Industry

Industry — Understand the Playing Field

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

1. Industry in One Page

The semiconductor industry sells the building blocks of every digital device — logic chips that compute, memory chips that store, power chips that switch, and the sensors and analog parts that connect them to the physical world. The companies that make those chips do not build their own equipment; they buy it from a separate industry called wafer fabrication equipment (WFE), which in 2026 is roughly a USD 115 billion global market growing at ~7% per year toward ~USD 163 billion by 2031 (Mordor Intelligence). WFE itself splits into two halves: front-end tools that build transistors on a round silicon wafer (lithography, etch, deposition, metrology — ASML, Applied Materials, Lam Research, KLA, Tokyo Electron) and back-end tools that take the finished wafer and turn it into individual packaged chips (grinding, dicing, bonding, test — Disco, Tokyo Seimitsu, BE Semiconductor, ASMPT, Kulicke and Soffa). Disco sits squarely in back-end, in three specific physical steps — cut, grind, polish — where it holds 60–80% global share depending on the product line.

What a newcomer usually misunderstands: this is not "another tech industry." The customer base is roughly 30 fabs and packaging houses worldwide; the same tool sells to TSMC, Samsung, SK Hynix, Intel, Micron, and a handful of OSATs (outsourced assembly and test). Demand swings with fab capex cycles, but the consumables (diamond blades, grinding wheels) attached to Disco's installed base run on customer wafer volume, not capex — a structural shock absorber that smooths the cycle. The AI buildout has now made one back-end step — wafer thinning for HBM (high-bandwidth memory) and CoWoS advanced packaging — into the most-discussed bottleneck in semiconductors.

Disco's seat: layer 3 — back-end equipment for thinning, cutting, and polishing wafers, with embedded consumables.

2. How This Industry Makes Money

Semiconductor equipment is sold three ways: tools (the boxes — dicing saw, grinder, polisher; one-time sale of $0.25–0.94 million per unit at Disco), consumables (the diamond blades and grinding wheels that wear out — sold per cut/per wafer, monthly run-rate), and services (installation, training, parts, leasing). The economics are very different from chipmaking. Equipment vendors do not need fabs; they need engineering talent, deep applications labs (Disco runs ~5,000 test cuts/year for customers), and an installed base. Once a fab qualifies a tool for a process, switching is painful — re-qualification can take quarters, blade compatibility is process-specific, and a yield difference of 50 basis points on a leading-edge wafer dwarfs equipment price differences.

Where bargaining power sits. With ~60–80% global share in dicing and grinding (Morningstar puts it at 60–70%; sell-side and trade press cite 70–80% on advanced nodes), Disco is the price-setter, not the price-taker, in its core categories. The countervailing pressure: its customers are the most concentrated buyer base in any industry — TSMC, Samsung, and SK Hynix alone drive a large share of leading-edge orders. The reason that imbalance does not crush margins is technology lock-in: every new packaging architecture (TAIKO ring-thinning for power devices, KABRA laser slicing for SiC ingots, stealth dicing for memory, hybrid-bond pre-thinning for HBM) requires Disco-specific application know-how validated through hundreds of test cuts. Capital intensity is low for the customer per dollar of value, and Disco's tools are typically a low single-digit percent of a fab's bill of materials.

Capital intensity for the vendor. Disco runs three production sites in Japan (Kure, Kuwabata, Chino) — onshore Japanese manufacturing rather than offshore arbitrage. Capex is rising (FY2026 saw $851M used in investing activities driven by $627M of time-deposit placements and new manufacturing real estate), but the underlying business is not capital-heavy: total property, plant and equipment is $1.4B against $2.74B of revenue, and FCF would have been positive ex-time deposits.

3. Demand, Supply, and the Cycle

Semiconductor equipment is the most cyclical sub-sector inside a cyclical industry. The customer is a fab planner deciding whether to add wafer capacity 12–24 months from now. When end demand is rising, fabs over-order; when it slows, equipment orders disappear faster than wafer volumes do. The chart below shows Disco's own revenue history through three completed cycles — the dot-com bust, the GFC, and the 2019/2023 downturns — plus the AI-driven upcycle.

Where the cycle hits first. Equipment is a forward-looking asset purchase, so order momentum turns several quarters before fabs cut wafer starts. Disco itself only guides one quarter forward; management explicitly says "drastic and rapid fluctuations in customer willingness to invest make it difficult to predict demand." Watch shipment value (not revenue) for the leading signal — it includes orders received and tools shipped but not yet recognized. FY2026 shipment value of $2.78B grew 10.3% YoY; the Q1 FY2027 forecast of $828M is up 18.8% YoY, a re-acceleration.

Downturns hit Disco less than peers. In FY2009 revenue fell 42% ($921M → $539M) but the company stayed near break-even because consumables and parts (then ~33% of mix) held in while equipment collapsed. FY2010 had GFC overhang. The 2019 memory downcycle was milder (-12%). The FY2024 power-semiconductor weakness barely registered in headline numbers because AI/HBM more than offset it.

4. Competitive Structure

The global back-end equipment industry is a federation of niche monopolies, not one consolidated market. Each of the three or four physical steps (thin/dice, attach, bond, test) has one or two dominant suppliers with 40–80% share, and almost no overlap between them. Disco's competitive set is narrow:

Structure read. Disco's dicing-and-grinding category is a global duopoly with Tokyo Seimitsu (Accretech), and a wide one — Accretech's total revenue ($1.05B FY2026) is ~38% of Disco's, and its product mix includes metrology that Disco does not sell. Disco is the technology leader on every advanced-process refinement (DBG/SDBG memory thinning, stealth dicing, TAIKO power-device thinning, KABRA SiC slicing, laser lift-off for LEDs). The two share leading-edge customers; trailing-edge and Chinese demand is more contested.

The China question. China's drive for semiconductor self-sufficiency has produced credible competitors in front-end deposition and etch, but precision dicing/grinding requires application know-how accumulated over decades (Disco runs 70+ private test booths in Tokyo; ~260 application engineers worldwide). Morningstar's bear case explicitly flags this as the long-term risk; near-term, Chinese fab capex is a net positive for Disco as installed-base expansion drives both equipment and consumables. Asia ex-Japan (Taiwan, Korea, China, Singapore) is 75% of Disco's FY2026 sales.

5. Regulation, Technology, and Rules of the Game

Semiconductor equipment lives inside the most actively-regulated trade corridor in modern industry. The rules below are not generic — each one shifts who gets paid in this specific niche.

The biggest technology shifts that change Disco's economics — not just buzzwords — are (i) HBM stacking, where each AI-server GPU now sits next to 4–8 HBM stacks of 12–16 ground-thin DRAM die each, multiplying Disco's grinding footprint per packaged chip; (ii) hybrid bonding, where Besi leads the bonder itself but the wafer must first be ground and prepared by a Disco-class tool — pre-bond and post-bond thinning are both Disco wallet; and (iii) SiC wafer slicing, where Disco's KABRA laser process recovers far more usable wafers per ingot than incumbent diamond wire-saws. None of these are forecasts; each is in production or qualification today at the named customers.

6. The Metrics Professionals Watch

A short list of what actually moves stocks in this industry. Disco's own targets are useful anchors: management runs a four-year cumulative ordinary-income margin target of ≥20% (FY2026 reading: 41.4%, a 10-year streak of hitting it) rather than annual targets — an admission of the cycle.

7. Where Disco Corporation Fits

Disco is the focused, high-margin specialist of the semiconductor equipment industry — the opposite of broad-portfolio leaders like Applied Materials or Tokyo Electron. It owns three steps, deeply, and refuses to extend.

Market caps approximate as of May 2026 (Morningstar, peer_valuations.json). Disco is small vs. front-end leaders but the largest pure-play back-end specialist.

8. What to Watch First

A six-signal checklist a reader can re-run each quarter to know whether the industry backdrop is improving or deteriorating for Disco — independent of company-specific noise.