Skip to content

Foresight · Next 5 Years

An AI-synthesized 5-year outlook derived from every PESTLE layer. Scenarios, not certainties.

Horizon: 2026–2031· Updated 2026-06-20· 31 layers

Big picture

Over the next five years MIXI operates inside six structural shifts running at once. (1) Collapsing inference cost makes AI-native production/ops the norm, letting small teams run large live-ops. (2) Gacha, betting, child-safety and AI regulation tighten globally in parallel, turning compliance into both a moat and a drag. (3) Japan's population, real wages and share-of-wallet shrink, pushing growth offshore (FamilyAlbum in North America/India, PointsBet in Canada, Strike World in India) and into sports/betting. (4) The web migrates to a post-app, post-search, agentic internet that rewrites discovery, checkout and monetization. (5) Taiwan-contingency, chip/rare-earth and energy fragmentation become a permanent cost/BCP bias. (6) Japan's own tail risks — the Nankai Trough, a Tokyo inland quake and a Mt-Fuji eruption — become preconditions for business continuity. The winning posture: compress unit cost with AI, diversify revenue across IP, sports, overseas and family, and convert regulation and geopolitics into a designed-in advantage.

AI Confidence: High by 2028

Collapsing inference cost makes AI-native production standard; small teams run large operations

Generative AI and agents for image, voice, video, 3D and code become permanent infrastructure across production, live-ops, marketing and support. A multi-vendor frontier race (Claude Opus 4.8, GPT-5.5, Gemini 3.1, Grok 4.3) and a 90%+ drop in API pricing move AI from 'efficiency' toward 'new-category creation.'

Why (PESTLE): Structural inference-cost decline, maturing per-modality capability, the spread of AI-assisted dev, and MIXI's own proof point — a 98% work-hour cut on a Monster Strike project, with AI use built into its evaluation system.

Implication for MIXI: Scale the 'AI built into the evaluation system' advantage — reinvest the 98%-effort-cut proven on a Monster Strike feature into new IP, new experiences and overseas localization, and use domestic open models (e.g. Rakuten AI 3.0) plus local compute to cut content cost.

Open Resolves 2028 — Frontier-model API price-per-token is down >=90% vs the 2023 baseline AND generative AI/agents are embedded as standing workflows across MIXI production, live-ops, marketing and support (verifiable across multiple shipped initiatives).
Games Confidence: High by 2029

Synchronized gacha regulation and store-fee redistribution rewrite the monetization model

Odds disclosure, spend caps, age checks and loot-box ratings (EU Digital Fairness Act, PEGI, Australia R18+) standardize, pressuring pure-gacha models; meanwhile the three-bloc fee regime (Japan's MSCA live Dec 2025, the US Epic ruling, EU DMA) opens external payments so web shops/D2C reclaim the ~15–30% store fee. With Monster Strike at $11.4B lifetime but under a third of peak and ~99% Japan-dependent, de-concentration is unavoidable.

Why (PESTLE): Regulation in AU/EU/Japan, the fluid platform policy around Epic v. Apple/Google, and rising social pressure for odds disclosure.

Implication for MIXI: Design odds disclosure, caps and minor protection in as a moat; scale the Monster Strike Web Shop into a group-wide D2C billing layer and validate the net fee upside to recover margin.

Open Resolves 2029 — External (out-of-store) payments are operationally permitted on major stores across Japan (MSCA), the US and the EU (DMA) AND the Monster Strike Web Shop runs as a group-wide D2C billing layer, with odds disclosure, spend caps and age checks implemented as standard.
Betting Confidence: Med by 2030

In betting, where regulation sets enterprise value, prediction markets and regional divergence redraw the revenue map

US CFTC-regulated prediction markets (Kalshi/Polymarket, monthly handle surging toward $24B) cement as a license-bypassing force, Australia's strictest-in-world ad rules go live in 2027, Japan's public monopoly (keirin/toto) persists against an estimated ¥6.5tn illegal offshore market, and India's blanket real-money-gaming ban hardens. PointsBet's growth weight shifts from Australia to Canadian multi-province iGaming.

Why (PESTLE): US prediction-market regulatory arbitrage, Australia's 2027 ad ban, AUSTRAC AML enforcement, Japan's public-competition tailwind (TIPSTAR revenue +55.8%) and Canadian iGaming (Ontario net revenue +58% half-on-half).

Implication for MIXI: Grow PointsBet but re-engineer acquisition for the 2027 ad ban; build responsible-gambling, age/payments and harm-detection tech as a cross-business moat across TIPSTAR and PointsBet, and pre-spec the prediction-market 'games x betting' fusion from JP/AU bases.

Open Resolves 2030 — US CFTC-regulated prediction markets (Kalshi/Polymarket et al.) reach >=$24B monthly handle, Australia's ad rules took effect in 2027, AND PointsBet's growth weight has shifted to Canadian multi-province iGaming (Australia's revenue share declines).
Sports Confidence: High by 2029

Arena economics, women's sports and broadcast-betting fusion cement sports as the second pillar

A revenue machine fusing media rights (NBA's 11-yr $76bn deal, F1's Apple exclusive, the 2026 FIFA World Cup), sponsorship, arenas, fan data and betting tie-ins expands. In Japan, B.LEAGUE PREMIER's launch plus an arena boom and J.League's autumn-spring shift run in parallel. MIXI's Sports segment grew to ¥65.8bn (+63.8%) in FY2026 as a clear second pillar.

Why (PESTLE): Chiba Jets' routinely sold-out LaLa arena, women's sports growing ~3x faster (>$3bn in 2026), and broadcast-data x betting integration (Sportradar/Genius).

Implication for MIXI: Make owned-club arena economics (Chiba Jets/FC Tokyo) a profit core, couple TIPSTAR/PointsBet with live broadcast data, evaluate early women's-sports exposure, and factor the autumn-spring shift into FC Tokyo planning.

Open Resolves 2029 — MIXI's Sports segment revenue holds at or above the FY2026 level of ¥65.8bn and remains the largest non-games revenue segment (the second pillar).
Family Confidence: High by 2030

Amid a global child-safety wave, the 'privacy-safe family lifecycle' becomes the differentiator

Age-assurance regimes (COPPA reform, GDPR-K, UK OSA, Australia's under-16 ban, India's DPDP, US state laws, the SCOTUS Paxton ruling) tighten worldwide at once. FamilyAlbum (30M users, 175 countries, 40% overseas) turns this to advantage as the 'private, never-AI-trained family app' and extends into a family-lifecycle stack: photo sharing → child GPS → telemedicine → eldercare.

Why (PESTLE): Rising global child-safety anxiety, domestic saturation under Japan's 1.14 birthrate (65%+ FamilyAlbum penetration), high-birthrate overseas markets (India, North America), and privacy-conscious on-device AI.

Implication for MIXI: Map FamilyAlbum against every age-assurance regime once and reuse globally; reconcile safety and monetization (premium, printing, goods) with on-device AI that explicitly 'doesn't train on your photos,' and make North America and India the flagship demographic hedge.

Open Resolves 2030 — FamilyAlbum's overseas user share holds at >=40% AND at least one family-lifecycle feature beyond photo sharing (child GPS, telemedicine, or eldercare) is officially launched.
Internet Confidence: Med by 2030

A post-app, post-search agentic internet rewrites discovery, checkout and monetization

An 'agentic web' emerges where AI browsers and agents search, compare and transact on a user's behalf, with interop protocols (Anthropic's MCP, Google's A2A, the Linux Foundation's Agentic AI Foundation) and agent payments (ACP/AP2, Visa rails) standardizing. Discovery shifts from app stores and SEO toward LLM recommendation and agent funnels, demanding a new identity/consent/payment layer — and structurally destabilizing ad/SEO/ASO-dependent acquisition.

Why (PESTLE): MCP/A2A agent-infra standardization, ACP/AP2 agent payments, TikTok's MCP server (ChatGPT/Claude), on-device/ambient AI enabled by cheap inference, and LLM discoverability becoming the top 2026 marketing issue.

Implication for MIXI: Grow mixi2 as an owned attention graph and identity rail to hedge distribution dependence (LINE/Meta/Google/stores); make FamilyAlbum, Monster Strike and sports reachable and transactable by agents via MCP, and get ahead on LLM-recommendation optimization (AI discoverability) and agent payments.

Open Resolves 2030 — An agent-payment protocol (ACP/AP2 with Visa rails) is standardized in production AND at least one major MIXI product (FamilyAlbum, Monster Strike, or sports) is reachable and transactable by agents via MCP or equivalent.
Experiences Confidence: Med by 2029

AI companion/character experiences go mainstream as a new 'relationship' category

AI companions/character chat — typified by Character.AI's 200M+ monthly visits and ~20M MAU — build a daily-use base in the tens of millions, and AI-character dialogue becomes a standing feature inside games, social and family apps. Japan's character culture x VTuber demand spawns a home-grown AI-character layer, while generative world-model/4D tools (Genie 3, Roblox 4D) slash UGC and live-ops production barriers.

Why (PESTLE): Mainstreaming of AI companions, Japan's character/VTuber demand, asset-cost collapse from world-model/4D generation, and cheap inference.

Implication for MIXI: Build AI companions/character chat into Monster Strike to capture relationship demand, construct a home-grown AI-character layer tied to mixi2 and family experiences, and adopt generative/world-model tooling in-house to accelerate UGC and live-ops — with safety, age-care and disclosure designed in.

Open Resolves 2029 — Leading AI-companion/character-chat services collectively reach a daily-use base on the order of 50M MAU AND AI-character dialogue ships as a standing feature inside a MIXI game (e.g. Monster Strike).
Experiences Confidence: Low by 2031

Smart glasses, spatial/AR, wearables and ambient AI emerge as the next-gen interface

Led by Meta smart glasses (7M+ in 2025, ~13.4M expected in 2026), smart glasses, spatial computing and wearables climb the adoption curve, and shoppable livestream, second-screen and on-device/ambient AI become new touchpoints for games, sports viewing and family experiences — with robotics/embodied AI and BCI frontiers moving past the experimental stage.

Why (PESTLE): Surging smart-glasses shipments, shoppable-stream (TikTok Shop) and second-screen demand, maturing on-device/ambient AI, and evolving spatial/4D generated experiences.

Implication for MIXI: Watch and small-scale-pilot smart glasses as a next-gen interface for Monster Strike and sports viewing, and integrate shoppable livestream and second-screen into TIPSTAR/sports live experiences — deferring large bets until adoption signals firm up.

Open Resolves 2031 — Annual global smart-glasses shipments reach >=30M units AND MIXI has run at least a pilot of a smart-glasses or second-screen experience for Monster Strike or sports viewing.
Consumers Confidence: Med by 2029

The fight for time and wallet intensifies; Japan's shrinking household squeeze exposes the ceiling on high-ARPU monetization

Global app usage hits 5.3T hours and 3.6h/day, with social (2.5T hrs) dwarfing games; MIXI competes for time against TikTok, YouTube and short video. In Japan, a weak yen and inflation stall real wages and shrink household spend, and Digital Entertainment falls -13.1%. The ceiling on high-ARPU gacha is exposed, and the share-of-wallet contest moves to betting, sports and overseas.

Why (PESTLE): Time concentrating in social/short video, Japan's stalled real wages and -3.0% household spend, subscription fatigue (Australia), and India's ultra-low-ARPU but vast time market.

Implication for MIXI: Reframe Monster Strike's rivals as TikTok/YouTube for time, make betting the engine of share-of-wallet capture, watch Japan's household squeeze as a high-ARPU risk, and position India as a time-first, money-later market via ads plus small UPI payments.

Open Resolves 2029 — Global app usage reaches ~5.3T hours/year and ~3.6h/day with social/short-video exceeding games in time share AND Japan's digital-entertainment market posts a year-over-year decline (high-ARPU gacha plateau).
Geopolitics Confidence: Med by 2031

Taiwan, chip, rare-earth and energy fragmentation become a permanent cost/supply tail risk

US-China chip/rare-earth rivalry (rare-earth flows to Japan down 80%), the Taiwan escalation ladder, AUKUS/Quad-driven Indo-Pacific realignment and energy fragmentation become structural, normalizing an upward bias on cloud/AI-hardware, component and power-linked costs under a weak yen. Export controls on cloud API access (the June 2026 Anthropic order) make model access itself a geopolitical risk.

Why (PESTLE): Structuralized US-China export controls, rare-earth restrictions on Japan, Taiwan-dependent compute, the active-cyber-defense law (Oct 2026), and energy/FX fragmentation.

Implication for MIXI: Diversify cloud regions and hedge de-China component/compute exposure, build a model-access BCP (multi-provider with a domestic fallback), price energy/FX fragmentation into COGS hedging, and prepare for flashpoint ad-sentiment.

Open Resolves 2031 — US-China chip/rare-earth export controls remain effectively in force AND export controls on cloud-API model access (the June 2026 Anthropic order lineage) have not been rescinded (model access remains a geopolitical risk).
Technology Confidence: High by 2030

On-device AI, post-quantum crypto and software-supply-chain resilience become baseline tech requirements

Mainstream on-device AI, collapsing inference cost, game-engine repricing/generational shifts (Unity pricing vs Unreal 6 vs WebGPU), Apache Iceberg standardization, post-quantum migration and security regulation (EU CRA) all advance at once. As npm supply-chain attacks show, SBOM, pinned deps, signed provenance and least-privilege become preconditions for continuity.

Why (PESTLE): Sharp inference-cost decline, maturing on-device stacks (Apple/Android), PQC migration mandates, the EU CRA/supply-chain rules, and engine-tech flux.

Implication for MIXI: Adopt on-device AI for FamilyAlbum's image/child-data processing, build a PQC migration plan for long-lived assets (FamilyAlbum photos, PointsBet payments/KYC), harden the supply chain with SBOM/pinned deps/signed provenance, re-evaluate engine strategy, and budget compute conservatively against Japan's DC power constraints.

Open Resolves 2030 — Security regulation such as the EU CRA is in force AND MIXI operates SBOM, pinned dependencies and signed provenance in production and has a documented post-quantum (PQC) migration plan for long-lived assets (FamilyAlbum photos, PointsBet payments/KYC).
Technology Confidence: Med by 2030

AI/ML talent becomes the hardest hire worldwide; India GCCs and inbound foreign talent are the route around domestic scarcity

AI/ML engineers become the single hardest role to hire globally; Japan faces a shortfall of up to 800,000 IT workers by 2030 alongside record new-grad pay. The US $100K H-1B fee reshapes talent flows and big tech sheds jobs to AI substitution (52,050 in Q1 2026), while India's 2,117 GCCs (2.36M professionals), J-Skip/J-Find and Australia's Skills-in-Demand visa open alternative routes.

Why (PESTLE): Global AI/ML scarcity, Japan's IT shortfall and surging new-grad pay, the US H-1B fee, India's GCC expansion, and big tech halving new-grad/junior hiring.

Implication for MIXI: Secure AI/data talent via India GCCs/remote to bypass domestic scarcity, re-baseline comp for AI/data/live-ops roles against Japan benchmarks, expand inbound hiring via J-Skip/J-Find and India university pipelines, and lock in PointsBet/TIPSTAR trading, risk and compliance talent ahead of need.

Open Resolves 2030 — Japan's IT-worker shortfall is confirmed by an official estimate to reach the hundreds of thousands (up to ~800,000) AND MIXI is actually sourcing AI/data talent via India GCCs/remote or J-Skip/J-Find-type routes.
Consumers Confidence: Med by 2030

India becomes the demographic-dividend flagship, attacked regulation-first and time-first

India — median age 28.8, 1.46bn people, UPI/Aadhaar digital public infrastructure — becomes the largest growth frontier, even as a blanket (criminally enforced) real-money-gaming ban and phased DPDP rollout (consent managers ~Nov 2026 → core duties ~May 2027) gate entry. Strike World (full launch Apr 2026) and FamilyAlbum build a foothold via F2P + ads + small UPI payments and local-language localization.

Why (PESTLE): India's #1 app downloads but ultra-low ARPU, the F2P/esports vacuum from the RMG ban, phased DPDP, esports' national-sport status (Krafton's $200M bet), and a deep family-forming cohort.

Implication for MIXI: Keep Strike World strictly F2P to capture the cleared market, target the young-family TAM with FamilyAlbum riding the WhatsApp + UPI + Reels/Shorts rails, run India-first co-dev from Bengaluru/Hyderabad/Pune, and bake DPDP and AI-labeling into pre-launch.

Open Resolves 2030 — India's DPDP core duties are in force AND Strike World (full launch Apr 2026) remains live in India as a strictly F2P title, with Strike World or FamilyAlbum having established a local user base (MAU).
Internet Confidence: Low by 2031

Stablecoin rails and agent payments become the new base layer for cash flows and fan experiences

After the speculative Web3-gaming collapse (93% of projects effectively dead, funding down 99%), 2026 begins a realistic rebuild toward fun-first, sustainable economies and stablecoin payments. Stablecoin supply exceeds $320bn, the US GENIUS Act, EU MiCA and Japan's stablecoin framework firm up, and functional NFTs/IP-utility plus USDC prediction markets become an adjacent frontier to the core betting business.

Why (PESTLE): Stablecoin-market growth plus GENIUS/MiCA/Japan's PSA, the Web3 token collapse and rebuild, agent payments (ACP/AP2), and USDC prediction markets (Polymarket/Kalshi).

Implication for MIXI: Evaluate stablecoin rails (JPYC, GENIUS-compliant) for TIPSTAR/PointsBet deposits and payouts, re-design sports-IP × functional NFTs as fan experience/betting tie-in rather than speculation, stay cautious about bolting tokens onto existing games, and monitor Japan's FIEA reclassification, tax and PSA enforcement.

Open Resolves 2031 — Total stablecoin supply holds at >=$320bn, the US GENIUS Act, EU MiCA and Japan's stablecoin framework are in force, AND MIXI has evaluated or adopted stablecoin rails (e.g. JPYC) for TIPSTAR/PointsBet deposits and payouts.
Consumers Confidence: Med by 2029

The export boom in Japanese IP, anime and the creator economy becomes a tailwind for MIXI's IP and community assets

The anime/manga IP market industrializes into an export business (record ¥3.84tn, 56% overseas), with the government targeting ¥20tn in content exports by 2033, while the ~$250B creator economy is reshaped by VTubers, UGC games, live commerce and AI creators. MIXI can connect Monster Strike/STRIKE WORLD anime IP, popular-IP collabs and its community DNA to the overseas boom and fan economy as a growth axis.

Why (PESTLE): Anime overseas revenue overtaking domestic, the government's ¥20tn Cool Japan target, high-ARPU VTuber fan economies (Cover/ANYCOLOR), UGC games (Roblox/UEFN), and India's anime fandom.

Implication for MIXI: Scale STRIKE WORLD anime IP onto the overseas boom and make popular-IP collabs a standing UA/monetization engine, elevate VTuber/creator collabs to a standing program, convert community heritage into co-creation products, and confine generative AI to non-creative work to avoid brand damage.

Open Resolves 2029 — The anime/manga IP market's overseas revenue share stays above 50% (not falling below the recent ¥3.84tn / 56%-overseas level) AND MIXI has scaled STRIKE WORLD anime IP or popular-IP collabs into a standing UA/monetization engine.

Bold scenarios (if AI keeps compounding)

Higher-variance, aggressive takes for an AI curve that accelerates faster than expected — possible upside, not the base case.

BOLD AI ~2028

Agentic AI does the majority of digital knowledge work as capability doubles on short cadences

This is NOT the base case — it is a deliberately bullish read of what happens if AI's compounding does not stall. At mid-2026 a Claude Opus 4.8-class model is the accessible frontier, inference cost has fallen 90%+ since 2023, MCP/A2A have standardized agent infrastructure, and Claude Code / Cursor have shifted dev from 'writing code' to 'delegating goals.' If capability keeps doubling on a months-long cadence, by 2028 agents autonomously run the majority of digital knowledge work — planning, implementation, QA, live-ops, support and marketing — and humans retreat to the decision, taste and accountability layer. The wall Gartner flagged ('88% of PoCs never ship') dissolves once doublings continue. One producer orchestrates dozens of agents to run what used to take a 100-person studio.

Implication for MIXI: Upside: MIXI already proved a 98% work-hour cut on a Monster Strike feature and built AI use into its evaluation system, giving it the best initial conditions to move first to 'agent-native operations.' Threat: if every rival wields the same lever, the cost advantage equalizes overnight and the contest shifts to non-replicable assets — IP, community, first-party data and brand trust. Pre-position: make an internal agent-orchestration platform (with permissions, review and security gates) the standard OS across all studios, make FamilyAlbum / Monster Strike / PointsBet reachable and operable by agents via MCP, and re-organize so humans concentrate on the 'what to build' decision.

BOLD Games ~2029

World models birth infinitely-replayable, on-the-fly games and engine-less production goes mainstream

A bullish scenario. Genie 3 generated controllable 3D worlds from text for minutes in Jan 2026, World Labs' Marble exports physics-ready meshes, and Roblox 4D creation makes objects with physical properties instantly. If that curve holds, by 2029 games where the world, enemies and story are generated fresh every session — never the same experience twice — become commercially viable. Most level design and asset production is replaced by run-time world-model generation, and the fixed-engine + handmade-asset mode retreats to a 'premium artisanal' category. Existing monetization that leans on the scarcity of gacha contents must be redefined when content is infinite.

Implication for MIXI: Upside: generating Monster Strike boards, raid bosses and collab content at run-time structurally erases live-ops production cost and supplies unique quests daily. Threat: infinite generation breaks 'scarcity = reason to pay,' hitting the pure-gacha revenue base directly. Pre-position: fold world-model / 4D tooling into the in-house pipeline (per the Sora-shutdown lesson, tie it to IP, human finishing and human participation rather than 'AI-only'), and shift monetization toward character/nurture/community belonging. In FamilyAlbum, explore IP-independent generative experiences such as improvising a 'storybook world made only for your child' from family photos.

BOLD Games ~2028

Near-zero marginal cost for AAA-quality content spawns swarms of few-person AI-native studios

A bullish scenario. Video runs $0.10–0.15/sec (Veo-class), images cost cents, 3D takes seconds, ElevenLabs is the voice standard, and agents generate the majority of code. If per-modality costs keep falling, by 2028 the marginal cost of one more unit of AAA-grade visuals, audio and narrative approaches zero. As a result, 3–5-person 'AI-native studios' mass-produce titles that once took hundreds of people, years and tens of millions of dollars — in months — and proliferate explosively worldwide. Stores saturate from oversupply, and the contest shifts entirely from 'being able to build' to 'being discovered, loved and trusted.'

Implication for MIXI: Upside: MIXI has the early proof of small-team/large-output (the 98% cut) and already holds the 'loved/trusted' moat via IP and character assets; it can run several in-house AI-native studios and a high-volume, high-mortality portfolio betting on many new IPs a year. Threat: in a zero-marginal-cost world, one hit from an unknown studio can drain Monster Strike's time overnight, and the supply explosion wrecks UA pricing and store discoverability. Pre-position: harden not raw production volume but mixi2's attention graph, community, IP and first-party data as the discovery/trust moat, and make AI discoverability (LLM-recommendation optimization) and agent funnels the top priority.

BOLD Experiences ~2028

Always-on multimodal AI companions become the primary surface for relationship and entertainment

A bullish scenario. Character.AI draws 200M+ monthly visits and ~20M MAU, ~50M people used AI companions daily as of Feb 2026, and ~40% of Japanese teens/twenty-somethings say they can emotionally connect with generative AI. If cheap inference, on-device AI and voice/video-native models keep advancing, by 2028 multimodal AI characters with voice, expression, memory and always-on context overtake social, video and games as the surface youth spend the most time with for relationship and entertainment. People live with, confide in and play alongside characters — a shift from 'watching' a feed to 'spending time with' a character.

Implication for MIXI: Upside: MIXI already holds three 'relationship' anchors — Monster Strike characters, mixi2's social graph and FamilyAlbum's family context — and is best positioned to win a 'safe, home-grown AI-character layer' differentiated from NSFW lines like xAI's Ani. Threat: if the relationship surface is captured by Character.AI or overseas platforms, youth attention drains wholesale and Monster Strike/SNS retention collapses. Pre-position: build AI companions into Monster Strike, run low-latency high-privacy character chat on on-device inference, and bake in CA SB243 / NY S9051-class safety, age and disclosure from the design stage; put a 'safe AI that recalls family memories' on FamilyAlbum and a 'responsible spectating companion' on betting.

BOLD Physical AI ~2030

Humanoids, embodied AI and robot companions enter the home

A bullish scenario. The 1X NEO is on pre-order at $20k / $499 per month, Figure 03 hits one robot per hour at BotQ, robot foundation models (GR00T N1, Gemini Robotics On-Device) provide the 'brain' as software, and the AI plush-toy market grows ~52% CAGR. If price declines and model generalization continue, by 2030 home humanoids and embodied AI companions become reality in a slice of advanced-economy homes. Japan's working-age population shrinks 12M by 2040, structuralizing embodied-AI demand for monitoring, eldercare and chores — AI characters standing off the screen, in the living room.

Implication for MIXI: Upside: MIXI's edge is not the machine but engaging characters, voice, personality and IP, so it can enter as an 'embodiment layer' atop sourced robot brains; FamilyAlbum's family/monitoring/eldercare sits right next to embodied AI. Threat: components are China-dependent (94% of permanent magnets), so moving into hardware imports the entire supply-geopolitics risk. Pre-position: stay in the model/software/character layer rather than build hardware, line up partnerships to put Monster Strike / FamilyAlbum character personas onto the 1X NEO, AI toys and smart glasses, design around child-safety (FTC-probe-class) and privacy, and bake supply-disruption BCP in from day one.

BOLD Internet ~2028

Real-time AI dubbing and localization collapse the language barrier to global distribution overnight

A bullish scenario. ElevenLabs offers ~48kHz natural multilingual synthesis and low-latency conversation (adopted for Spotify's AI-voiced audiobooks), Veo-class models generate video-native audio, and Japanese-specialized open models like Rakuten AI 3.0 have arrived. If dubbing quality — matching voice, lip-sync and cultural context — keeps rising, by 2028 games, character dialogue, livestreams and family content can ship to every world language in real time at near-zero added cost. The 'make it in Japanese, sell it worldwide' localization step becomes automatic and concurrent with planning. Content borders effectively vanish.

Implication for MIXI: Upside: FamilyAlbum is already in 175 countries (40% overseas) and Monster Strike / STRIKE WORLD can ride the export boom in Japanese IP (anime market ¥3.84tn, 56% overseas). With real-time dubbing, MIXI can ship home-grown IP, character dialogue and livestreams in many languages at once and maximize the tailwind of the government's ¥20tn-by-2033 content-export target. Threat: once the barrier falls, overseas mega-IP also pours in flawlessly localized into Japanese, intensifying the fight for the home core. Pre-position: build AI dubbing into FamilyAlbum's overseas localization and Monster Strike/sports multilingual live-ops, clear voice-actor consent and voice rights (protected identity) at the design stage, and stand up a 'day-one, all-languages, worldwide' release capability for home-grown IP.

BOLD Experiences ~2029

Personalized AI-generated entertainment replaces a slice of broadcast and feed consumption

A bullish scenario. Global app usage is 5.3T hours and 3.6h/day with social/short-video dominating time, and Sora's shutdown exposed the thin demand for an 'AI-only feed.' But if video/world-model generation quality and personalization compound and fuse with human IP and interaction, by 2029 'shows, stories and live content generated for you, in the moment' will measurably eat into time spent on legacy broadcast and algorithmic feeds. The shift is from passively watching what streams in to experiencing what is generated for your context — changing the very arena of the fight for attention.

Implication for MIXI: Upside: MIXI holds the raw materials for personalized generation — Monster Strike IP, sports broadcast/data and FamilyAlbum's family data. Think AI commentary and highlights optimized per viewer, a personal show or storybook built from a child's growth in FamilyAlbum, or a bespoke Monster Strike quest program — replacing passive consumption with active generative experience. Threat: if personalized AI entertainment is owned by a platform, both attention and data get disintermediated. Pre-position: build personalized generation in-house tied to human IP, community and first-party data (the Sora lesson: AI-only doesn't last), grow mixi2 as the attention/identity rail for generative entertainment, and wean UA off broadcast/feed dependence.

BOLD Internet ~2029

Agentic AI drives prediction markets and sports betting, and AI companions become the betting surface

A bullish scenario. US CFTC-regulated prediction markets (Kalshi/Polymarket) are surging toward $24B monthly handle, broadcast-data x betting integration (Sportradar/Genius) is advancing, and agent payments (ACP/AP2) plus agent browsers are starting to own discovery and checkout. If AI capability keeps compounding, by 2029 personal AI agents go mainstream as 'betting companions' that combine real-time odds analysis, automated bets and responsible-limit management, moving the entry point of betting from apps to chat/agents. Fused with the prediction-market 'games x betting' convergence, the betting experience itself becomes conversational and always-on.

Implication for MIXI: Upside: MIXI uniquely combines PointsBet (Australia/Canada), TIPSTAR, the broadcast and fan data of Chiba Jets / FC Tokyo, and character/companion technology — a rare position to build a 'safe, responsible betting companion' in-house and pre-spec a conversational, always-on betting experience from its JP/AU bases. Threat: if the betting entry point shifts to agents, the distribution surface and brand are owned by outsiders, and harm, compliance and age/payment risks amplify. Pre-position: build responsible-gambling, harm-detection and age/payment tech as a cross-business moat across TIPSTAR and PointsBet, develop an AI-commentary/betting companion fused with broadcast data using on-device and safety-by-design, and watch prediction-market regulatory arbitrage while piloting conversational-bet UX early.

Wildcards (low-probability, high-impact)

The Nankai Trough, a Tokyo inland quake, or a Mt-Fuji eruption

low-probability, very-high-impact

An upgraded Nankai Trough probability (60–90%), a Tokyo inland quake (up to 18,000 deaths, ¥83tn in damage, up to ~8.4M stranded commuters), or Mt-Fuji ashfall could hit MIXI's HQ at Shibuya Scramble Square and its game infrastructure directly. Monster Strike (on-prem + AWS), FamilyAlbum and public betting all assume 24/7 uptime; a prolonged outage cascades into revenue, trust and regulatory-reporting failures. Geographic dispersion, graceful degradation, disaster-mode operations and pre-staged inventory/power become continuity preconditions.

An AI copyright/regulation shock (training-data litigation, labeling/high-risk duties, model-access cutoff)

medium-probability, high-impact

Training-data-provenance litigation (Anthropic's $1.5bn settlement, Universal Music's ~$3bn suit) spills onto downstream users, while the EU AI Act's GPAI enforcement and heavy fines (Aug 2026), labeling duties (Dec 2026) and high-risk rules (Aug 2027) converge. On top, cloud-API model access becomes export-controlled (the June 2026 Anthropic order), so a top-tier model can vanish overnight — flipping the cost and legal basis of AI-dependent production, marketing and character experiences in an instant.

A Taiwan contingency / semiconductor supply shock

low-probability, very-high-impact

Escalating military tension over Taiwan would abruptly tighten Taiwan-fab compute and component supply, spiking cloud COGS, AI hardware and device costs. Combined with rare-earth flows to Japan down 80% and a Hormuz disruption (17% of Qatari LNG lost), it becomes a twin sourcing-and-FX shock under a weak yen — destabilizing MIXI's infrastructure planning, compute budget and ad sentiment at once.

An AI-capex bubble correction with higher-for-longer rates

medium-probability, high-impact

With a ~$700B AI-capex boom colliding with a 1.0% BOJ policy rate, a 30-year-high 2.8% 10-year JGB and a Fed that won't turn dovish, a break in AI-return expectations would bring rising capital cost and a valuation correction together. MIXI's Investment-segment marks, M&A (PointsBet amortization), FX and inference/power-cost assumptions all come under stress — making conservative payback assumptions and budget stress-tests essential.

An AI synthesis of public-information PESTLE analysis into forward scenarios — not deterministic predictions.