.5 at USD 4 input and USD 20 output per million tokens, and OpenAI halved GPT-6 Sol to USD 2/10. The usage-weighted average price fell below USD 1 per million tokens for the first time in August and has since levelled at about USD 0.96. Cost per task follows cache reuse, context length and dated promotions more than the headline rate. Duties now extend to third-party checks: EU AI Act Article 50 and California's SB 942 have applied since 2 August.
›Evidence and background
Evidence weight: four layers (ai, technology, games, marketing) carry independent sources: vendor price lists (Anthropic, OpenAI, Google), a usage-weighted price index, EU and California law, Gallup's international survey and CESA's survey. Part of the price is still dated: GPT-5.6 Sol's promotional price is guaranteed only to 21 November 2026, and Gemini 3.7 Flash's price doubles on 1 January 2027. Gemini 4 Argon opened at an introductory USD 2/10 that rises to USD 4/20. Fable 5.1 cut cache reads 75% to USD 0.25 per million tokens. In September California signed SB 813 and AB 1405 on independent verification and an auditor registry, and SB 1119 on independent audits of chatbots used by minors. Enforcement lags: only 9 of 27 EU member states have designated national authorities. Base rate: the technology layer records an order-of-magnitude fall in inference cost, and in August usage rose 47% while spending rose 7%. Counter-evidence: the average price has levelled at about USD 0.96, and long context doubles the input rate, so cost per task can rise while cost per token falls. The Intelligence Index changed version three times in September, so scores do not compare across months, and in Gallup's 37-country study only 36% of Americans expect AI to improve their lives. Exposure: group AI cost. The layers hold no disclosed figure for MIXI's AI spending, so frontier list prices have no measured link to any segment. The only figure is MIXI's self-reported effect of roughly JPY 1bn a year, about 5% of its FY2027 operating-income guidance of JPY 19.5bn, without outside verification. Group AI runs on outside models, and a US export directive once halted two Anthropic models worldwide for 18 days. The item stays in the set on evidence weight, and its cost and duty terms reach every business that uses generative AI. Its exposure is small among the predictions, and its confidence is set to medium to match. MIXI's own programme is not counted as evidence that the area matters.
Why (PESTLE): Price cuts at the top tier and cache-driven cost, and standing duties on detection, audits and minors' defaults. Against them: indices that change version month to month, dated promotions, public opinion that stays negative, and uneven enforcement capacity.
Implication for MIXI: The exposure is group AI cost, and the layers hold no disclosed figure for MIXI's AI spending. The only figure is MIXI's self-reported effect of roughly JPY 1bn a year, about 5% of FY2027 operating-income guidance of JPY 19.5bn. It has no outside verification and describes internal cost and lead time, not revenue.
›Evidence and background
The self-reported use above 99% and the Chief AI Officer from 1 August 2026 are facts about MIXI, not evidence about the market. Competitors can buy the same tools, so an efficiency edge is likely to be temporary. Because cost per task depends on cache and on promotion end dates, internal cost plans follow vendors' repricing dates. The effect after quality requirements, rights clearance and disclosure duties is not disclosed.
Open Resolves 2029 — Confirmed if, in 2029, the standard list prices (excluding introductory and promotional prices) of the tiers named in the prediction, Claude Opus 5.5 (USD 4/20) and GPT-6 Sol (USD 2/10), and Google's Gemini 4 Argon at its announced standard price (USD 4/20), or their same-tier successors, are at or below those levels AND a majority of major generative-AI providers offer machine-readable provenance marking or a public detection tool, verifiable from public price lists and vendor documentation.
AI Confidence: Med by 2030
Generative AI becomes routine behind the scenes, while its visible use in characters, voices and promotion needs consent and fan acceptance on top of legality
Use and acceptance diverge. In CESA's first survey 85.8% of 1,349 Japanese game developers use generative AI at work, while in the GDC 2026 survey 52% called it harmful to the industry, up from 30% a year earlier. On Steam, titles disclosing generative AI were 30.8% of 2026 releases but only 10-27% of estimated sales. Rights are being defined: Japan's Ministry of Justice placed the voice within publicity rights on 7 August 2026, and Anthropic's USD 1.5bn training-data settlement is final.
›Evidence and background
Evidence weight: five layers (ai, games, anime, creator-economy, marketing) carry independent sources: an industry body (CESA), a developer survey (GDC), Steam disclosure data, government (the Ministry of Justice), courts (Anthropic's settlement, GEMA v. Suno) and platform policy (YouTube). YouTube made mass-produced AI content ineligible for monetization and is widening likeness and voice detection. Backlash against visible use now has concrete cases: Level-5 drew criticism in September for AI-generated promotional material. Japan's training-data disclosure moves to comply-or-explain with no penalties, possibly from this autumn. Base rate: the share of Steam releases disclosing generative AI rose from 10.9% in 2024 and 19.9% in 2025 to 30.8% in 2026, and GDC's 'harmful' answer rose from 30% to 52% in one year. Counter-evidence: in one survey only 10% of viewers disliked AI-made ads, against 48% expected by buyers (a single self-reported survey). The G20 favours sector rules over new AI-specific regulators. Exposure: Digital Entertainment's Q1 FY2027 growth came from IP collaborations, where the stance of anime rights holders and voice actors sets the terms, and Monster Strike carries voice-acted characters. MIXI's own postponed project is recorded as a fact about MIXI on this question, not as market evidence.
Why (PESTLE): Generative tools that cut production cost and time, routine use inside studios, clearer rights over voice and likeness, and platform monetization rules. Against them: backlash against visible use, training-data litigation, disclosure duties, and an international preference for sector-by-sector rules.
Implication for MIXI: On 10 September 2026 MIXI announced an AI voice project modelled on a Monster Strike character, built from recordings the voice actor had permitted, and postponed it on 24 September after criticism. Digital Entertainment's growth relies on IP collaborations whose terms rights holders set.
›Evidence and background
MIXI reports that its Runway partnership cut video production from 21 days to 3, so its main use of generative AI is in outward-facing video. Acceptance is a separate test from permission: material made with permission can still be rejected.
Open Resolves 2030 — Confirmed if, by 2030, Japan has binding rules (statute or final judgment) on AI use of voice or likeness AND a majority of large Japanese game companies publish a policy on generative AI in released products. Defeated if titles disclosing generative AI take a share of Steam's estimated sales within 5 points of their share of releases.
Technology Confidence: Med by 2030
Build pipelines, app binaries and the sandboxes agents run in form one attack surface, while compute cost tracks power permits, memory and single-region risk
Securing the build pipeline and the client binary stops being optional. The Sckit worm, reported on 23 September 2026, hid in npm and PyPI releases of an AI memory plugin. Attackers decompiled 1.8 million Android APKs to harvest embedded secrets and reused stolen AI API keys. In August ChainDrop poisoned 444 packages in under four hours. Reporting clocks multiply: EU CRA reporting began on 11 September, and Japan's active-cyber-defense duties took effect on 1 October for 258 designated operators.
›Evidence and background
Evidence weight: three layers (technology, bcp, ai) carry independent sources: a security firm (StepSecurity), a model provider's threat report (Anthropic), Japan's National Police Agency, the European Commission, Cloudflare and affected companies' disclosures. The surface includes the infrastructure agents run on: on 24 September Cloudflare disclosed that newly allocated disk blocks in Containers were not zeroed (it says it has no evidence customer data was compromised). Sakura Internet found intrusion lasting about three years with up to 1.36m accounts possibly affected, and Gyazo exposed about 23.62 million accounts. Cost tracks permits and physical risk: Texas halted data-centre environmental permits, and on 15 September AWS said data held only in its conflict-damaged Bahrain region cannot be restored. Conventional DRAM contract prices rose 58-63% quarter on quarter in Q2 2026, and three US providers hold about 63% of cloud. No public data yet shows that memory price rises pass through to cloud storage prices. Base rate: ransomware has been IPA's top threat for six years running, and Japan's first-half 2026 reports reached a half-year record of 123, with VPN equipment about half of intrusion routes. Counter-evidence: staged publishing at registries, defensive AI and falling token prices on new hardware. Exposure: Monster Strike reportedly runs on a hybrid of on-premise data centres and AWS, and emergency maintenance for server-equipment failures recurred in 2023, 2025 and June 2026. FamilyAlbum stores 30 million users' family photos for years, and PointsBet handles payment and identity data. Japan's amended APPI introduces its first surcharge, with the effective date still unset.
Why (PESTLE): Worms that reach AI tooling, mass harvesting of secrets from shipped apps, dated reporting duties (the EU CRA, Japan's active cyber defense), and power, permits and memory setting compute cost. Against them: staged publishing and human-presence checks at registries, defensive use of AI, and falling cost per token on new hardware.
Implication for MIXI: Monster Strike reportedly runs on a mix of on-premise hardware and AWS, and emergency maintenance for server-equipment failures recurred as recently as June 2026. FamilyAlbum stores family photos for years and PointsBet handles payment and identity data, so secrets in client code and build provenance are concrete questions.
›Evidence and background
Japan's active-cyber-defense duties attach to designated critical-infrastructure operators; firms outside that designation meet them mainly through designated partners and vendors. Under the Cabinet Office draft, a cloud provider's notice counts as recognition for SaaS and some PaaS, so reporting clocks can start from a supplier. The Gyazo case shows that for photo services the metadata generated for search, such as location and OCR text, is itself a leak surface. MIXI does not publish its position on these questions.
Open Resolves 2030 — At least one major package registry makes staged or approval-gated publishing, or provenance verification, the default path for releases AND NIST's final migration guidance keeps a deprecation date for RSA-2048 and ECC P-256 of 2030 or earlier, verifiable from registry documentation and NIST's final publication.
Technology Confidence: Med by 2030
Japan's engineering labour stays short and costly while AI narrows the entry tier and cross-border hiring routes tighten
The domestic shortage persists while the entry point narrows. METI projects Japan's IT-talent gap widening from about 300,000 in 2020 to as much as 800,000 by 2030, and in 2026 67.5% of Japanese firms raised new-graduate starting pay, by a record JPY 9,462 on average. At the US tech majors entry-level hiring is about 65% below 2019, while engineers rose from 46% to 55% of new hires. Cross-border routes tighten: Japan raised its permanent-residency fee on 1 October, reports date new income, language and pension tests to October or April 2027, and Kyodo reports a planned JLPT N2 requirement for new Engineer/Humanities visa entrants; the US renewed its USD 100,000 H-1B requirement by proclamation.
›Evidence and background
Evidence weight: five layers (talent-network, jp-macro, ai, technology, games) carry independent sources: METI, Teikoku Databank, SignalFire, Challenger, the ASGC layoff tracker and the US government. 51.3% of firms report a shortage of regular staff, above 50% for a fourth year. AI-attributed cuts fell to 3,462 in August, the lowest since December 2025, but the year-to-date total is 116,175 and employment of 22-25-year-old developers is down about 20%. Base rate: METI's projection assumes the gap more than doubles in ten years, and the direction of that series has not changed. Counter-evidence: games-industry layoffs reached 9,781 by 26 July 2026, with the year-end forecast raised to 14,259, releasing experienced people. US announced cuts in August were the lowest for the month since 2022, and India's GCCs, 2,117 centres with 2.36M professionals, offer a sourcing option. Exposure: group-wide payroll and capability. MIXI's consolidated headcount reached 2,116 in FY2026 (+23.2%), the increase from the acquisition, and operating profit per employee fell. New-graduate engineer pay of about JPY 300,000 a month sits high in Japan's distribution, but Sony, NEC and DeNA pay up to JPY 10M for AI roles. The next stage named in the CAIO set-up depends on the hardest-to-hire tier of AI-building talent.
Why (PESTLE): An ageing, shrinking workforce, scarcity and pay premiums for AI roles, a narrowing entry tier, and frequent changes to cross-border rules. Against them: experienced people released by games-industry cuts, India's capability centres as a sourcing option, and falling US layoffs.
Implication for MIXI: MIXI's consolidated headcount rose to 2,116, mainly from consolidating PointsBet. New-graduate engineer pay of about JPY 300,000 a month is high in Japan but far below AI-specialist outliers, and the next stage of the CAIO set-up depends on the hardest-to-hire tier.
›Evidence and background
On the betting side, demand rises for trading, risk, compliance, AML and responsible-gaming staff, a demand that exists because operators must limit gambling harm, and it arrives with payroll, systems and penalty exposure. Australia's advertising rules from 1 January 2027 also change the premises for PointsBet's acquisition staff.
Open Resolves 2030 — Defeated if, by 2030, METI or an equivalent official estimate shows Japan's IT-talent gap shrinking below 300,000, or if Teikoku Databank's survey shows fewer than half of firms raising starting pay in two consecutive years.
Geopolitics Confidence: Med by 2030
The 2026 energy shock and synchronized rate hikes end Japan's zero-rate era: a higher floor for yields, a weak and volatile yen, and a narrower listing exit
Rates settle at a level Japan has not seen in three decades. The BoJ raised its policy rate to 1.25% on 18 September 2026, the highest since April 1995, and says it will keep raising; the 10-year JGB briefly reached 3.115% on 25 September, the highest since August 1996. The Fed, the ECB and, on 29 September, the RBA also hiked, Brent is above USD 100, and the yen fell back into the 158s after the hike. Debt service takes a quarter of the FY2027 budget request, with interest up about 30%. The listing exit is narrower: 17 IPOs in the first half of 2026.
›Evidence and background
Evidence weight: seven layers (jp-macro, global-macro, us-macro, eu-macro, au-macro, in-macro, geopolitics) carry independent sources: decisions by the BoJ, the Fed and the ECB, the Finance Ministry's budget requests, Jiji Press market reporting and market data. Government projections put the assumed rate at 4.6% by FY2029, with interest payments reaching JPY 45.2tn by FY2035. Saudi Arabia's East-West pipeline shut on 11 September, so both Gulf exits are contested. Base rate: the 10-year JGB yield was 1.28 points higher on 4 September than a year earlier. IPOs fell from 123 in 2021 to 17 in the first half of 2026, down about 30% on a year earlier, with 11 on the Growth market. Counter-evidence: August core CPI was 1.7%, below 2% for a seventh month, and two board members dissented for a hold. Tokyo's September core CPI, published on 2 October, rose to 2.7%, above 2% for the first time in eight months, as policy effects reversed, and power and gas relief ended with September usage. US and Iranian officials reopened indirect talks during the UN General Assembly, and Iran offered to reopen the strait within seven days if the blockade is lifted. Lower energy prices would weaken the case for hikes. Exposure: Investment segment revenue was JPY 4.44bn in FY2026 (-22.0%), a reaction to prior-year gains on Timee shares, so it depends on the exit market. The rise in net-income guidance to JPY 25.0bn rests on a one-off gain of about JPY 12.2bn from the bitbank share transfer, completed on 1 October, which does not depend on the listing market. PointsBet's Australian- and Canadian-dollar revenue and dollar-denominated cloud and AI costs all move with the yen.
Why (PESTLE): Underlying inflation approaching 2%, synchronized hikes abroad, crude above USD 100 and a swelling debt-service bill. Against them: core CPI below 2%, the chance of cheaper oil from US-Iran talks, and the case for holding if hikes cool the economy.
Implication for MIXI: Investment segment revenue fell 22.0% to JPY 4.44bn in FY2026 and depends on sale gains. The share price was JPY 3,830 on 25 September; the 9.97x forecast PER includes the one-off gain and is about 18.5x on the pre-revision plan. PointsBet revenue and dollar costs move with the yen.
›Evidence and background
The FY2027 plan predates September's rate moves and keeps operating income at JPY 19.5bn, down 12.4%. A higher discount rate lowers the value of holdings and lengthens payback on new projects. Fewer domestic IPOs mean fewer exits and sale gains for the Investment business. MIXI does not disclose its direct import dependence or hedging. In the September Tankan large firms assumed JPY 154.23 per dollar for FY2026, about JPY 4 stronger than JPY 158.22 on 1 October, and the MOF reported no intervention from 27 August to 28 September. PointsBet's revenue translates through the Australian and Canadian dollars against the yen, which official action on dollar-yen does not target.
Open Resolves 2030 — Confirmed if in 2030 the BoJ policy rate is at or above 1.0% and the 10-year JGB yield averages above 2% for the year. Defeated if the BoJ returns the policy rate to 0.5% or below before 2030, verifiable from BoJ releases and market data.
An AI synthesis of public-information PESTLE analysis into forward scenarios — not deterministic predictions.