History

Figures converted from JPY at historical FX rates — see data/company.json.fx_rates (frankfurter / ECB). Ratios, margins, and multiples are unitless and unchanged.

The Story DISCO Has Been Telling

DISCO's story has not changed much; the world has. The same management team — under CEO Kazuma Sekiya, in the seat since March 2009 — has told an unusually consistent narrative for 17 years: stay inside Kiru, Kezuru, Migaku (cut, grind, polish), refuse to diversify, run the company on internal "Will Accounting / PIM" mechanics, and let the semiconductor cycle do the work on top. What changed is the cycle. Between FY2018 and FY2025 (year ending March 2025), operating margin stepped from 30.5% to 42.4%, and revenue tripled — driven first by power semiconductors for EV, then by IC for generative AI and HBM. The narrative is simpler than it was a decade ago. The numbers are larger. Credibility, measured by guidance discipline, has improved through this run, not deteriorated.

1. The Narrative Arc

Disco's published financials go back to FY2001 (year ending March 2001). The shape of the story is visible in one chart: two brutal cyclical busts (2002, 2009), a long climb, and an AI-era step-change.

Five inflection points actually moved the story:

The unusual thing is the flatness of the narrative across these inflections. Disco's 2020 Corporate Report ("Aiming for Excellence in Corporate Activities") and its FY2025 4Q earnings deck (April 2026) describe the company in essentially the same language: precision Kiru, Kezuru, Migaku; PIM (Performance Innovation Management) culture; "Will Accounting" internal-currency model; refuse to deviate from the core. The story did not pivot. The market did.

2. What Management Emphasized — and Then Stopped Emphasizing

Disco does not publish full earnings call transcripts; its filings are the Financial Review PDF plus a slide deck plus a Q and A summary. Tracking what the deck mentions, and how often, reveals the application mix shift that's actually driving the business — much more than anything management says about strategy.

Three things stand out:

• Generative AI / IC went from 0 to 5 in two quarters — from FY2023-Q3 (Jan 2024) to FY2023-Q4 (Apr 2024), exactly when NVIDIA's H100 cycle was lifting all back-end equipment names. The Q4 FY23 deck (April 2024) is the first to explicitly note "significantly increased for IC due to increased demand for generative AI."
• Power semi / EV quietly went the other way — from a "5" through FY2023 (year ending March 2024) to a "1" by FY2024-Q4 (April 2025), with explicit language: "demand for power semiconductors remained sluggish against the backdrop of a slowdown in the demand for EV" (FY2025 4Q, April 2026). Disco did not over-promise on EV; they simply stopped emphasizing it. That is healthier than retracting it.
• FX tailwind disappeared from the script — it was a "4–5" through FY2023 when GP margin was being explained by yen weakness. In FY2025 it dropped to "1" once the yen reversed. Management was direct about this: FY25-Q1 deck (July 2025) said "GP margin: YoY decreased due to the exchange rate, although high value-added products contributed." No spin.

The constant is "high value-added products" — a "4–5" throughout. This is the actual structural margin story management is asking investors to believe: even when FX flips against them, mix shift toward AI/HBM grinders and DGP (Grinder-Polishers) keeps GP margin around 70%.

3. Risk Evolution

Disco's formal Corporate Report "risk factors" section is unusual — it focuses on BCM (business continuity, earthquake/tsunami protection, pandemic response) rather than financial or market risk. The same text appears in FY2023 through FY2026 reports (a flag itself — they are not updating the formal disclosure). The real risk discussion lives in the Financial Review's "Future outlook" paragraph and the forward-looking-statement disclaimer.

What changed in the risk register:

• Capacity expansion execution moved from a footnote to a top-three risk. The Gohara Plant resolution (April 2025) commits $221M for Phase 1 alone (at the April-2025 FX rate), with Phases 2 and 3 still TBD. Combined with the Haneda R and D Center rebuild ($50M impairment booked in FY2023, construction starting April 2025), this is the largest capex commitment in DISCO's history. Investing cash outflow jumped from $98M (FY2022) → $108M (FY2023) → $450M (FY2024) → $908M (FY2025). That is a different company, balance-sheet-wise, even if equity ratio (~79%) remains fortress-like.
• China exposure crept up the risk lens. China was 34–38% of revenue through FY2023, then dropped to 32% in early FY2024 as customers diversified. Management has not flagged this in narrative form, but the regional mix has clearly become a thinner line to walk; Morningstar's "Bears Say" view explicitly cites the China self-sufficiency drive as the main bear case.
• Single-application concentration is the new tail risk. When IC for generative AI is a "5" and power semis is a "1," the symmetry has broken. The FY2025 4Q deck warned plainly: variations in demand were seen for each application.
• Earthquake/tsunami BCM didn't drop — it's a structural feature of running plants in Hiroshima and Nagano. The Gohara Plant move is partly to relocate production out of the tsunami-risk Kure Plant footprint. That risk-mitigation framing is a constant. The standing-disclosure text has not been updated meaningfully in three years.

4. How They Handled Bad News

DISCO's bad-news handling is terse and prompt, in the Japanese disclosure style. There is no investor-day theatrics. Three episodes are diagnostic:

The $50M Haneda R and D Center impairment (FY2023 4Q, April 2024). Disclosed in the same press release as record sales, treated as an extraordinary line item, explained in one sentence: "although an impairment loss amounting to approximately 7.5 billion yen due to the reconstruction of the Haneda R and D Center was included as an extraordinary loss, it was absorbed by the increase in operating income." No re-statement, no preview-and-bury, no obfuscation. The same deck explicitly displayed it inside the capex bar chart with "Haneda R and D Center" labeled in a separate color through FY2019–FY2024. Investors had been told this was coming for two years.

The EV/power-semi slowdown (FY2024 4Q, April 2025). Power semis had been a "5"-emphasis theme since FY2022. By April 2025, the deck stated plainly: "Decreased for power semiconductors due to a slowdown in the demand for EV." Then it pivoted attention to IC/AI without claiming the prior trend was wrong, and guided next-quarter shipments down 0.9% YoY — a tacit admission. Management did not pretend the cycle was over; they did not pretend it was fine.

FY2025 1Q guide-down (April 2025). The 1Q FY2025 forecast was net sales −9.4%, operating income −28.7%, net income −29.6% YoY (the harshest single forecast in five years). Issued the same day as record FY2024 results. Management's read-across was honest: the AI cycle would continue but the FY2024 Q3-Q4 burst rate (sales above $670M per quarter) was not the new baseline.

What did not happen: no buyback announced to defend the share price as it fell 50% from the July 2024 peak; no "we see strong demand normalization" reassurance; no walking back the guide-down once 1Q FY2025 actually came in +8.6% YoY instead of the −9.4% guided. Beats are simply reported as beats.

"We have always been conservative in our forecasts each time, so there is no particular reason or background for them." — DISCO Q1 FY2026 (June 2025 quarter) earnings call, on a $60M upward revision

The quote is revealing because it matches the data. Management is not just saying they are conservative; the guidance track record below confirms it.

5. Guidance Track Record

DISCO issues one-quarter-ahead guidance only, plus a revised full-year forecast at Q3. The cited reason: "The drastic and rapid fluctuations in customer willingness to invest make it difficult to predict demand." This conservative framing is structural — it is repeated verbatim in every Financial Review.

Every single forecast in the available 12-quarter window was a beat. The smallest beat is +4.3% (FY25 full-year sales revised in January 2026, beat by April 2026). The largest is +44.8% (FY25-Q1 operating income). The pattern is sandbagging — operating-income beats average roughly 25%, sales beats average roughly 12%. This is not noise; this is policy.

Credibility score: 9 / 10. Justification: 18 of 18 published forecasts beaten across FY2023-Q4 through FY2025-Q3; the company avoids forward-looking enthusiasm even when results would justify it; bad news (Haneda impairment, EV slowdown, FX reversal) is disclosed in the same release as good news without softening; the 4-year cumulative ordinary-income margin target of ≥20% has been hit for 10 consecutive years (FY2016 onward). The deducted point is for over-conservatism — guidance that is too easy to beat is its own form of low information content. Some investors would prefer a forecast range rather than the consistently low single point.

6. What the Story Is Now

The story today is shorter than it was. Three claims, each verifiable:

1. The structural margin step from ~25% to ~42% operating margin held through a yen reversal and a power-semi slowdown. FY2026 (year ended March 2026) operating margin was 42.3% with the yen appreciating against the dollar from its 2024 trough. The mix-shift story is intact: high-value-added IC dicing/grinding/DGP for generative AI and HBM is now the dominant earnings stream, and it carries the GP margin even without FX.

2. Capital allocation is reverting to "fortress + reinvest," not "buyback + payout." Dividends per share went from $2.05 (FY23) to $2.76 (FY24) to $3.38 (FY25) — record highs each year, set by a one-third-of-surplus formula. But the big move is the capex pivot: investing-cash-outflow $908M in FY2025, the Gohara Plant ($221M Phase 1, plus $17M for the land, Phases 2-3 TBD), Haneda R and D rebuild starting April 2025, and $627M added to time deposits. Net cash is being preserved against the next cycle; growth is being prepaid.

3. Management credibility is higher than at any point in the data window. Guidance discipline + transparent bad-news handling + zero scandal in the search record + a stable team (CEO since 2009, CFO since 2011, CTO since 2009) is unusual at a stock that has compounded 240%+ over three years.

What still looks stretched

• Application concentration. Generative AI/HBM/advanced logic is doing the heavy lifting; power semis have rolled over; consumer (smartphone/PC) is "gradually recovering" but small. If the AI capex cycle inflects, FY27 forecasts will look very different.
• China revenue (~32–38% range). Not visibly hedged in management commentary; the Morningstar bear case (China self-sufficiency) is the externally articulated risk.
• Multi-year capex execution. Gohara Phase 1 alone is ~10% of net assets at FY25 (year ending March 2025). Three phases plus Haneda R and D rebuild is a multi-year, multi-billion-dollar build. Disco has historically run lean — this is the first major capacity commitment of the current chapter.
• No buyback. With ~$1.79B cash and a 79% equity ratio at FY26 year-end, the dividend-only return policy is generous but inflexible compared to peers.

What to believe vs. discount