Koei Tecmo Holdings (TSE:3635) is trading at half its five-year peak. The stock sits near its year-to-date low at around ¥1,458, with a consensus of “Neutral” from nine analysts. The market has moved on to defense contractors and semiconductor equipment makers. Nobody is watching this one.
That might be a mistake.
The Numbers That Should Not Be Ignored
| Metric | FY2025 | FY2026 Guidance |
|---|---|---|
| Revenue | JPY 88.4bn | JPY 90.0bn (+1.8%) |
| Operating Profit | JPY 37.2bn | JPY 32.0bn (-13.9%) |
| Operating Margin | 42.0% | ~35.6% |
| Net Profit | JPY 42.8bn | JPY 31.0bn (-27.6%) |
| Dividend per Share | ¥66 | ¥48 (forecast) |
| EPS | ¥131.77 | ¥95.38 (forecast) |
A 42% operating margin is exceptional by any standard. Capcom — widely considered Japan’s most profitable game developer — operates at 30–35%. Koei Tecmo’s margin is built on something older and more durable: intellectual property that has been fully amortized over four decades.
What This Company Actually Is
Koei Tecmo is not simply a game publisher. It is a hybrid: a game studio layered on top of what functions as a corporate investment fund.
Founder and chairperson Keiko Erikawa personally oversees a financial portfolio of approximately ¥150 billion, concentrated in US tech equities — GAFA, AI, cloud, and cybersecurity — alongside structured bonds generating roughly ¥14 billion in annual interest income. The company even established a dedicated financial subsidiary, Koei Tecmo Corporate Finance, in February 2025.
The result is a P&L that requires careful reading:
- FY2025 operating profit: ¥37.2bn (game business)
- Non-operating income: ¥39.9bn gross, but ¥20.1bn in losses on securities sales
- Net non-operating: ~¥19.8bn added to ordinary income
In plain terms: FY2025 ordinary income of ¥57.0bn overstates game business performance. The underlying game engine generates roughly ¥37bn annually, with investment returns adding or subtracting on top. The ¥17.8bn in realized securities losses in FY2025 suggests the company was cutting losing positions — using a strong game year as cover to clean up the portfolio.
The Win Scenario: AI That Nobody Has Thought Through
The market narrative on game stocks is straightforward: post-COVID demand normalization, mobile saturation, Switch 2 platform transition uncertainty. All correct. All already priced in.
What is not priced in is Koei Tecmo’s structural advantage in the one area where AI could most immediately transform gaming: historical simulation.
Consider what the company already has:
Every warlord in Nobunaga’s Ambition and Romance of the Three Kingdoms is defined by dozens of parameters — intelligence, warfare, charisma, loyalty, personality archetype, factional relationships. These parameters have been refined across 40 years of releases. They are, in effect, pre-written system prompts for a large language model.
Connect those parameters to an LLM, and Tokugawa Ieyasu reasons about whether to betray an alliance based on the actual game state — not a probability table. Oda Nobunaga responds differently to the same diplomatic proposal depending on how his campaign is going. Every playthrough becomes genuinely unique.
The implementation barrier is lower here than at almost any other game company. The character data exists. The historical text corpus exists. The game logic already models personality-driven decision-making. AI does not need to replace the system — it needs to inhabit it.
Add AI voice synthesis using decades of existing voice actor libraries, and the production cost of making this fully voiced drops dramatically.
This would not be an incremental improvement to existing titles. It would make Nobunaga’s Ambition a fundamentally different game — one where replay value becomes effectively infinite, and where the 40–60 year old lapsed player base has a genuine reason to return.
The Loss Scenario: Concentrated in the Wrong Direction
The risk is structural and specific.
Koei Tecmo’s investment portfolio is concentrated in the same sector as its business: technology. GAFA equities, AI-adjacent companies, cloud infrastructure, structured bonds with complex derivative exposure. When tech sentiment is positive, both the game business and the portfolio perform. When tech sentiment turns, they fall together.
There is no real hedge here. No heavy industry, no energy, no utilities, no commodities. The portfolio that is meant to buffer game business volatility is correlated with it.
The worst-case scenario: an AI bubble correction hits US tech equities sharply, game market sentiment softens alongside it, and structured bond positions face mark-to-market pressure from rate movements — simultaneously. The investment buffer disappears at the same moment the core business weakens.
This is not a base case. But it is the only scenario that makes Koei Tecmo genuinely dangerous to hold, and it is worth understanding clearly.
The Race
The bull case and the bear case share a timeline: both depend on what happens to AI in the next 12–24 months.
If AI valuations hold and Koei Tecmo announces an AI-enhanced simulation title — even a preview — the stock re-rates. An “AI Nobunaga” announcement from a company with the right IP at the right moment would attract attention that no amount of earnings guidance can generate.
If AI valuations correct sharply before that announcement comes, the investment portfolio takes the first blow, ordinary income collapses, and the narrative shifts to “game company with investment losses” rather than “hidden AI play.”
At ¥1,458 per share — 11x forward earnings, 3.3% dividend yield, 42% operating margins in the core business — the question is simply which arrives first.
The answer is not obvious. That is what makes it interesting.
One More Thing: There Is an Audience That Has Been Waiting Thirty Years
Set aside the investment thesis for a moment.
The players who most want AI-driven historical simulation are the ones who already love this genre. They have been playing Nobunaga’s Ambition since the PC-88 era. They want Ishida Mitsunari to reason, adapt, and betray — not follow a script. They have been waiting for this without knowing it was possible.
Make each warlord a fully autonomous AI NPC, and the design space expands without proportional cost. An immersive single-warlord mode. Dynamic reactions when players break historical alliances. Unpredictable chains of betrayal that produce a different unification story every time. The content surface area multiplies with minimal incremental development investment.
But there is a genuine tension here: release speed versus historical depth and quality.
Ship a rough AI implementation early, and disappointment calcifies. Wait for something truly worthy, and another company may arrive first — though no competitor exists with anything close to Koei Tecmo’s 40-year asset base in this specific genre. Time, however, is not unlimited.
If Koei Tecmo can deliver something that moves players before anyone else does — something that makes the genre feel genuinely new — the fixed audience returns, a new generation discovers it, and AI-driven warlords, scenarios, and derivative content keep expanding. The next thirty years build on the last forty, stronger than before.
The market does not yet understand what Koei Tecmo might actually be investing in. That gap in understanding is where the opportunity lives.
Source: Original filing (TDnet) | 日本語版
Disclaimer | This article is for informational purposes only and does not constitute investment advice.