← back to picks

HQ Taoyuan · Taiwan Reporting TWD Credit synth Aaa/AAA · net cash ~NT$9.7bn Valuation 2026-06-02 Damodaran FCFF · 1000-iter MC FCFF · Dark v3

Good company, wrong price — priced for a margin band-break it has never delivered · −24.6% margin of safety

Market sits ABOVE the entire p5–p95 MC fan — P(intrinsic < market) = 99%

Disciplined intrinsic value NT$172.58/share (after a 10% governance discount; pre-haircut NT$191.75) vs market NT$215 — a −24.6% margin of safety. The Monte Carlo p95 tail still tops out at NT$202, below today's price: every plausible draw lands under NT$215, so P(intrinsic < market) = 99.1%. Even crediting the full AI/optical regime-change steelman (margin 9%, growth 7%) lifts intrinsic only to ~NT$217 ≈ price — still no cushion either way.

SectorEMS / contract SMT-PCBACountry mixChina 80% · TW 10% · KR 5%β / MC σ1.30 levered · ±NT$17/sh (1000 runs)GovernanceWu family · 10% haircutQualityNet cash · 21% ROIC · synth AAARe-rateNT$91 → ~NT$215 in 4 months

Intrinsic / share

NT$172.58

post 10% gov · pre NT$191.75

Market / share

NT$215.00

2026-06-02 · TWSE · 52-wk NT$91–254

Margin of safety

-19.7%

vs intrinsic

Enterprise value

NT$46.36B

74.3% terminal

Cost of equity / debt

9.25% / 2.44%

β 1.30 · CRP 0.85%

Terminal ROIC / g

12.00% / 2.00%

spread ~344bp (ROIC 12.00% vs WACC 8.56%)

What it sells, where it sells

Operating segments

Display / TFT-LCD boardsLegacy core — LCD/TV motherboards for Chinese panel makers (BOE, CSOT, HKC); soft, the commodity end~46%

Mini-LED backlightCollapsed from ~32% to ~12% — the swing factor behind the cyclical revenue volatility~12%

DRAM modules / automotiveMicron DDR/DRAM (Xi'an, ~10% rev) + auto dashboards — the steadier, higher-ASP non-display base~32%

AI / optical / DDR5~5% of FY25, guided to 10–15% — the entire bull thesis rests here. Higher margin, certification-gated, back-half-loaded~10%

~58% of revenue is still display-linked (TFT-LCD + Mini-LED) and structurally soft, while the entire re-rating bet rides on the AI/optical/DDR5 sliver (~5% of FY25, guided 10–15%) carrying the margin of the whole company up — a mix the consolidated P&L has not yet shown. 1Q26 went the wrong way: op margin fell to 5.0%, EPS −26% YoY.

Country mix (revenue-weighted CRP input)

🇨🇳China80%

🇹🇼Taiwan10%

🇰🇷South Korea5%

🇺🇸United States4%

🇩🇪Germany (EMEA proxy)1%

Unlike export-EMS peers, TSMT's customers are themselves predominantly Chinese panel makers (BOE, CSOT/TCL, HKC), so its revenue is a genuine China end-market that loads real China CRP into the WACC (blended ERP ~5.08%). One upside lever the base does not credit: if the AI/optical pivot is genuinely ex-China (US hyperscalers), the prospective China weight falls and the discount rate eases — but that is the very band-break the price already assumes.

Background — five things to know before the case

Context the share price doesn't carry on its face. Skim this once and the bull / bear bullets below stop reading as inside baseball.

The stock ran ~2.2× in four months (NT$91 → ~NT$215) on an AI-server / optical-comms / DDR5 narrative. P/E went from ~9× to ~22× trailing while the underlying business — surface-mount/PCBA assembly of LCD/TV boards for Chinese panel makers — stayed a low-margin, price-taker contract manufacturer. The consensus 1-year price target (NT$163) sits below the market price; the stock has run past even the sell-side.
Operating margin has never escaped a 5.0–8.5% band in 10 years. 5.08% low (FY18) → 8.49% peak (FY22 super-cycle) → 7.03% (FY25). At NT$215 you are paying for the AI/optical/DDR5 mix to break that band upward; as of FY25 it had not shown in the consolidated number, and 1Q26 went the wrong way — op margin ~5.0%, EPS −26% YoY on higher expenses.
Cyclical revenue normalizing off a super-cycle peak. NT$30.6bn (2018) → NT$68.5bn (2022 all-time peak) → NT$46.5bn (2023, −32%) → NT$49.9bn (2025, +10%). The bears' "−5.6%/yr 5-yr EPS CAGR" is an artifact of measuring off the 2021/22 top; the defensible mid-cycle base is ~NT$50bn, not the NT$68bn peak nor the NT$45bn trough.
Wu-family controlled, clean balance sheet — a real, quality business. Chairman Kai-Yun Wu (31 years, since 1995) plus son/president Yun-Chung Wu; ~14% insider, 4 independents on an 11-seat board. Net cash ~NT$9.7bn (~16% of mcap), interest cover ~20×, 10 straight profitable years, durable mid-teens-to-low-20s ROIC even at the cycle trough. This is a good company at the wrong price, not a broken one.
Cash reaches minorities, but the buyback does not return capital. Dividend real and growing (NT$5.5 2025 vs 5.0 2024), ~58–63% payout, 28 consecutive years of cash dividends. But the Dec-2025 buyback is earmarked for employee transfer, not cancellation, so it is a dilution-offset, not return-of-capital; 8 mainland plants add a China asset/repatriation overhang above the revenue CRP.

Where we diverge from the sector

Each row asks: if we'd used the sector median for this one assumption instead of our override, how much would the share price change? Negative = our override is more conservative (worth less); positive = more aggressive (worth more). Skim down the rightmost column to see which bets are doing the work.

Assumption

Our base case

Sector / what price implies

Value impact

01Revenue growth10-year revenue CAGR off the normalized mid-cycle base

~4.6% highThrough-cycle normalization: ~NT$50bn → ~NT$78bn by Y10; non-display mix outgrows the soft display core

~7% to justify priceRegime-change steelman — ~NT$87bn by Y10, crediting the AI/optical ramp the company guides but has not delivered

−NT$18

02Operating marginYear-10 target — the load-bearing call

7.5%Inside the 10-yr 5.0–8.5% band (FY25 ran 7.03%); a modest ~50bp mix tailwind, NOT a band-break

9.0% to justify priceBREAKS above the decade-long band on the bet AI/optical/DDR5 re-rates the whole company; 1Q26 ran 5.0%

−NT$44

03Sales-to-capitalReinvestment efficiency — working-capital-heavy EMS

1.30×Above total asset turnover (~0.9–1.0); incremental lines less capital-hungry than the WC base

~1.5×Electronics (General) cross-sector reference

~NT$0

04Terminal growthYear 10+ steady state vs TWD risk-free ceiling

2.0%TWD long-run inflation; held ≤ the blended 2.65% risk-free

2.65% ceilingrisk-free (65/35 blend); terminal g held below it

NT$0

05Cost of capital10y WACC — β + China-weighted CRP + TWD rf

8.56%β 1.30 (non-suppressed 5Y regression, nudged toward sector) · rf 2.65% (65/35 blend) · China-0.80 CRP · D/V ~10%

~9.3% Ke at β 1.71sector-relevered β → a higher WACC and lower intrinsic still

−NT$7

Net read: disciplined base lands ~25% below price

Every disciplined input — through-cycle 4.6% growth, an inside-band 7.5% margin, a non-suppressed β — lands intrinsic at NT$172.58, ~25% below the NT$215 tape. To justify the price you must pencil in the right-hand column: ~7% growth AND a 9.0% margin that breaks a decade-long band for the first time. Even granting both, the steelman only reaches ~NT$217 ≈ price — no cushion either way.

−NT$42

Our disciplined base Sector / what NT$215 implies Base sits below the price-implied input Second-order

The story & the five claims

The 10-year story this DCF is built on, plus the five anchor claims that translate the story into model inputs.

The 10-year story

Taiwan Surface Mounting Technology is a mature, cyclical contract assembler — it surface-mounts and stuffs the boards inside Chinese panel makers' LCD/TV sets, DRAM modules, and car dashboards, at a thin 5–8% operating margin it has never sustainably escaped in a decade. In four months the stock doubled-and-then-some on a single story: that a small AI-server, optical-communications, and DDR5 sliver (~5% of revenue, guided to 10–15%) carries higher margins and will re-rate the entire company. Held to the margins and growth it has actually delivered — through-cycle revenue of ~NT$50bn compounding ~4.6% and an inside-band 7.5% operating margin — the disciplined base is worth about NT$173 a share after a 10% governance haircut, roughly a quarter below the NT$215 price. That is not a harsh model: even crediting the full regime-change steelman (a band-breaking 9.0% margin on 7% growth) only reaches ~NT$217, essentially equal to the price, so there is no margin of safety in either direction. The first hard datapoint on the pivot — 1Q26 operating margin of 5.0% and EPS down 26% — pointed the wrong way, and the sell-side's own NT$163 target sits below the tape. This is a good company at the wrong price: the verdict is SELL / avoid, and the place to revisit is a pullback, not the melt-up high.

Two debates worth pressure-testing

Is this a broken business the model is right to discount?

Our view: No — the quality is real. Net cash ~NT$9.7bn (~16% of mcap), ~20% ROIC even at the cycle trough, 10 straight profitable years, a 28-year growing dividend, and a long-tenured family operator. The SELL is not a quality call; it is a price call. At NT$215 you pay ~22× trailing for a low-margin cyclical EMS on the bet that a ~5%-of-revenue AI/optical sliver lifts the margin of the whole company above a band it has held for a decade. Good company, wrong price — the GOOGL-style counter-example, not a value trap.

Does the AI/optical steelman rescue the price?

Our view: Not enough to matter. Credit the regime-change in full — 9.0% year-10 margin (breaking the 5.0–8.5% band) on 7% growth — and intrinsic reaches only ~NT$217 ≈ NT$215, with P(intrinsic < market) still ~61%. So there is no margin of safety even on the bull inputs. On the disciplined band-respecting base it is −24.6% with P(intrinsic < market) = 99.1%, and the entire p5–p95 Monte Carlo fan (NT$144–202) sits below the price. Terminal value is ~71% of EV, so the year-10 margin and β — not cash-in-hand — carry the number. Disciplined entry for a real cushion sits near NT$130–145, not at the high.

CLAIM 01Revenue compounds ~4.6% off the normalized mid-cycle base.growth_high: 4.57% · terminal: 2.0% · NT$50B → ~NT$78BThrough-cycle extrapolation, not the FY2027 analyst peak. FY18→25 secular revenue is +63% even through a −33% drawdown; the non-display mix (DRAM, auto, a small AI/optical sliver) outgrows the soft display core.

CLAIM 02Year-10 operating margin holds 7.5% — INSIDE the decade 5.0–8.5% band.target_op_margin: 7.5% (inside band)The load-bearing claim. Margin has never sustainably escaped 5.0–8.5%; FY25 ran 7.03% and 1Q26 fell to 5.0%. A ~50bp mix tailwind is credited; a double-digit band-break is rejected — the 10-year record gives no basis, and the price requires it.

CLAIM 03Working-capital-heavy but capital-efficient; ROIC stays mid-teens-plus.S2C: 1.30× · terminal ROC 12%Total asset turnover ~0.9–1.0 but PP&E is light vs revenue; reported ROIC 18–26% across the cycle. Terminal ROC faded to 12% — a modest ~3–4pt perpetual spread over the ~8.6% WACC, not a commodity-EMS excess return.

CLAIM 04Mature assembler — margin converges by Year 5, growth decays to terminal by Year 10.year_of_convergence: 5Already at steady state, not young-growth. China panel in-sourcing and single-customer optical concentration cap how long any outsized growth lasts.

CLAIM 05No failure risk; governance handled as a separate 10% haircut.terminal_g: 2.0% · failure: 0% · gov haircut: 10%Net cash, ~20× interest cover, 10 straight profitable years. The 10% haircut captures family entrenchment, China asset/repatriation, and the non-cancellation (employee-transfer) buyback — partly offset by a real, growing dividend.

Where we diverge from sell-side

We find no margin of safety — and the sell-side agrees on direction. The consensus 1-year target (NT$163) sits below the NT$215 price; the stock has run past even its paid analysts. Our disciplined intrinsic (NT$173) is in the same neighbourhood — both independent anchors say the price overshot.
The AI/optical pivot is already IN the price, not on top of it. Most bullish notes frame the pivot as upside layered on the current price. We show that even crediting it in full (9.0% margin, 7% growth) only gets intrinsic to ≈ the market — the optionality is priced, not free.
Cyclical base normalized to mid-cycle, not the peak. The bear "−5.6%/yr EPS" is an artifact of measuring off the 2021/22 super-cycle top; we anchor on FY25's ~NT$50bn, neither the NT$68bn peak nor the NT$45bn trough.
China is a genuine end-market, so it carries real CRP. Unlike export-EMS peers, TSMT's customers ARE Chinese panel makers, so China earns full CRP weight in the WACC (blended ERP ~5.08%) — a heavier discount than a China-located-but-US-serving assembler would get.
β set to 1.30 off a non-suppressed regression. The 5Y regression β (1.25, identical across three sources) is genuinely high after the doubling, not a thin-float artifact — so the "don't trust suppressed Taiwan betas" caveat does not apply; we nudge modestly toward the 1.71 sector relever, which would lower intrinsic further.

Two-sided case — bear anchors

The band holds — which is the base case, not a bear. At the historically-respected 7.5% margin, intrinsic is NT$173 and the stock is ~25% overvalued. The first datapoint (1Q26 margin 5.0%, EPS −26%) supports it. This is the central estimate, and the council pressure-tested it to PASS at NT$215.
The AI/optical ramp is back-half-loaded, certification-gated, and single-customer-concentrated. The 2027 EPS hockey-stick (~NT$19 sell-side) depends on orders moving from pipeline to signed and on certifications reaching mass production. Slippage breaks the entire re-rate that justifies today's price.
China panel in-sourcing. BOE/CSOT could in-source SMT/PCBA as they scale, commoditizing TSMT's display core; optical orders could likewise commoditize on volume against Hon Hai / Wistron.
China asset / FX leakage. 8 mainland plants and ~80% China end-market revenue concentrate profit-repatriation, capital-control, and cross-strait risk — captured in the 10% governance haircut and the China-weighted CRP, but a fat tail if either worsens.
Terminal-heavy, β-sensitive. ~71% of EV is terminal value; a move of β toward the 1.55 sector level (sampled in the MC) compresses intrinsic materially, since the discount rate — not near-term cash — drives the number.

Risks to thesis (tail, not bear case)

Momentum / melt-up can run further (the SELL risk)High

A name that 2.2×'d in four months on an AI narrative can stay irrational and print higher before reverting. Being "~25% right" on value means nothing if it tags NT$280 first. This is an avoid/trim call for a long-only book, not a timing signal — and certainly not a short.

The pivot is real and durableMed

If named customers, signed-vs-pipeline orders, and segment margins confirm the AI/optical/DDR5 mix genuinely re-bases the company above the 5.0–8.5% band, intrinsic moves toward the ~NT$217 steelman — closing the gap. The durability evidence is the single thing that would flip the verdict.

Ex-China demand lowers the discount rateMed

If AI/optical demand routes to US hyperscalers rather than Chinese panel makers, the prospective China-weighted CRP falls and WACC eases — a legitimate upside lever the trailing-80%-China base does not credit.

Net cash under-rewardedMed

Net cash ~NT$9.7bn (~16% of mcap) is a real floor the per-share multiplier discounts. Combined with the cycle-trough downside, it brackets the bear floor near NT$70–100 if AI orders stall — cushioning, but well below today's price.

β / terminal-value sensitivityLow

~71% of EV is terminal value; a β drift toward the 1.55 sector relever (sampled in the MC) compresses intrinsic further. The discount rate, not near-term cash, drives the number.

Family entrenchment / buyback not cancellingLow

31-yr founder-chairman with son succession, ~14% insider, 4 independents on 11 seats; the 2025 buyback is earmarked for employee transfer, not cancellation. Priced via the 10% haircut; offset by a real, growing dividend.

10-year forecast

Revenue NT$52.2B → NT$72.4B over 10y (~4.6% through-cycle ramp); operating margin lifts modestly from ~7.1% toward the inside-band 7.5% target by Y5 — held to the 5.0–8.5% band the company has actually delivered, not the double-digit step-up the price assumes.

Monte Carlo distribution

The entire Monte Carlo fan sits below today's price: the p95 best-case draw tops out at NT$202, still under NT$215, and the median is NT$170 — so P(intrinsic < market) = 99.1%. Even the most favourable correlated stress draw does not reach the tape; the price is outside the distribution, not at its high tail.

Mean NT$171.63 ± NT$17.30/sh, 1000 iterations (0 failed). P(intrinsic < market NT$215.00) = 99.1%.

Even the regime-change steelman doesn't rescue the price

The disciplined base is the published case. But the natural bull objection is: what if the AI/optical/DDR5 mix genuinely re-rates the company above its decade-long band? We ran that steelman in full — and it still doesn't clear the bar.

Disciplined band-respecting base — margin 7.5%, growth 4.6%: intrinsic ≈ NT$172.58 (post-10% gov; pre NT$191.75) vs market NT$215 → −24.6% margin of safety; P(intrinsic < market) = 99.1%, the whole p5–p95 fan (NT$144–202) below the price. This is the published verdict: SELL. The independent council pressure-tested it to PASS at NT$215.
Regime-change steelman — margin 9.0% (band-break), growth 7%: intrinsic only reaches ~NT$217 ≈ market → MoS ~0%, with P(intrinsic < market) still ~61%. Crediting the bull in full barely gets you to fair value, never to a cushion.
Reconciled read: NT$215 is priced for the AI/optical pivot to succeed — no margin of safety in either direction. The asymmetry favours patience: a 20–25% cushion sits near NT$130–145 (band base). Revisit on a pullback, not at the melt-up high. Per the low-rate-currency discipline, the TWD risk-free is headlined at the blended 2.65% (bookends: a 1.13% local-bond floor, a normalized 4.3% high) — but even the most generous rate regime does not close a ~25% gap.

Cost of capital build

Risk-free rate	2.65%
Mature-market ERP	4.23%
Levered β	1.30
Weighted CRP	0.85%
Cost of equity	9.25%
Pre-tax cost of debt (synth Aaa/AAA)	3.05%
D / V	~10%
WACC	8.56%

Full year-by-year DCF

Year	Revenue	Op mgn	EBIT	EBIT(1−t)	Reinvest	FCFF	PV
1	NT$52.16B	7.12%	NT$3.71B	NT$2.91B	NT$1.75B	NT$1.15B	NT$1.06B
2	NT$54.54B	7.22%	NT$3.94B	NT$3.08B	NT$1.83B	NT$1.25B	NT$1.06B
3	NT$57.03B	7.31%	NT$4.17B	NT$3.26B	NT$1.92B	NT$1.35B	NT$1.05B
4	NT$59.64B	7.41%	NT$4.42B	NT$3.46B	NT$2.00B	NT$1.45B	NT$1.05B
5	NT$62.36B	7.50%	NT$4.68B	NT$3.66B	NT$2.10B	NT$1.56B	NT$1.04B
6	NT$64.89B	7.50%	NT$4.87B	NT$3.83B	NT$1.95B	NT$1.88B	NT$1.15B
7	NT$67.19B	7.50%	NT$5.04B	NT$3.98B	NT$1.77B	NT$2.21B	NT$1.26B
8	NT$69.23B	7.50%	NT$5.19B	NT$4.12B	NT$1.57B	NT$2.55B	NT$1.35B
9	NT$70.97B	7.50%	NT$5.32B	NT$4.24B	NT$1.34B	NT$2.90B	NT$1.43B
10	NT$72.39B	7.50%	NT$5.43B	NT$4.34B	NT$1.09B	NT$3.25B	NT$1.50B

Methodology & flags

Damodaran FCFF DCF, 10y explicit + perpetuity, in TWD. Risk-free 2.65% (local-currency government bond). Synthetic credit Aaa/AAA. CRP 0.85% from revenue-weighted country mix × Damodaran 2026 CRPs. Terminal ROIC faded to 12.00%; 10% governance haircut applied post-DCF. Monte Carlo: 1000 iterations. Engine v1.0.0 · result: /Users/valentin/Documents/notes_jiliac_labs/Finance/Damodaran/valuations/tsmt/output/2026-06-02-result.json

R&D cap: OFF · Lease cap: OFF · Failure: OFF · ESO: OFF
Governance haircut: 10% applied post-DCF (NT$191.75 > NT$172.58)
Sensitivity tornado: not run

FCFF DCF · dark theme v3. Generated 2026-06-02T07:18:50 from /Users/valentin/Documents/notes_jiliac_labs/Finance/Damodaran/valuations/tsmt/output/2026-06-02-result.json.