← back to picks

HQ Taipei · Taiwan Reporting TWD Credit synth Aaa/AAA · net cash ~NT$6bn Parent Kindom 冠德 · group ~56% Valuation 2026-06-04 Damodaran FCFF · 1000-iter MC FCFF · Dark v3

Net cash and a real franchise — but the cash a minority gets is already in the price · fairly valued to modestly rich, no margin of safety

FCFF +100% is a NAV ceiling; the dividend a minority can actually realize ≈ market

We do not underwrite the NT$169 perpetual-FCFF print (+100%). It capitalizes NT$6bn of trapped net cash plus 100% of peak FY2025 earnings that a 56%-Kindom-controlled minority — paid ~50%, with no catalyst — never sweeps. Strike the value on the dividend that actually reaches a minority, normalized off the non-repeatable TSMC-fab earnings spike, and fair value is a cash-to-minority DDM of ~NT$80 (range NT$70–95). The market at NT$84.20 is pricing that approximately correctly: a fine ~5%-yield income holding, a WATCH, not the +96% BUY the screener advertised.

SectorEngineering / ConstructionCountry mixTaiwan 100%β / MC σ0.80 levered · ±NT$13/sh (1000 runs)GovernanceKindom-controlled ~56% · RP margin drag · 8% haircutQuality~22% ROE · ~11% ROIC · net cash ~NT$6bn · synth AAAIncomeNT$4.8 cash dividend · ~5% yield · 20-yr record

Intrinsic / share

NT$80.00

cash-to-minority DDM ~NT$80 (range 70–95) · NAV ceiling NT$169 unrealizable

Market / share

NT$84.20

late-May / early-Jun 2026 · TWSE

Margin of safety

-5.0%

vs intrinsic

Enterprise value

NT$17.44B

52.4% terminal

Cost of equity / debt

6.32% / 2.17%

β 0.80 · CRP 0.78%

Terminal ROIC / g

10.00% / 2.00%

spread ~380bp (ROIC 10.00% vs WACC 6.20%)

What it sells, where it sells

Operating segments

Tech-fab constructionTSMC AP7/AP8/F22 fab + 南科物流 jobs — >50% of backlog, the source of the non-repeatable FY25 recognition spike~50%

Public civil / railMRT, infrastructure, public works — long 5yr+ duration, third-party margins 5.5–25.2%~30%

Kindom internal (RP)Captive parent developer work at 2.03–4.63% gross margin — value routed up; shrinking from ~35% (2020) to ~20%~20%

The donut is whole-company revenue at a peak — FY2025 NT$21.5bn is +51% off a stable ~NT$14bn 2020–24 plateau as a cluster of TSMC fab jobs hit their recognition phase at once. The ~20% captive Kindom slice at 2–4.6% gross margin is the structural minority-unfriendly tell: it is modeled in the flows as a permanent ~6.5% operating-margin cap, not bolted on as a discount.

Country mix (revenue-weighted CRP input)

🇹🇼Taiwan100%

Entirely a Taiwan construction story — semiconductor-fab, public civil/rail, and parent-developer work, all domestic. That concentrates the thesis on the Taiwan tech-fab capex cycle and the property/construction cycle under the central bank's credit tightening, with no geographic diversification to cushion a domestic order-intake slowdown.

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 NT$169 perpetual-FCFF print is a NAV ceiling, not a target. It capitalizes 100% of peak FY2025 earnings plus all ~NT$6bn (~NT$46/share) net cash at face value. But Kedge is ~56%-controlled by the Kindom group (冠德, 2520; ~34% direct), pays out only ~50%, and the cash earns ~1.5% on the balance sheet — a minority owns a claim on Kindom's distribution policy, not on the firm. We report NT$169 as the unrealizable ceiling and strike value on the dividend that actually reaches a minority.
FY2025 EPS 9.52 is a recognition peak, not a run-rate. Revenue jumped +51% (NT$14.2bn → 21.5bn) on a wave of TSMC fab recognition; the prior three years all sat at ~NT$14bn. New-order signings are falling (231億 peak 2021 → 152億 in 2025), so the spike unwinds rather than compounds. The DDM is re-struck on a normalized through-cycle EPS of ~NT$7.0–7.8, giving a ~NT$3.4–4.2 cash dividend (vs the NT$4.80 peak).
Related-party margin extraction is the key minority-unfriendly tell. ~20% of revenue is captive Kindom work at 2.03–4.63% gross margin vs 5.5–25.2% on third-party jobs — value routed up to the parent, recurring, already visible in the thin ~6.5% consolidated operating margin. Mitigant: the related-party share is structurally shrinking (~35% in 2020 → ~20%) as tech-fab and public work grow.
Net cash, ~22% ROE, ~11% ROIC, 20-year growing dividend. ~NT$6bn cash vs ~NT$0.3bn debt → net cash ~NT$46/share against the ~NT$84 price; synthetic AAA balance sheet, 0% failure assumption. Twenty unbroken years of dividends, record NT$4.8 cash for FY2025, ~14–16% decade cash-dividend CAGR — the cash genuinely reaches and compounds for minorities. This is the one condition the name passes decisively.
No catalyst with teeth. No Value-Up plan, no buyback-and-cancellation, no holdco/opco merger, no activist disclosure. Nothing forces the gap between NT$169 NAV and the ~NT$80 cash-to-minority value to close. Mild ongoing stock-dividend dilution (~+1.2%/yr) is the opposite of a buyback. Absent a catalyst, realized return ≈ the ~5% dividend yield.

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 input

Sector median

Value impact

01Revenue growth10-year revenue CAGR vs Engineering/Construction median

−3.3% Y1-5DEGROWTH — fade the NT$21.5bn TSMC-fab recognition peak toward a through-cycle ~NT$18bn by yr5; signings falling

~4.5% CAGREngineering/Construction, global cross-sector median

−NT$30

02Operating marginThrough-cycle target vs sector median EBIT margin

6.5%≈ FY25 base 6.67%; the ~20% captive Kindom slice at 2–4.6% GM is a structural cap modeled IN the flow, not bolted on

6.49%Engineering/Construction US median EBIT margin (48 firms)

NT$0

03Sales-to-capitalReinvestment efficiency vs sector median

2.5×Asset-light, negative-WC cycle (-85.8 days) — payables float funds operations; above Te Chang's 2.0

~4.36×Engineering/Construction industry standard

~NT$0

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

2.0%TWD long-run inflation; held ≤ the 2.31% risk-free (Damodaran ceiling)

2.31%TWD risk-free ceiling (50/50 blend); terminal g held below it

NT$0

05Cost of capital10y WACC vs WACC implied by sector-median β

6.20%β 0.80 (floored — 5Y regression 0.27–0.35 is the suppressed-beta trap) · rf 2.31% · CRP 0.78% · net-cash, D/V ~0

~6.0%Damodaran global E&C β_u 0.7593 re-levered at net-cash → 0.776, floored to 0.80

~NT$0

Why we discount this ladder

This ladder produces the perpetual-FCFF NT$169 NAV ceiling, which we explicitly disown as a target. Two artifacts inflate it even after honest flow inputs: (1) the low-yield-TWD WACC capitalizes NOPAT at ~16×, and (2) the engine credits all NT$6bn trapped, controlled cash at face value plus 100% of earnings. The decision number is not this ladder but the cash-to-minority DDM: a normalized ~NT$3.4–4.2 dividend discounted at 6.3–8.5% → fair value ~NT$80 (range NT$70–95), against a NT$84.20 price.

fair ~NT$80

Our override Sector median Override adds value Override subtracts value

The story & the five claims

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

The 10-year story

Kedge is a net-cash, asset-light Taiwan construction contractor inside the Kindom group, which owns ~56%. Its FY2025 NT$21.5bn revenue is a non-repeatable peak — +51% off a stable ~NT$14bn plateau as a cluster of TSMC fab jobs hit their recognition phase at once — recognized at a thin ~6.7% operating margin, with new order signings already falling. The engine, fed honest flows (modest degrowth, a thin ~6.5% margin, faded terminal returns), still prints NT$169 a share, roughly double the price. We do not underwrite that number. It capitalizes ~NT$6bn of trapped cash and 100% of peak earnings that a controlled minority — paid ~50%, with no catalyst to force a payout — never actually receives. The honest read values the dividend that does reach a minority: normalize the earnings off the TSMC-fab spike to ~NT$7.0–7.8, take the ~50% payout to a ~NT$3.4–4.2 cash dividend, and discount it. That lands fair value around NT$80 (range NT$70–95), against a NT$84.20 price — fairly valued to modestly rich. An independent council (2026-06-04, 4-of-5 WATCH) confirmed the screener's +96% was a triple artifact: peak EPS, a low-yield-TWD WACC, and crediting trapped controlled cash at face value. A fine ~5%-yield income holding, not a double.

Two debates worth pressure-testing

If it is net cash with a growing 20-year dividend, why only a WATCH?

Our view: Because the cash backing the NT$169 NAV is trapped: a ~56%-controlled minority cannot sweep ~NT$6bn earning ~1.5% when payout is ~50% and no catalyst forces a distribution. Value what a minority actually gets — a normalized ~NT$3.4–4.2 dividend, discounted — and fair value is ~NT$80. At NT$84.20 you are paying roughly fair for the cash that reaches you. Passes the cash-return condition decisively, but fails the catalyst condition, so the NAV is a ceiling, not a target.

Wasn't the screener flagging +96% upside?

Our view: Yes — and it was a triple artifact the council told us to strip. (1) It anchored on FY2025 peak EPS (9.52), the all-time high on a +51% revenue spike that is not a run-rate. (2) The low-yield-TWD WACC capitalizes NOPAT at ~16×. (3) It credited all NT$6bn of trapped, controlled cash at face value. Fix all three and the +96% collapses to roughly fair. Re-rate triggers to BUY: a payout step-up or special dividend distributing the cash, a holdco action routing value to minorities, or a retreat to ~NT$60–65.

CLAIM 01The decision number is a cash-to-minority DDM ~NT$80; NT$169 is a NAV ceiling.normalized div ~NT$3.4–4.2 ÷ (Ke 6.3–8.5% − g 2.0–2.5%) → ~NT$70–95A ~56%-controlled minority owns a claim on Kindom's distribution policy, not on the firm. Value the dividend that actually reaches it; the NT$169 FCFF print capitalizes trapped cash + 100% of earnings and is unrealizable.

CLAIM 02FY2025 NT$21.5bn revenue is a recognition peak — normalize DOWN, don't compound.growth_high −3.3% Y1-5 · NT$21.5bn peak → ~NT$18bn by yr5+51% off a stable ~NT$14bn 2020–24 plateau on a cluster of TSMC fab jobs; new signings are FALLING (231億 → 152億). The spike unwinds rather than compounds.

CLAIM 03Through-cycle margin is thin ~6.5% — the captive Kindom slice is a structural cap in the flow.target_op_margin 6.5% ≈ FY25 base 6.67% ≈ E&C median 6.49%~20% captive parent work at 2.03–4.63% GM vs 5.5–25.2% third-party — modeled in the flow per the controlled-company rule, not double-counted as a bolt-on discount.

CLAIM 04Asset-light, negative-WC cycle; net cash, terminal ROC faded to 0.10.S2C 2.5× · net cash ~NT$6bn · terminal ROC 0.10 · g 2.0%Negative net operating cycle (-85.8 days) self-funds working capital; ROE ~22% is float-amplified, ROIC only ~11%. Terminal ROC faded to 0.10 (Te Chang precedent), just above terminal WACC.

CLAIM 05No catalyst — passes cash-return (A), fails catalyst (B); β floored against the suppressed-beta trap.β floor 0.80 (regression 0.27–0.35 rejected) · gov haircut 8% · failure 0%A growing 20-yr dividend (A ✅), but no Value-Up, buyback-and-cancel, or holdco action (B ❌). Float ~65–88M of 130.5M, controlled, illiquid → measured β is suppressed; floor to 0.80.

Where we diverge from sell-side

We disown our own screener pick. A council pressure-test (2026-06-04, 4-of-5 WATCH) moved this from a +96% screener BUY to WATCH: the headline upside was a triple artifact (peak EPS + low-yield WACC + trapped cash at face value). Fair value is the cash-to-minority DDM ~NT$80, not +100%.
The NT$169 FCFF is a NAV ceiling, reported but not underwritten. It hands a minority all ~NT$6bn net cash and 100% of earnings; in a ~56%-controlled, ~50%-payout, no-catalyst name, none of that excess is sweepable. Read NT$169 as the unrealizable ceiling, ~NT$80 as the realizable fair value.
Earnings normalized off the FY2025 peak. EPS 9.52 is an all-time high on a +51% TSMC-fab recognition spike with signings already falling. The DDM is struck on a normalized ~NT$7.0–7.8 EPS and ~NT$3.4–4.2 dividend, not the NT$4.80 peak payout (which would inflate fair value to NT$98–129).
Related-party drag modeled in the flow, not bolted on. ~20% captive Kindom work at 2–4.6% GM caps the consolidated margin at ~6.5%; the residual 8% governance haircut carries only leakage/dilution optionality, per the controlled-company rule.
Governance haircut 8% — below Te Chang's 10% and Cathay RE's 15%. The growing, compounding 20-year dividend and the genuine diversification away from the parent (RP 35% → 20%) justify a lighter residual haircut than the trapped-NAV developers.

Two-sided case — bear anchors

The cash a minority gets is already in the price. Absent a payout step-up or special dividend, realized return ≈ the ~5% dividend yield. The ~NT$6bn net cash earns ~1.5% on the balance sheet and is not swept; the gap to NT$169 NAV does not close on its own.
FY2025 is a TSMC-fab recognition peak that cannot compound. Revenue is +51% off a ~NT$14bn plateau as a cluster of fab jobs hit recognition simultaneously; new signings have fallen 231億 → 152億. As the spike unwinds, EPS and the dividend normalize down toward ~NT$7.0–7.8 / ~NT$3.4–4.2.
Related-party margin extraction is recurring and structural. ~20% of revenue is captive Kindom work at 2.03–4.63% gross margin vs 5.5–25.2% third-party — value routed up to the parent, permanently capping the consolidated ~6.5% operating margin. The mitigant (shrinking RP mix) is real but slow.
Thin, fixed-price, cyclical contractor economics. Lump-sum (總價承攬) tech-fab work shifts cost-inflation risk onto Kedge; FY2025 gross margin already fell to 8.5% (from 11–13%), and 26Q1 GM was back to 7.8%. A domestic capex or property-cycle slowdown hits order intake directly with no geographic offset.
Controlled micro-cap with the 0%-downside tell. MC P(intrinsic < market) = 0% on a 0-estimate, illiquid, 56%-controlled name is a sign of mis-specification (the un-anchored firm-level FCFF spread), not a riskless double. Mild stock-dividend dilution (~+1.2%/yr) is the opposite of a buyback.

Risks to thesis (tail, not bear case)

Trapped cash / no catalystHigh

~NT$6bn net cash (~NT$46/sh) earns ~1.5% and is not swept; payout ~50%, no Value-Up, buyback-and-cancel, or holdco action. The gap between NT$169 NAV and ~NT$80 cash-to-minority value does not close without a catalyst.

Peak-earnings normalizationHigh

FY2025 EPS 9.52 is an all-time high on a non-repeatable +51% TSMC-fab recognition spike; signings falling (231億 → 152億). As the peak unwinds, EPS and the dividend normalize down toward ~NT$7.0–7.8 / ~NT$3.4–4.2.

Related-party margin extractionMed

~20% captive Kindom work at 2.03–4.63% GM vs 5.5–25.2% third-party — recurring value routed to the parent, capping the consolidated ~6.5% margin. Mitigated only slowly by the shrinking RP mix (~35% → ~20%).

Thin fixed-price margin revertsMed

Lump-sum tech-fab work shifts cost-inflation risk to Kedge; gross margin fell to 8.5% (from 11–13%), 26Q1 back to 7.8%. Steel/labour inflation plus a ~53k-worker shortage could push the through-cycle margin below 6.5%.

Taiwan construction cycleMed

100% domestic; exposed to the TSMC capex cycle and a property/construction market under central-bank credit tightening. A domestic capex slowdown hits order intake with no geographic offset.

Stock-dividend dilutionLow

Annual stock dividends (盈餘轉增資, ~+1.2%/yr) mildly dilute share count — pro-rata and small, but the opposite of a buyback-and-cancellation that would route value to minorities.

10-year forecast

Revenue NT$20.8B → NT$18.1B over 10y as the FY2025 NT$21.5bn TSMC-fab recognition peak fades toward a through-cycle mid-point (−3.3% Y1-5); operating margin held thin at 6.5% (≈ the E&C median and the FY25 base, with the captive Kindom slice capping it). This is the honest flow path behind the NT$169 print we treat as a NAV ceiling, not a target.

Monte Carlo distribution

The Monte Carlo spans NT$152 (p5) to NT$194 (p95) and every draw clears the NT$84.20 price, so P(intrinsic < market) = 0% — but that band only stress-tests the perpetual-FCFF NAV ceiling we disown, not the thesis. A 0%-downside read on a 0-estimate, controlled, illiquid micro-cap is mis-specification, not a riskless double. The thesis stands on the cash-to-minority DDM: a normalized ~NT$3.4–4.2 dividend → fair value ~NT$80, roughly the market price.

Mean NT$171.16 ± NT$12.69/sh, 1000 iterations (0 failed). P(intrinsic < market NT$84.20) = 0.0%.

Cost of capital build

Risk-free rate	2.31%
Mature-market ERP	4.23%
Levered β	0.80
Weighted CRP	0.78%
Cost of equity	6.32%
Pre-tax cost of debt (synth Aaa/AAA)	2.71%
D / V	~3%
WACC	6.20%

Full year-by-year DCF

Year	Revenue	Op mgn	EBIT	EBIT(1−t)	Reinvest	FCFF	PV
1	NT$20.79B	6.64%	NT$1.38B	NT$1.10B	NT$-284M	NT$1.38B	NT$1.30B
2	NT$20.10B	6.60%	NT$1.33B	NT$1.06B	NT$-274M	NT$1.33B	NT$1.18B
3	NT$19.44B	6.57%	NT$1.28B	NT$1.02B	NT$-265M	NT$1.28B	NT$1.07B
4	NT$18.80B	6.53%	NT$1.23B	NT$978M	NT$-257M	NT$1.23B	NT$970M
5	NT$18.17B	6.50%	NT$1.18B	NT$940M	NT$-248M	NT$1.19B	NT$880M
6	NT$17.77B	6.50%	NT$1.15B	NT$920M	NT$-163M	NT$1.08B	NT$754M
7	NT$17.56B	6.50%	NT$1.14B	NT$910M	NT$-84M	NT$994M	NT$651M
8	NT$17.54B	6.50%	NT$1.14B	NT$910M	NT$-8M	NT$919M	NT$565M
9	NT$17.70B	6.50%	NT$1.15B	NT$920M	NT$66M	NT$854M	NT$493M
10	NT$18.06B	6.50%	NT$1.17B	NT$939M	NT$142M	NT$797M	NT$433M

Methodology & flags

Damodaran FCFF DCF, 10y explicit + perpetuity, in TWD. Risk-free 2.31% (local-currency government bond). Synthetic credit Aaa/AAA. CRP 0.78% from revenue-weighted country mix × Damodaran 2026 CRPs. Terminal ROIC faded to 10.00%; 8% governance haircut applied post-DCF. Monte Carlo: 1000 iterations. Engine v1.0.0 · result: /Users/valentin/Documents/notes_jiliac_labs/Finance/Damodaran/valuations/kedge/output/2026-06-04-result.json

R&D cap: OFF · Lease cap: OFF · Failure: OFF · ESO: OFF
Governance haircut: 8% applied post-DCF (NT$183.49 > NT$168.81)
Sensitivity tornado: not run

FCFF DCF · dark theme v3. Generated 2026-06-05T06:20:51 from /Users/valentin/Documents/notes_jiliac_labs/Finance/Damodaran/valuations/kedge/output/2026-06-04-result.json.