← back to picks

HQ New Taipei · Taiwan Reporting TWD Credit synth Aaa/AAA · net cash Valuation 2026-05-31 Damodaran FCFF · 1000-iter MC FCFF · Dark v3

Cheap, but the verdict is margin- and rate-contingent · +114% at base, +13% at stress

HOLD — survives a margin cut OR rate normalization, not both

Base-case intrinsic NT$408/share vs market NT$190.52 (+114%). But the load-bearing 25% terminal operating margin is mildly optimistic: at the honest 22% the upside is still +89% at Taiwan's 1.13% base rate yet only +13% at a normalized 4.3% stress rate (NT$215). Terminal value is ~163% of EV under stress, so the terminal margin is the valuation — and margin is already rolling over (Q1-26 gross margin 48.9%, guided down up to 4pts).

SectorMemory connectors · ElectronicsCountry mixUS 40% · CN 20% · KR 15%β / MC σ1.30 levered · ±NT$46/sh (1000 runs)GovernanceNo haircut · 0.0QualityDebt-free net cash · 17 straight dividend yearsShare>50% SO-DIMM / M.2

Intrinsic / share

NT$408.06

Market / share

NT$190.52

2026-05-31 close · TPEx

Margin of safety

+114.2%

vs intrinsic

Enterprise value

NT$34.84B

75.2% terminal

Cost of equity / debt

7.38% / 1.22%

β 1.30 · CRP 0.58%

Terminal ROIC / g

10.00% / 1.13%

spread ~262bp (ROIC 10.00% vs WACC 7.38%)

What it sells, where it sells

Operating segments

SO-DIMM + M.2 (legacy core)>50% share each, ~60% of revenue combined; commoditizing (Deren, Luxshare on price), SO-DIMM guided −10% on notebook cuts~59%

R-DIMM / Long-DIMM (server)DDR5 server memory, into two CSPs; 2026 R-DIMM units est. +100%, Long-DIMM +30%+ — the AI-server slice that ships today~11%

Metal Bar micro-stampingsSole supplier to a US handset/notebook major (incl. foldables); 2026 volume +20%+ — premium-ASP optionality~15%

Other connectors + SOCAMM2 (future)LPCAMM2 laptop sockets (Lenovo-concentrated); flagship NVIDIA SOCAMM2 is 2027+ with Argosy at ZERO revenue vs three incumbents~15%

What ships today is the legacy SO-DIMM/M.2 core (~59%, commoditizing and guided lower) plus a thin DDR5 R-DIMM server slice (~11%). The AI-server thesis the cheapness is supposed to ignore is mostly optionality: the flagship NVIDIA SOCAMM2 socket has three incumbents ahead and Argosy at zero revenue, and 2027+ delivery. The terminal margin carries the valuation, not this year's mix.

Country mix (revenue-weighted CRP input)

🇺🇸United States40%

🇨🇳China20%

🇰🇷South Korea15%

🇹🇼Taiwan15%

🇯🇵Japan10%

Revenue routes through US (40%) and Asian memory hubs (CN 20%, KR 15%, TW 15%, JP 10%) by end-customer geography, so the CRP is light. The earnings risk is not geographic but concentration: growth is bound to AI-server ramp and a single US handset customer for the premium Metal Bar parts, and a strong TWD compresses the reported NT$ numbers the market anchors on.

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.

Record FY25 — but FY25's 31.5% EBIT margin is a memory-cycle peak. 2025 revenue NT$4.0bn (+19.4%), net profit NT$1.143bn (+12.8%), EPS NT$12.69 — all records, and a NT$10 dividend (~6.3% yield at ~NT$158). The base case fades that 31.5% peak to a 25% terminal margin held to perpetuity. The honest through-cycle norm is lower: 2018–2022 averaged ~24%, and the recurring annual trough is 20.7%.
Margin is already rolling over. Q1-26 gross margin was 48.93%, with management guiding it down up to 4 points; SO-DIMM revenue is guided down ~10% on notebook-shipment cuts. Q1-26 EPS NT$2.82 fell 4.8% YoY and 12.4% QoQ. Management calls Q1 the year's trough, but the roll has begun before the AI ramp is proven.
>50% share in the commoditizing legacy lines; AI-server is mostly optionality. SO-DIMM and M.2 each carry >50% share (~60% of revenue combined) and are commoditizing as Deren and Luxshare compete on price. What ships into AI servers today is DDR5 R-DIMM (~11%) and LPCAMM2 laptop sockets (Lenovo-concentrated). The flagship NVIDIA SOCAMM2 socket has three incumbents ahead — Amphenol, Foxconn, Lotes — and Argosy at zero revenue, 2027+.
Debt-free net cash, 17 straight dividend years. Screen D/E 0.21, almost debt-free, with NT$1.928bn cash and a clean balance sheet. A Vietnam plant entered volume production in Q1-26 as a China+1 second base. The balance-sheet quality is real; it is the terminal margin, not solvency, that the valuation hangs on.
Terminal value is ~163% of EV under stress (145% at base). The terminal margin is not a tail assumption — it is the valuation. The cushion is regime-dependent, not margin-dependent: Argosy survives a margin cut to its historical trough OR a Taiwan-rate normalization, but not both at once.

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

Honest / sector

Value impact

01Revenue growth10-year revenue CAGR off the AI-server ramp

15% Y1-5Fading to 1.13% terminal — R-DIMM/Long-DIMM + Metal Bar; SOCAMM2 excluded (zero revenue)

~6% CAGRElectronics sector median, global cross-sector

+NT$60

02Terminal operating marginThe load-bearing input — TV is ~163% of EV at stress

25.0%Faded from the 31.5% FY25 peak — mildly OPTIMISTIC: through-cycle norm ~24%, recurring trough 20.7%

~22%The honest terminal margin — above the trough, below the peak-contaminated average

−NT$47

03Sales-to-capitalReinvestment efficiency Y1-10

1.70×Connector tooling + Vietnam plant build; held flat Y1-10

~2.0×Electronics industry standard

−NT$6

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

1.13%Pinned at the TWD 10y government-bond yield — Damodaran's stable-growth ceiling

1.13%TWD risk-free, identical — no override

NT$0

05Cost of capitalBase 7.38% vs normalized stress 10.55%

7.38%β 1.30 (sector-aware, anti-suppression) over the 0.42 OTC regression · CRP 0.58% · debt-free

10.55%Stress: rf 4.3%, terminal g 2.8% — Taiwan-rate normalization

−NT$166

The two swing factors

The base case nets to NT$408 (+114%). Two corrections each cut hard: the honest 22% margin takes it to NT$361 (+89%), and a Taiwan-rate normalization to NT$242 (+27%). Apply both — 22% margin AT the 10.55% stress rate — and intrinsic falls to NT$215 (+13%). The breakeven margin at stress is ~19.3%, which sits right on the all-time annual trough (20.7%). The thesis survives either stress alone, not both.

−NT$193

Our base input Honest / stress Adds value 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

Argosy is the >50%-share memory-connector maker the AI-server cycle is supposed to re-rate — SO-DIMM and M.2 leadership, debt-free net cash, and seventeen straight dividend years, at ~15x earnings on a record FY25. The base case fades the 31.5% FY25 EBIT peak to a 25% terminal operating margin held to perpetuity, and that single number — because terminal value is ~163% of enterprise value under stress — is essentially the whole valuation. The honest read is that 25% is mildly optimistic, not conservative: the pre-peak through-cycle norm is ~24%, the recurring annual trough is 20.7%, and margin is already rolling over (Q1-26 gross margin 48.9%, guided down up to 4 points; SO-DIMM guided −10%). At the defensible 22% terminal margin the stock is still +89% at Taiwan's 1.13% base rate — but only +13% once rates normalize to 4.3%, with the stress breakeven margin (~19.3%) riding right on the all-time floor. The AI-server ramp that is meant to underwrite the premium is mostly optionality: the flagship NVIDIA SOCAMM2 socket has three incumbents ahead and Argosy at zero revenue. The call is therefore a HOLD — cheap, but contingent. It survives a margin cut OR a rate normalization, not both at once; the genuine anchor is the conservative NT$215.

Two debates worth pressure-testing

Is the 25% terminal margin conservative, as filed?

Our view: No — it is mildly optimistic. The defense cited the ~25.4% 8-year average, but that average is peak-contaminated: 2018–2022 averaged ~24% and 2023–2025 averaged ~29.8%, and the step-up is the DDR5/AI mix shift whose permanence is the open question. The honest terminal margin is ~22% — above the 20.7% recurring trough, below the peak. At 22% the base-rate upside is +89%, so the BUY does not depend on 25% holding; the stress-rate cushion does.

Why HOLD and not BUY, when base MoS is +114%?

Our view: Because the comfortable margin of safety lives almost entirely in Taiwan's 1.13% base-rate regime. Five of six margin×rate cells survive, but the call bifurcates on rates, not margins: at the honest 22% margin the stress MoS is a thin +13% (NT$215), with breakeven ~19.3% on the all-time floor, and it breaks outright in the double-stress corner (18% margin AND 4.3% rate → −6%). That is a macro bet as much as a business bet — hence HOLD, with NT$215 as the conservative anchor.

CLAIM 01Revenue compounds 15% Y1-5 on the server ramp, fading to 1.13%.growth_high: 15% · terminal: 1.13% · NT$4.1B → NT$11.5BR-DIMM units est. +100% and Long-DIMM +30%+ into two CSPs, plus Metal Bar +20%+. Excludes the flagship SOCAMM2 socket, where Argosy has zero revenue vs three incumbents.

CLAIM 02Terminal operating margin 25% — and this is the load-bearing call.target_op_margin: 25% from Y5Faded from the 31.5% FY25 peak. Honest read ~22% (norm 24%, trough 20.7%); margin already rolling over. Terminal value is ~163% of EV at stress, so this number IS the valuation.

CLAIM 03Reinvestment stays connector-/Vietnam-build intensive.S2C: 1.70× Y1-5 and Y6-10Connector tooling plus the Vietnam China+1 plant ramp. Held flat below the electronics standard — a conservative drag, not a stretch.

CLAIM 04Margin and growth converge by Year 5 (2030).year_of_convergence: 5Ramp and margin maturation complete by year 5; no further expansion in years 6-10, with growth fading linearly to the 1.13% terminal rate.

CLAIM 05No failure risk — debt-free net cash, no governance haircut.terminal_g: 1.13% · failure: 0% · gov haircut: 0%Debt-free, NT$1.93bn cash, synth Aaa/AAA, 17 straight dividend years; terminal ROIC faded to 10% from the engine's ~19%. The risk is the terminal margin and Taiwan rates, not solvency.

Where we diverge from sell-side

The verdict is HOLD, not the base-case BUY. Base MoS is +114%, but the cushion is regime-dependent: at the honest 22% terminal margin the stock is +89% at Taiwan's 1.13% base rate and only +13% (NT$215) once rates normalize to 4.3%. It survives a margin cut OR a rate normalization, not both — so the chip is HOLD and the conservative anchor is NT$215.
25% terminal margin treated as mildly optimistic, not conservative. The filed defense leaned on the peak-contaminated ~25.4% 8-year average; the honest through-cycle margin is ~22% (norm 24%, recurring annual trough 20.7%). Margin is already rolling over — Q1-26 gross margin 48.9%, guided down up to 4pts; SO-DIMM guided −10%.
Terminal value is ~163% of EV at stress, so the terminal margin IS the valuation. The rate regime just rescales the discount on that same slug. "Rates matter more than margin" is misleading — both feed the one perpetuity that carries the entire number.
The AI-server ramp is mostly optionality, not base. What ships today is DDR5 R-DIMM (~11%) and LPCAMM2 (Lenovo-concentrated). The flagship NVIDIA SOCAMM2 socket has three incumbents ahead and Argosy at zero revenue, 2027+. The base credits the explicit ramp, not the flagship win.
β set to 1.30 by sector-anti-suppression, not the 0.42 OTC regression. The thin-trade TPEx small-cap regression β is implausibly suppressed; 1.30 lifts WACC honestly to 7.38% rather than flattering the valuation with a low β. Governance haircut is 0% — debt-free, 17 dividend years, clean structure.

Two-sided case — bear anchors

The double-stress corner breaks the thesis. 18% terminal margin (the all-time quarterly trough) AND Taiwan rates normalizing to ~4.3% simultaneously → NT$178, a −6% MoS (mild overvaluation). Under either stress alone it holds; the failure is the simultaneous corner.
Margin reverts to the historical trough. If the DDR5/AI mix shift does not permanently offset cyclicality and active commoditization, the 25% terminal margin reverts toward 20.7% (annual trough) or 18.4% (worst quarter). Each 1pp costs ~NT$15.8 at base rates and ~NT$9.1 at stress.
Rate normalization compresses the perpetuity. With terminal value ~163% of EV at stress, a move from the 1.13% risk-free toward 4.3% takes intrinsic from NT$408 to NT$242 at 25% margin (and to NT$215 at the honest 22%). The genuine risk to underwrite is Taiwan's low-rate regime persisting.
SOCAMM2 / customer-concentration miss. The flagship NVIDIA socket has three incumbents ahead and Argosy at zero revenue; premium Metal Bar parts depend on a single US handset customer. If the AI/server ramp slips, the explicit slice that justifies 15% growth shrinks and the optionality stays optionality.
TWD strength compresses reported earnings. Argosy reports in NT$; a structurally strong TWD shrinks the reported numbers the market anchors on even at flat USD volumes, and gold/raw-material price swings pressure the gross margin the whole valuation rests on.

Risks to thesis (tail, not bear case)

Terminal margin reverts below 22%High

The load-bearing input. 25% is faded from a 31.5% cycle peak; the honest norm is ~22% and the recurring trough 20.7%. Margin is already rolling over (Q1-26 GM 48.9%, guided −4pts). Each 1pp ≈ NT$16 at base / NT$9 at stress.

Taiwan rate normalizationHigh

~163% of EV is terminal at a 1.13% risk-free. A move to ~4.3% takes intrinsic to NT$242 (25% margin) or NT$215 (22%). The MoS lives in the low-rate regime — this is the binding contingency, hence HOLD.

SOCAMM2 / AI-ramp shortfallMed

Flagship NVIDIA SOCAMM2 socket has three incumbents ahead and Argosy at zero revenue (2027+). If the explicit R-DIMM/Long-DIMM/Metal Bar ramp slips, the 15% growth slice that the case credits shrinks.

Legacy commoditizationMed

SO-DIMM/M.2 (~60% of revenue) are commoditizing as Deren and Luxshare compete on price; SO-DIMM guided −10% on notebook cuts. Price erosion in the core base directly pressures the terminal margin.

TWD strength / raw-material costLow

A strong TWD compresses reported NT$ earnings at flat USD volume; gold and metal feedstock swings pressure gross margin. Cyclical headwind, already partly visible in the FY25 optics.

Customer / liquidity concentrationLow

Premium Metal Bar parts depend on a single US handset customer; TPEx small-cap with ~6% foreign holding means thin liquidity and coverage. Tail risk, not base case.

10-year forecast

Revenue NT$4.77B → NT$11.49B over 10y (15% Y1-5 on the server ramp, fading to 1.13% terminal). Operating margin fades from a 31.5% FY25 peak through ~30% Y1 to a 25% terminal held from Y5 — the load-bearing call; the honest through-cycle margin is closer to 22%.

Monte Carlo distribution

Across 1,000 correlated draws the base-case p5 still sits at NT$336 — far above today's NT$190.52. But the Monte Carlo holds the 25% terminal margin and the 1.13% base rate fixed; it does not sample the rate-normalization or honest-22%-margin regimes where the real contingency lives. The honest verdict is HOLD, anchored on NT$215.

Mean NT$406.82 ± NT$45.60/sh, 1000 iterations (0 failed). P(intrinsic < market NT$190.52) = 0.0%.

⚠ Active diagnostic: terminal_growth (0.0113) >= risk_free_rate (0.0113); Damodaran's stable-growth ceiling is the risk-free rate

Cost of capital build

Risk-free rate	1.13%
Mature-market ERP	4.36%
Levered β	1.30
Weighted CRP	0.58%
Cost of equity	7.38%
Pre-tax cost of debt (synth Aaa/AAA)	1.53%
D / V	~0%
WACC	7.38%

Full year-by-year DCF

Year	Revenue	Op mgn	EBIT	EBIT(1−t)	Reinvest	FCFF	PV
1	NT$4.77B	30.19%	NT$1.44B	NT$1.14B	NT$366M	NT$775M	NT$721M
2	NT$5.49B	28.89%	NT$1.58B	NT$1.26B	NT$421M	NT$834M	NT$724M
3	NT$6.31B	27.59%	NT$1.74B	NT$1.38B	NT$484M	NT$895M	NT$723M
4	NT$7.25B	26.30%	NT$1.91B	NT$1.51B	NT$557M	NT$955M	NT$718M
5	NT$8.34B	25.00%	NT$2.09B	NT$1.65B	NT$640M	NT$1.01B	NT$709M
6	NT$9.36B	25.00%	NT$2.34B	NT$1.86B	NT$600M	NT$1.26B	NT$824M
7	NT$10.25B	25.00%	NT$2.56B	NT$2.04B	NT$521M	NT$1.52B	NT$933M
8	NT$10.93B	25.00%	NT$2.73B	NT$2.18B	NT$403M	NT$1.78B	NT$1.03B
9	NT$11.36B	25.00%	NT$2.84B	NT$2.27B	NT$251M	NT$2.02B	NT$1.11B
10	NT$11.49B	25.00%	NT$2.87B	NT$2.30B	NT$76M	NT$2.22B	NT$1.16B

Methodology & flags

Damodaran FCFF DCF, 10y explicit + perpetuity, in TWD. Risk-free 1.13% (local-currency government bond). Synthetic credit Aaa/AAA. CRP 0.58% from revenue-weighted country mix × Damodaran 2026 CRPs. Terminal ROIC faded to 10.00%; 0% governance haircut applied post-DCF. Monte Carlo: 1000 iterations. Engine v1.0.0 · result: /Users/valentin/Documents/notes_jiliac_labs/Finance/Damodaran/valuations/argosy/output/2026-05-31-result.json

R&D cap: OFF · Lease cap: OFF · Failure: OFF · ESO: OFF
Sensitivity tornado: not run

FCFF DCF · dark theme v3. Generated 2026-05-31T20:00:59 from /Users/valentin/Documents/notes_jiliac_labs/Finance/Damodaran/valuations/argosy/output/2026-05-31-result.json.