Tokyo Electron

Phase A — Understand the business

Company Overview

Tokyo Electron ("TEL") is the world's #3-4 wafer-fab-equipment (WFE) maker and Japan's largest — a critical-tools supplier that sits between raw silicon and finished chips. Founded 11 Nov 1963 as Tokyo Electron Laboratories by Tokuo Kubo and Toshio Kodaka, seeded by Tokyo Broadcasting System; began as a trading house importing Thermco diffusion furnaces and selling car radios, then pivoted to in-house R&D/manufacturing in the 1980s. HQ Akasaka, Minato-ku, Tokyo; ~15,000-18,000 employees.

How it makes money. TEL sells the process tools that etch, deposit, coat, develop, clean, thermally treat and test wafers — sold as multi-million-dollar capital equipment to chipmakers, then monetized again for a decade+ through Field Solutions (service, spare parts, upgrades/retrofits on the installed base). Reporting segments: Semiconductor Production Equipment (SPE) — >95% of sales; Flat Panel Display (FPD) equipment — small; Other.

FY2026 shape:

• Net sales ¥2,443.5B (~$15-16B) — record, +0.5% YoY
• Field Solutions ¥626.0B (+16.3% YoY) → SPE new-equipment ≈ ¥1,817.5B
• Advanced packaging ~¥200B and climbing
• AI-related demand ≈ 40% of sales

Customers are the five leading-edge fabs — TSMC, Samsung, SK Hynix, Intel, Micron — plus Chinese mature-node fabs (SMIC, CXMT, YMTC and the domestic long-tail). Concentration is high (a handful of buyers = most of new-equipment sales) but the mix of buyers rotates with the cycle (China-heavy 2023-24 → Taiwan/Korea leading-edge 2025-26). Contracts are project-based capital orders (no take-or-pay); the recurring, sticky revenue is Field Solutions, which grows with wafer starts and installed-base utilization.

Supply Chain

Upstream into TEL: precision machined parts, vacuum/robotics subsystems, RF-power and gas-delivery components, optics and control electronics from a fragmented Japanese/global component base (VAT vacuum valves, MKS/Advanced Energy-class subsystems, precision-machining SMEs) — no single-source choke that publicly binds TEL.

TEL's position in the AI compute chain: Equipment (ASML · Applied Materials · Lam · KLA · **TEL**) → Foundry/Memory (TSMC · Samsung · SK Hynix · Intel · Micron) → Fabless (NVIDIA · AMD · Broadcom) → Advanced packaging (TSMC CoWoS) → Hyperscalers (MS · Meta · Google · Amazon).

The load-bearing adjacency: TEL's coater/developer track runs inline with ASML's EUV scanner. Every EUV exposure needs a photoresist coat-and-develop step, and TEL owns ~100% of the EUV coater/developer market — it supplied the track integrated with ASML's High-NA EXE:5000 at the imec-ASML lab in Veldhoven. So TEL is complementary to ASML and KLA (litho, metrology) and competitive with Lam and Applied (etch, deposition, clean). That means TEL is levered to both the EUV ramp (litho-adjacent) and the deposition/etch layer-count explosion (3D NAND, GAA, advanced DRAM) — a rare two-sided exposure.

Downstream chokepoints TEL feeds: CoWoS advanced packaging (TSMC, sev. 9, easing late-2026 to ~130K wafers/mo), HBM3E/HBM4 (SK Hynix > Samsung > Micron, sev. 9, structurally tight through 2026), and data-center power (7 GW US shortfall 2026). TEL sells etch/bonding/probing into the HBMHigh-bandwidth memoryMemory stacked in layers beside a processor so data reaches it faster. The usual bottleneck in AI chips. and advanced-packaging build-out — it is a supplier to the bottleneck, which is the profitable place to stand.

Competitive Advantages (moats)

TEL is a collection of niche near-monopolies bundled inside a top-4 WFE franchise:

• Coater/developer tracks: ~90-91% global share — and ~100% of EUV-layer tracks. This is the crown jewel: a de-facto standard that every leading-edge fab must design around.
• GAA isotropic etch: ~100% share — TEL owns the etch step that 2nm gate-all-around transistors require.
• Wafer probers: ~38% (#1); batch deposition, diffusion furnaces, gas-chemical etch: #1 worldwide.
• Dry etch ~23% / deposition ~27% — here TEL is a strong #2-3; Lam Research leads etch (~45-50%), Applied leads deposition. Positions oscillate with 3D NAND and GAA nodes.

Durable moats: (1) Co-development lock-in — process tools are qualified into a customer's recipe over years; switching mid-node is a yield gamble almost no fab takes. (2) Installed base → Field Solutions annuity (¥626B, +16%) that compounds and smooths the capital-order cycle. (3) Process breadth — TEL can sell a coat-etch-deposit-clean-probe suite, giving it integration leverage Applied uses in reverse. (4) EUV-era criticality — as litho gets harder, the track and the isotropic etch become gating.

Bargaining power is shared — the top-5 WFE oligopoly (AMAT, ASML, TEL, Lam, KLA) and the ~5 leading-edge fabs need each other; neither side has runaway pricing power. TEL's leverage is highest where its share is highest (tracks, GAA etch) and lowest in contested etch/deposition where Lam and Applied press on price with scale.

Segments

By product (FY2026, directional — TEL does not itemize all product revenue publicly):

. All new-equipment lines guided up sharply into FY2027 — the growth is broad, not one-tool.

By geography (new-equipment composition, FY2026):

• China 34.1% full-year, but decaying fast: Q4 FY2026 China 26.8% vs 47.4% in Q4 FY2024 — a ~21-pt collapse in two years as export controls bit and mature-node spend rolled over. China ¥ sales fell to ~¥175.5B from ~¥279.4B.
• Q4 FY2026: Korea 24.3%, Taiwan 22.0% — leading-edge logic (TSMC N2/GAA) and HBM/DRAM (SK Hynix, Samsung, Micron) are now the growth engines, replacing China's mature-node bulk.

The trend that matters: the revenue base is rotating from China quantity to leading-edge quality. That is healthier mix (higher-value tools, AI-durable) but it removes the China shock-absorber that let TEL fill soft quarters 2023-24. Concentration risk moved from "one country" to "three customers' AI-capex plans."

---

Phase B — Measure performance

Earnings Result (FY2026, reported 2026-04-30)

A record top and bottom line — with a margin warning underneath:

• Net sales ¥2,443.5B, +0.5% YoY — record, but flat (FY2025 was already ¥2,431.6B)
• Operating income ¥624.9B, op margin 25.6% — down hard from FY2025's ¥697.3B / 28.7%
• Gross margin 45.3% vs 47.1% prior — ~1.8pt compression
• Net income ¥574.5B (record, +5.6%); diluted EPS ¥1,250.88
• FCFFree cash flowCash left after paying to run and maintain the business. Unlike profit, it is hard to flatter with accounting choices. ¥330.7B — conflict: NineScrolls reports FY2026 FCFFree cash flowCash left after paying to run and maintain the business. Unlike profit, it is hard to flatter with accounting choices. ¥433.2B "record"; the gap likely reflects operating-cash-flow vs FCFFree cash flowCash left after paying to run and maintain the business. Unlike profit, it is hard to flatter with accounting choices. definitions — flagging, not resolving.

The tell: revenue made a record while operating profit fell ~10%. Margins were squeezed by an FX headwind (~¥70B), 9-10% wage inflation, and mix. But the exit velocity is violent — Q4 revenue ¥711.8B (+28.9% QoQ), Q4 op income ¥205.6B (+77.1% QoQ), Q4 gross margin 46.8% (+4.1pt). The year was a soft-front-half / explosive-back-half shape as AI-server tool demand accelerated.

Guidance (FY2027, 1H only):

• 1H net sales ¥1,570B (+33% YoY) vs ¥1,420B consensus — a large beat-on-guide
• 1H op income ¥431B (+42.2%), op margin 27.5% (record on a half-year basis)
• TEL will stop giving full-year forward guidance — disclosing 1H only, citing larger/more-volatile customer spend. Read this as reduced visibility, not confidence — management is hedging its own forecast risk.

Market reaction: stock +14% in a session post-print (June 2026), part of a +123%/12-month run to record highs — the tape treated the guide as validation of the AI-WFE super-cycle.

Earnings Calls (sentiment trend)

Speakers: CEO Toshiki Kawai, SVP Finance Hiroshi Kawamoto.

Tone: the most bullish in years, with an honest margin caveat. The trajectory across FY2026 calls: Q1 "revenue dips as investment patterns shift" (cautious) → Q3 "revenue declines amid optimistic AI-driven outlook" (bridging) → Q4 "demand still very strong, customers request delivery acceleration almost weekly" (conviction). Kawai's recurring themes: AI server demand, HBM/DRAM interconnect share wins, GAA gate/isotropic etch, advanced packaging +60%. What they stopped leaning on: China as a growth driver (now framed as "stable mid-30%s," a managed decline).

The one thing management won't promise: near-term margin. Kawai committed to 50%+ gross margin "within two years" and conceded the 35% operating-margin target "remains challenging" against FX and labor. That candor is a credibility positive and a valuation negative — the AI story is intact, the profitability story is a work-in-progress.

Comps

WFE oligopoly, mid-2026 — the entire group has re-rated on AI-capex:

Read: TEL at ~57x trailing is not a lone outlier — ASML sits at 55x, AMAT ~40x. The whole complex has doubled off 2024 levels on AI-capex optimism. TEL screens at the premium end (with ASML) and above Lam/KLA (~28-34x forward). TEL's justification vs ASML is weaker: ASML has an absolute EUV monopoly; TEL's monopoly is in the (smaller) track/GAA-etch niches while it contests the big etch/deposition pools with Lam and Applied. The multiple prices TEL like ASML without ASML's litho monopoly. Peer multiples are; where unsourced, marked n/a rather than fabricated.

Stock-Price Catalysts (5-yr, moves >5%)

:

• Oct 2022 — US export controls launched → TEL cut profit outlook >25% (Nov 2022). Down.
• 2023 — 3-for-1 stock split; China surged to ~43% of revenue (Chinese fabs pre-buying mature-node tools ahead of controls). Up.
• Jul 2023 — Japan's own export controls (23 SME categories). Overhang.
• 2024 — further Japan tightening; market capMarket capitalisationThe share price multiplied by the number of shares. What the market says the equity is worth. ~$114.6B, 3rd-most-valuable in Japan.
• Spring 2025 — semi/AI-capex-doubt selloff (tariff + DeepSeek-efficiency fears); trough ~¥19,870 (52-wk low).
• 2026 — AI-WFE melt-up: +123% / 12 months, +14% single-session post-Q4, record highs ~¥71,060 (2026-07-09). Up violently.

Pattern: TEL trades on (1) export-control headlines (China is the swing factor) and (2) the AI/leading-edge capex narrative — not primarily on its own quarterly beats, which have been mixed. The market treats TEL as a high-beta call option on the WFE cycle and US-China policy. That is exactly why the current ~57x multiple is fragile: it was set by narrative, and narrative reverses faster than order books.

---

Phase C — Judge people & books

Management

CEO Toshiki Kawai — long-tenured (President & CEO since 2016; ~40-year TEL lifer), a professional operator, not a founder. Track record: steered TEL through the 2019 and 2023-24 down-cycles without losing its share crown, and positioned the company early for GAA/HBM/advanced-packaging — the exact vectors now driving the AI re-rating.

Capital allocation — shareholder-friendly and net-cash disciplined:

• ~50% payout-ratio policy tied to net income; FY2026 DPS ¥601 (record); 1H FY2027 interim ¥361.
• Buybacks: ¥149.9B repurchased in FY2026 (3.6M shares cancelled); fresh authorization up to ¥150B / 7.5M shares through Mar-2027. Explicitly framed as ROE-accretive + offsetting SBC dilutionDilutionIssuing new shares, so each existing share owns a smaller slice of the same company..
• Total FY2026 shareholder return ~¥437.4B (record) — dividends + buybacks.
• ROE ~28.6%, "close to 30%," against a debt-free, ¥506B-cash balance sheet.

Red flags: none egregious. The nuance: the balance sheet is over-capitalized — ¥506B cash, ¥713B inventory, zero debt, ¥2.0T equity — which dilutes ROE and is why management leans on buybacks to hit the 30-35% ROE target. A more aggressive capital structure would lift returns; TEL's conservatism is culturally Japanese and unlikely to change fast. Midterm plan targets — ¥3T+ sales, 35%+ OP margin, 30%+ ROE — the sales line is within reach (FY2027 tracking ~¥3.1T ); the 35% margin is a stretch (FY2026 was 25.6%, 1H FY2027 guided 27.5%).

Forensic Red Flags

Accounting posture: clean and conservative, with two cyclical watch-items:

• Inventory ¥713.1B against ¥2,443.5B revenue (~29% of sales / ~107 days) — elevated, but consistent with a WFE maker building tools ahead of a delivery-acceleration ramp customers are demanding "weekly." Watch: if the AI-capex pull-in stalls, this inventory becomes the write-down risk. Not a fraud flag — a cycle flag.
• China receivables — as China shrank from 47% to 27% of new-equipment sales, collection/return risk on the fading mature-node book is worth monitoring; not disclosed as impaired.
• Zero debt, ¥506B cash, ¥2.0T equity — no leverage games, no goodwill tower (TEL grows organically, not by acquisition — the 2015 AMAT merger was blocked). SBC exists but is modest and offset by buybacks. Cash flow tracks earnings (FCF ¥330-433B — definition conflict noted in Lens 5), no earnings-quality divergence visible.

Regulatory findings:

• SEC: None — TEL has no CIK and files nothing with the SEC; no EDGAR LR/AAER search possible. Verified via our model, 2026-07-10.
• US antitrust: none current. Historical: the 2015 TEL-Applied Materials $10B merger was cancelled on US DOJ antitrust concerns — relevant to why the WFE oligopoly stayed 5-wide.
• Trade-secret matter (material, disclose): In a TSMC trade-secret case ruled ~April 2026, a former TEL-Taiwan employee received a 10-year sentence and a ~$4.6M fine; TEL's Taiwan unit pleaded guilty and settled with TSMC (TSMC received T$100M of the fine). TEL states no organizational involvement and no external leakage were found, and agreed enhanced trade-secret protections. A governance/reputational item against TEL's most important customer relationship — low financial impact, non-trivial optics.
• Export controls: TEL operates under Japanese METI controls (July 2023, 23 SME categories) aligned with US/Netherlands; ongoing compliance burden, not an enforcement finding.
• Net: No material accounting or securities-enforcement findings — verified via SEC EDGAR EFTS (no CIK), web search, and public reporting as of 2026-07-10. The trade-secret case is the one live legal item.

---

Phase D — Project & stress-test

Forward Projection (FY2027-FY2029, fiscal years end March)

Built bottom-up from FY2026 actuals + 1H FY2027 guidance. All outputs; inputs labeled.

Anchors: FY2026 rev ¥2,443.5B, EPS ¥1,250.88, ~458M shares, net margin 23.5%. 1H FY2027 guided ¥1,570B / +33%. SEMI: total equip $145B (2026) → $156B (2027), WFE +9.0%/+7.3%. TEL's own WFE call $150-170B/yr runs above SEMI's $135.2B WFE line — flag: TEL's outlook is broader/more bullish than SEMI's; base case leans toward SEMI.

• Bull: AI-capex holds, gross margin recovers toward 50%, OP margin → 30%+ → FY2029 EPS ~¥2,500+.
• Bear: WFE cycle rolls over 2027-28, China worsens, FX reverses → FY2028 EPS flat-to-down ~¥1,400-1,500.

Valuation frame: at ¥71,060, FY2027 base EPS ¥1,630 → ~44x forward; even FY2029 base ¥2,010 → ~35x three years out. The stock needs the bull path to make today's price look reasonable on normal WFE multiples (~20-25x). (No our model logged — watchlist/unattended mode.)

Bull vs Bear

Bull case. TEL is a toll road on every leading-edge wafer: 90%+ coater/developer, ~100% EUV track and GAA isotropic etch, #1 in probers/furnaces/batch-depo. The AI build-out is simultaneously lifting logic (2nm GAA at TSMC/Samsung/Intel), DRAM/HBM (etch + bonding + probing), advanced packaging (+60% guide) and a NAND recovery — TEL sells into all four. 1H FY2027 +33%/+42% proves the inflection is real, not hoped-for. Net-cash, ~29% ROE, ~50% payout, ¥150B buyback. Secular tailwind: SEMI sees records into 2027 ($156B). The pre-mortem's failure modes are all cyclical/policy, not structural — the moat itself is not in question.

Bear case (2-3 permanent-impairment / de-rating risks).

1. Valuation is the risk. ~57x trailing / ~44x forward on a cyclical WFE name at what may be mid-to-late cycle. GuruFocus flags "significantly overvalued" (GF value $103.58 vs $245 ADR). A re-rate to even 30x on flat EPS is -30%+ with no operational error.
2. Margins are going the wrong way while revenue makes records — OP margin 25.6% (vs 28.7%), gross margin down 1.8pt, FX (~¥70B) and 9-10% wage inflation biting. The 35% margin target looks unreachable; a growth-at-declining-margin name doesn't deserve a growth multiple.
3. China is a one-way door. From 47% to 27% of new-equipment sales in two years, with US lawmakers pushing tighter controls. The shock-absorber is gone; growth now rests entirely on three customers' AI-capex plans (TSMC, Samsung, SK Hynix) that can be cut in a single board meeting.

Pre-mortem (18 months out, thesis broke): AI-capex digestion hits in 2H CY2027 (DRAM/NAND tool orders that were "coming" get pushed); TEL's ¥713B inventory writes down; a China control-tightening removes the residual mature-node book; the multiple mean-reverts from 44x to 25x. EPS misses the ¥1,630 base by 15%, and the stock is -40% — not because the franchise broke, but because the price did.

Contrarian view (what the market refuses to see): the crowd treats TEL as a secular AI compounder and pays 55x. It is actually a best-in-class cyclical whose earnings fell in the down-legs of 2019, 2023 and 2024 — and whose China tailwind just became a headwind. The market is extrapolating the back-half-of-FY2026 acceleration as the new baseline.

Devil's Advocate (short-seller)

Dismantling the bull:

• Where revenue concentrates: ~5 fabs; and increasingly just the three AI-capex spenders now that China's long-tail is throttled. Lose one Samsung/SK Hynix capex reset and a quarter cracks.
• Moat weaker than bulls think: the big pools — etch and deposition — are Lam's and Applied's, not TEL's. TEL's absolute monopolies (tracks, GAA isotropic etch) are the smaller WFE segments. Bulls pay an ASML multiple for a franchise that only has ASML-like dominance in a minority of its revenue.
• Most dangerous competitor bulls underrate: Lam Research. Etch is the largest WFE segment and Lam leads it; every GAA/3D-NAND node reshuffles etch share, and Lam's innovations have "caused year-to-year market-share swings". A Lam win in GAA etch would hit TEL's best growth story.
• Capital allocation nit: over-capitalized balance sheet flatters nothing and needs buybacks to reach ROE targets — a tell that organic returns on the ¥2T equity aren't 35%.
• What must hold for today's price: WFE grows 20%+ two years running (TEL's number, above SEMI's), TEL holds/gains share in contested etch/depo, margins recover to 30%+, China doesn't get worse, and FX cooperates. All five, simultaneously.
• -20-30% growth-disappointment scenario: FY2028 rev ~¥2,800B not ¥3,500B; net margin slips to ~21%; EPS ~¥1,300; at a de-rated 25x → ~¥32,500, roughly -55% from ¥71,060.
• Single permanent-impairment scenario (low probability): a hot US-China escalation that pulls Japan into an outright leading-edge-tool ban to all of China and triggers Chinese retaliation on Japanese suppliers — plausibility low-to-moderate, impact large.

Management Questions (ordered by information value)

1. Your WFE outlook of $150-170B/yr for CY2026-27 sits above SEMI's ~$135B — what exactly is in your number that SEMI's WFE definition excludes, and what would make you cut it?
2. Operating margin fell to 25.6% at record revenue. Bridge us from here to the 35% target — how much is FX, how much wage/mix, and which is structural?
3. You committed to 50%+ gross margin "within two years" — what are the 2-3 concrete levers (pricing, productivity, new products), and what's the risk each doesn't land?
4. China is 27% of Q4 new-equipment sales and falling. At what floor does it stabilize, and how much of the residual is at risk from further US/Japan controls?
5. You stopped giving full-year guidance. Is that reduced visibility or reduced confidence — and what changed in customer ordering to force it?
6. Inventory is ¥713B (~107 days). How much is committed against firm AI-server orders vs. speculative build, and what's the write-down exposure if 2H CY2027 orders slip?
7. In contested etch and deposition (where Lam and Applied lead), are you gaining or ceding share at 2nm GAA and advanced DRAM — with numbers?
8. HBM/advanced-packaging (+60% guide): what's your share of the bonding/etch/probe steps, and how defensible is it against Applied and BESI/ASMPT in hybrid bonding?
9. If a customer asks you to accelerate deliveries "weekly," where is your own supply-chain/component constraint, and what's the lead-time floor?
10. Capital structure: with zero debt and ¥506B cash, why not lever modestly to lift ROE past 35% rather than relying on buybacks?
11. GAA isotropic etch is ~100% yours today — what stops Lam or Applied from qualifying a second source at a major fab within two nodes?
12. What is your Field Solutions attach-rate trajectory, and can it reach a level that materially de-cyclicalizes the P&L?
13. Post the TSMC trade-secret case, what specific controls changed, and how do you protect the co-development trust that is your moat?
14. Where would you deploy the balance sheet if the AI-capex cycle paused for 12 months — counter-cyclical R&D, M&A (post-2015), or returns?
15. What's the one leading indicator you watch that would tell you the AI-WFE cycle is peaking before the order book does?

---