← back to picks

HQ Hsinchu · Taiwan Reporting TWD Founder Miin Wu (Chair+CEO, ~age 77) Ownership dispersed / institutional Valuation 2026-06-02 Damodaran FCFF · 1000-iter MC FCFF · Dark v3

A peak-cycle memory price-shock priced as a permanent monopoly · −90% to base DCF

SELL — overvalued in every scenario, every β, every MC path (P(intrinsic<market)=100%)

Base intrinsic NT$14.4/share (post 8% governance haircut; pre NT$15.7) vs market NT$149.5. The stock has 4×'d on a genuine but small low-density-eMMC supply shock (big-4 exiting; Q1'26 op margin 18.5% vs a −13% FY25). But even the full bull case — a durable specialty-analog monopoly at 17% margins, 10% growth, a real 14% terminal ROC — is worth only NT$39 (−74%). A reverse-DCF says the price requires a ~55-60% operating margin held for a decade — roughly 2× Macronix's all-time-best gross margin.

SectorSemiconductor (memory IDM)Country mixCN 36% · TW 24% · JP 20% · DE 12% · US 7%β / WACC1.61 regression · WACC 9.73% (stress 11.4% @ 1.98)GovernanceFounder Chair+CEO ~77 · 8% haircutCyclicalityOp margin +29% to −29% over 15y · 10 loss quarters into Q1'26Reverse-DCFprice implies ~58% perpetual op margin

Intrinsic / share

NT$14.41

post 8% gov · pre NT$15.66

Market / share

NT$149.50

early-June 2026 · TWSE · 52wk NT$18-178 (+660%)

Margin of safety

-90.4%

vs intrinsic

Enterprise value

NT$37.21B

90.8% terminal

Cost of equity / debt

10.28% / 2.41%

β 1.61 · CRP 0.72%

Terminal ROIC / g

6.50% / 2.00%

spread ~-323bp (ROIC 6.50% vs WACC 9.73%)

What it sells, where it sells

Operating segments

NOR Flash — #1 globally (~17% share)58% of revenue, +47% YoY on AI/datacenter pull; ~38% is high-value auto/industrial/medical. The real top-line driver.~58%

NAND / eMMC — the shortage story30% of revenue (was 11% a yr ago), +382% YoY; sole low-density MLC supplier as the big-4 exit. Small TAM (~US$1-2bn), price-spiking.~30%

ROM — structural declineNintendo (via MegaChips) Switch-1 EOL; Switch-2 did not adopt Macronix. Was 34% of revenue in 2023.~8%

FoundryFlat, non-core.~4%

Q1'26 product mix. NOR Flash (58%, #1 globally) is the real top-line driver, tightening on AI/datacenter; NAND/eMMC (30%) is the shortage story — sole low-density MLC supplier as Samsung/Hynix/Micron/Kioxia exit, but a small ~US$1-2bn niche and a price spike, not durable volume. ROM (8%) is in structural decline (Switch-1 EOL). The bull blog figures (revenue to NT$305bn) are physically impossible at the hard ~25k→30k wafer/month capacity ceiling.

Country mix (revenue-weighted CRP input)

🇨🇳China36%

🇹🇼Taiwan24%

🇯🇵Japan20%

🇩🇪Germany12%

🇺🇸United States7%

Revenue-weighted by end-customer / shipment-destination geography (FY2025 audited segment note), not fab location — all fabrication is in Hsinchu, Taiwan, whose single-site risk is carried in the governance discount. Asia-ex-Japan (~39%) proxied to China; Japan (20%, ROM via MegaChips→Nintendo) is structurally declining. All keys low-CRP, so the ~0.72% blended add-on is insensitive to the split.

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 verdict is SELL — and it is robust, not an artifact. Base intrinsic NT$14.4 vs NT$149.5. Every scenario (bear NT$5 · base NT$14 · bull NT$39), every β (1.61→1.98), every 2×2 margin×WACC cell (NT$2.5–25.6), and all 1,000 Monte Carlo draws land far below the price. P(intrinsic<market)=100%.
This is a cyclical at a peak, priced as a monopoly. Macronix has swung from +29% to −29% operating margin over 15 years and posted 10 consecutive loss quarters before Q1'26. The base year is normalized to an 8% mid-cycle margin (discarding both the −2% TTM and the 18.5% Q1'26 spike).
The reverse-DCF is the headline. To justify NT$149.5 at WACC 9.7% you need a ~55-60% perpetual operating margin — roughly 2× the best gross margin (53%, 2010) and ~2.5× the peak operating margin (29%) the company has ever achieved. Even granting 12% perpetual growth, ~47%.
Cross-checks all confirm. 23× peak-cycle EPS (NT$6.48, FY2021); no net-cash floor (slight net debt, ≈ −NT$3/share); market cap is 7× the DCF enterprise value.
The eMMC story is real but bounded. Last-supplier-standing economics are genuine, but the niche is ~US$1-2bn, capacity is hard-capped until H1'27, and the filings themselves flag that the supply cliff can be filled by TLC migration / big-4 re-entry over 3-5 years.

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

Note

01Revenue growth10y CAGR vs Semiconductor median

~5% CAGR7% Y1-5 (NOR + specialty NAND, ROM drag, capacity-capped) fading to 2.0% TWD inflation — council-moderated from 9%

~6% CAGRSemiconductor global cross-sector median

~in line

02Operating marginYear-10 target vs the 15y envelope

11.0%Structural step-up over the low-single-digit through-cycle average, but ≪ the 18.5% Q1'26 spot / 22% prior peak — the load-bearing lever

27%Semiconductor median (fabless-logic-heavy — the wrong benchmark for a commodity memory IDM)

+NT$11 → 16%

03Sales-to-capitalReinvestment efficiency

0.70×The company's own capital-intensive through-cycle asset turnover (0.34-0.71×) — two fabs, not fabless

1.11×Semiconductor median (fabless-skewed)

capital-heavy

04Terminal ROCPerpetual excess return

6.5%≈ terminal cost of capital → value-NEUTRAL growth (council fix). The eMMC moat is real near-term but assumed competed away in perpetuity

~14%Engine triple if a moat were assumed durable

value-inelastic

05Cost of capitalWACC vs sector-relevered β

9.73%β 1.61 (5Y regression, Yahoo+StockAnalysis) · rf 2.31% TWD blend · CRP 0.72% — the CONSERVATIVE pin for a sell

11.4%Damodaran Semiconductor β_u 1.869 re-levered → 1.98 → WACC 11.4% (even more overvalued)

−NT$1.8 @ 1.98

Net read

The NT$135/share gap to price is not an override artifact. Using the (wrong, fabless-skewed) 27% sector margin still only lands ~NT$60 (−60%). The price requires a ~58% perpetual operating margin — 2× the all-time-best gross margin. Every lever has been stress-tested; the SELL survives all four corners and the full bull case.

−90% MoS

Our input Sector median 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 model inputs.

The 10-year story

Macronix is a deeply cyclical Taiwanese non-volatile-memory IDM — #1 in NOR Flash, a rising NAND/eMMC franchise, and a declining Nintendo-linked ROM business. After 10 consecutive loss quarters it swung violently to profit in Q1'26 (gross margin 40.8%, operating margin 18.5%) on a low-density MLC/eMMC supply shock: Samsung, SK Hynix, Micron and Kioxia are permanently exiting the 4-32Gb MLC niche, leaving Macronix the last supplier standing and letting eMMC contract prices double. The stock 4×'d. The structural story is genuine — but it sits on a base that has destroyed capital across two 15-year troughs, the niche is small (~US$1-2bn), and capacity is hard-capped until H1'27. Normalizing the base to a mid-cycle 8% margin and converging to a structural-but-modest 11% year-10 margin, the FCFF DCF values the equity at ~NT$14/share — roughly a tenth of the NT$149.5 price. The market is pricing the peak-cycle shortage as a permanent monopoly.

Two debates worth pressure-testing

Is the 18.5% Q1'26 margin a new regime, or a price spike that mean-reverts?

Our view (and the council's clash): The whole valuation reduces to this one probability — does the last-supplier rent survive Macronix's own H1'27 capacity and eventual TLC/big-4 re-entry? The bull (Expansionist) reads it as a Texas-Instruments-style specialty-analog monopoly (17%+ margins for a decade); the bear reads it as a 12-18-month spike reverting to the mid-single-digit through-cycle average. We model 11% in the base — but it barely matters: even the full bull case (17%) is worth NT$39, still −74%.

Did normalizing the base "manufacture" the SELL?

No. The reverse-DCF removes the normalization entirely and asks what margin the price implies: ~55-60%, roughly 2× the all-time-best gross margin. And the 2×2 grid shows the SELL survives at margins from 6% to 16% and βs from 1.61 to 1.98. The conclusion is independent of the base-year choice.

CLAIM 01Revenue grows ~7% Y1-5, fading to 2.0% TWD inflation.growth_high: 7% · terminal: 2.0% · NT$33.2B → ~NT$47BStructural NOR (AI/datacenter) + specialty NAND growth off an already-elevated FY26 run-rate, partly offset by ROM decline. Hard-bounded by the ~25k→30k wafer/month capacity ceiling (equipment slips to H1'27). Council-moderated from 9% — the "reclaim the 2021 peak" framing was incoherent with a mid-cycle margin.

CLAIM 02Year-10 operating margin steps up to 11% — not the 18.5% spot.target_op_margin: 11% by Y5 (base normalized at 8%)A real structural step-up over the low-single-digit through-cycle average, earned by the sticky specialty-memory mix — but well below the 18.5% Q1'26 spot and 22% prior peak, because the supply cliff partially fills over 3-5 years. THE load-bearing input; the bear/bull fan spans 6%/17%.

CLAIM 03Reinvestment at the company's capital-intensive 0.70× sales-to-capital.S2C: 0.70× both phasesMacronix is a two-fab IDM, not a fabless designer; its through-cycle asset turnover is 0.34-0.71×, not the fabless-skewed 1.11 sector figure. Using the sector number would manufacture a fictitious terminal ROC.

CLAIM 04Terminal ROC = terminal cost of capital — value-neutral, no perpetual moat.override_roc: 6.5% ≈ terminal WACC · terminal_g: 2.0%Council fix: don't assert perpetual value destruction (the raw 6.2% triple) on a firm that just gained pricing power, nor a permanent monopoly. Value-neutral is the honest Damodaran default for a temporary moat competed away. Value-inelastic regardless.

CLAIM 05Small failure tail; founder-succession governance haircut.failure: 3% · gov haircut: 8% · net debt ≈ −NT$3/shTwo 3-year loss troughs in 15 years and a suspended dividend justify a 3% failure tail, offset by positive operating cash flow through the FY25 loss. 8% governance haircut: NOT a controlled company (dispersed institutional ownership), but a founder Chair+CEO at ~age 77 with boilerplate succession + chronic sub-cost-of-capital reinvestment. No net-cash floor under the price.

Where we diverge from sell-side

Market NT$149.5 vs base DCF NT$14.4 — a ~10× gap. Not driven by a single contestable override: the reverse-DCF implied margin (~58%) and the all-corners grid both confirm it independently of the normalization.
The bull case is fully credited and still loses. A durable specialty-analog monopoly (17% margins, 10% growth, 14% terminal ROC, 6% governance) is worth NT$39 — still −74%. You have to believe in margins above the all-time peak, held forever, to touch the price.
The β pin is the conservative choice. 1.61 (regression) gives a lower WACC and a higher value than the sector-relevered 1.98; the SELL is robust at the friendlier number, and only deepens at the stress (NT$12.6).

Two-sided case — bear anchors

The shortage mean-reverts when Macronix's own H1'27 capacity lands. The filings flag TLC migration / YMTC ramp / big-4 re-entry filling the cliff over 3-5 years. Margin reverts toward the mid-single-digit through-cycle average → bear case NT$5 (−97%).
It is a commodity memory maker that has destroyed capital across cycles. 10 loss quarters into Q1'26; through-cycle ROE swings to −3% in troughs. The 4× re-rating is the market extrapolating one boom quarter.
No floor. Slight net debt, suspended dividend, 23× peak-cycle EPS. Nothing catches a fall if the eMMC narrative cracks.

Risks to thesis (tail, not bear case)

eMMC shortage mean-revertsHigh

The 18.5% Q1'26 margin is a price spike. When Macronix's own +50% capacity (H1'27) and any TLC migration / big-4 re-entry fill the ~US$1-2bn niche, margins revert toward the mid-single-digit through-cycle average. Bear case NT$5 (−97%).

Peak-cycle priced as permanentHigh

At NT$149.5 the market implies a ~58% perpetual operating margin (2× the all-time gross-margin peak) and 23× peak-cycle EPS. Memory cycles historically de-rate hard off peaks; the 4× run prices a monopoly the cash flows do not support.

Founder successionMed

Miin Wu (founder, Chair+CEO, ~age 77) is credited with every turnaround; succession planning is boilerplate. A leadership transition in a capital-intensive cyclical is a real key-person risk — carried in the 8% governance haircut.

Capex into the peakMed

The NT$22bn (~11× FY25) capacity plan is the classic memory-industry pattern of adding supply near a top — value-destructive if pricing normalizes before the volume lands.

ROM roll-offLow

Nintendo Switch-1 EOL; Switch-2 did not adopt Macronix ROM. ~8% of revenue and falling — a steady structural drag, already in the blended growth.

Upside tail: durable monopolyLow

If the supply cliff truly holds for a decade (TI/ADI-style specialty-analog economics), the bull case is NT$39 — still below price, but the scenario the longs are buying. Sampled at the top of the MC margin axis.

10-year forecast

Revenue NT$33.2B → ~NT$47B over 10y (7% Y1-5, fading to 2.0% TWD inflation). Operating margin normalized at 8% (vs the −13% FY25 / +18.5% Q1'26 spot) and converging to a 11% year-10 target by Y5. Terminal value is 91% of enterprise value — but the whole enterprise value (NT$41.7B) is a seventh of the NT$296B market cap, so the SELL does not hinge on the terminal.

Monte Carlo distribution

The Monte Carlo varies the five load-bearing inputs — including the contested year-10 margin (6-16%) and a β-walk (1.45-1.98) the house method flags as the dominant WACC lever. All 1,000 draws land below the NT$149.5 price (p95 = NT$22.5); P(intrinsic<market) = 100%. Unlike the usual MC caveat, there is no un-sampled bull corner hiding upside here — the discrete bull scenario (NT$39) is already well below price.

Mean NT$14.21 ± NT$4.65/sh, 1000 iterations (0 failed). P(intrinsic < market NT$149.50) = 100.0%.

⚠ Active diagnostic: stable_state.override_roc (0.0650) < WACC (0.0973); every dollar reinvested in stable state destroys value

Cost of capital build

Risk-free rate	2.31%
Mature-market ERP	4.23%
Levered β	1.61
Weighted CRP	0.72%
Cost of equity	10.28%
Pre-tax cost of debt (synth A1/A+)	3.01%
D / V	~7%
WACC	9.73%

Full year-by-year DCF

Year	Revenue	Op mgn	EBIT	EBIT(1−t)	Reinvest	FCFF	PV
1	NT$35.54B	8.60%	NT$3.06B	NT$2.66B	NT$3.32B	NT$-662M	NT$-604M
2	NT$38.02B	9.20%	NT$3.50B	NT$3.04B	NT$3.55B	NT$-510M	NT$-424M
3	NT$40.68B	9.80%	NT$3.99B	NT$3.47B	NT$3.80B	NT$-334M	NT$-252M
4	NT$43.53B	10.40%	NT$4.53B	NT$3.94B	NT$4.07B	NT$-130M	NT$-89M
5	NT$46.58B	11.00%	NT$5.12B	NT$4.46B	NT$4.35B	NT$104M	NT$66M
6	NT$49.37B	11.00%	NT$5.43B	NT$4.65B	NT$3.99B	NT$657M	NT$378M
7	NT$51.84B	11.00%	NT$5.70B	NT$4.80B	NT$3.53B	NT$1.27B	NT$678M
8	NT$53.92B	11.00%	NT$5.93B	NT$4.91B	NT$2.96B	NT$1.95B	NT$961M
9	NT$55.53B	11.00%	NT$6.11B	NT$4.97B	NT$2.31B	NT$2.66B	NT$1.22B
10	NT$56.65B	11.00%	NT$6.23B	NT$4.98B	NT$1.59B	NT$3.40B	NT$1.47B

Methodology & flags

Damodaran FCFF DCF, 10y explicit + perpetuity, in TWD. Risk-free 2.31% (local-currency government bond). Synthetic credit A1/A+. CRP 0.72% from revenue-weighted country mix × Damodaran 2026 CRPs. Terminal ROIC faded to 6.50%; 8% governance haircut applied post-DCF. Monte Carlo: 1000 iterations. Engine v1.0.0 · result: /Users/valentin/Documents/notes_jiliac_labs/Finance/Damodaran/valuations/macronix/output/2026-06-02-result.json

R&D cap: OFF · Lease cap: OFF · Failure: ON · ESO: OFF
Governance haircut: 8% applied post-DCF (NT$15.66 > NT$14.41)
Sensitivity tornado: not run

FCFF DCF · dark theme v3. Generated 2026-06-02T10:47:58 from /Users/valentin/Documents/notes_jiliac_labs/Finance/Damodaran/valuations/macronix/output/2026-06-02-result.json.