← back to picks

HQ Taipei · Taiwan Reporting TWD Credit synth Aaa/AAA · net cash, zero debt Valuation 2026-06-02 Sum-of-the-parts · 1000-iter MC FCFF · Dark v3

Fairly valued once the AI-server stake is marked, not consolidated · −1.4% margin of safety

Spot sits inside the MC band — clears only at the low-rate bookend

This is a sum-of-the-parts valuation of the wholly-owned stub (consumer motherboards + VGA + ASRock Industrial + peripherals), with the 46.2% stake in separately-listed ASRock Rack (7711.TW) carried as a marked-to-market cross-holding (NT$13.35bn), not consolidated. Intrinsic value NT$258.7/share (after a 13% governance haircut; pre-haircut NT$297.4) vs market NT$262.5 — −1.4% at the methodology-correct 2.65% TWD risk-free. The Monte Carlo straddles the price: p5 NT$226 / p50 NT$256 / p95 NT$292, and P(intrinsic<market)=61.4%. It clears spot only at the 1.13% low-rate bookend (+11.5%) — a rate bet, not a margin of safety. WATCH.

FrameSum-of-the-parts · stub + marked Rack stakeSectorComputers / PeripheralsCountry mixUS 30% · DE 22% · CN 16%β / MC σ1.25 levered · ±NT$20/sh (1000 runs)GovernancePegatron 46% control · 13% haircutQualityNet cash · zero debt · synth AAA · ~5% yield

Intrinsic / share

NT$258.73

post 13% gov · pre NT$297.39 · SOTP stub + marked Rack stake

Market / share

NT$262.50

2026-06-02 close · TWSE

Margin of safety

-1.4%

vs intrinsic

Enterprise value

NT$20.30B

56.2% terminal

Cost of equity / debt

8.50% / 2.44%

β 1.25 · CRP 0.45%

Terminal ROIC / g

15.00% / 2.00%

spread ~650bp (ROIC 15.00% vs WACC 8.50%)

What it sells, where it sells

Operating segments

MotherboardsHistorically the highest-margin leg, now the fastest-shrinking — shipments guided −37% in 2026 (4.3M→2.7M units), the steepest cut of the DIY big four~55%

VGA / graphics cardsFY25 inflated by a one-off order pull-forward ahead of ~100% memory-price spikes; 2026 volume ~−30% as orders reverse~25%

ASRock IndustrialEdge-AI / IPC (Siemens, Arrow) — small but genuinely growing +24.5%; the higher-quality leg~12%

Peripherals / mini-PCsAIO coolers, PSUs, monitors, gaming gear — growing off a small base~8%

The stub's FY2025 revenue of NT$23.0bn (+36% off FY2024) is a low-quality peak: the largest leg (motherboards) is in structural decline and the second-largest (VGA) was inflated by a one-off pull-forward — the small, higher-quality industrial/peripherals growers cannot offset both, so the honest base case models the near term DOWN ~10-20% then stabilising. The AI-server engine (ASRock Rack) is NOT in this mix — it is the separately-marked stake.

Country mix (revenue-weighted CRP input)

🇺🇸United States30%

🇩🇪Germany22%

🇨🇳China16%

🇹🇼Taiwan12%

🇬🇧United Kingdom8%

🇯🇵Japan6%

🇰🇷South Korea4%

🇸🇬Singapore2%

Geographic disclosure is sparse (low-confidence revenue-weighting): the stub sells globally into developed DIY/PC and industrial channels, so demand is a Western-consumer build-cycle story, not a domestic-Taiwan one — but the equity bridge and the marked Rack stake are pure TWD, which is why the risk-free regime moves the answer more than the geo mix does.

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.

This is a sum-of-the-parts, not a blended company DCF. ASRock Inc is becoming a partial holding company: its AI-server subsidiary ASRock Rack (7711.TW) IPO'd Nov 2025 and now contributes ~52% of consolidated revenue. The prior blended DCF averaged a declining high-margin board core with a fast-growing low-margin server engine and produced a mix-shift artifact (~5.74% blended OM, ~3.5% growth) that described neither business. This rebuild values the stub on its own and carries the 46.2% Rack stake as a marked cross-holding.
The stub is a cyclical, partly-declining hardware mix — not the AI growth engine. Motherboards (the historically highest-margin leg) are guided −37% on units in 2026, the steepest of the DIY big four, with no platform catalyst and memory-price inflation choking DIY builds; DigiTimes calls the slump "worse than the GFC or early COVID."
FY2025 stub revenue is an inflated peak. NT$23.0bn (+36% off FY2024) was lifted by a one-off VGA order pull-forward booked ahead of ~100% memory-price spikes — a borrow-from-the-future event that reverses in 2026 (VGA volume ~−30%).
The value lives on the balance sheet, not the income statement. Asset-light, zero interest-bearing debt, net cash, ~100%-prepayment terms (sales-to-capital ~4.0) — which lets even a flat, ~7%-margin business clear its cost of capital. 24 consecutive years of dividends at the parent; ~5% yield, ~45% payout.
Controlled by Pegatron (46.3%). Live related-party channels with the ASUS/Pegatron group, weaker ROC minority protection, and a NT$2.4bn non-core Beitou property purchase (May 2026) routing cash into real estate rather than buybacks/minorities — the live capital-allocation signal behind the 13% governance haircut.

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 growthStub Y1-5 blended CAGR off an inflated 2025 base

−1% Y1-5Near-term DOWN ~10-20% (MB −37% units + VGA pull-forward reversal), only partly offset by industrial(+24.5%)/peripherals; decays UP to 2.0% terminal

~5% CAGRComputers/Peripherals cross-section

−NT$24

02Operating marginYear-10 stub OM vs sector EBIT margin

7.0%A touch below the FY25 stub 7.78% for adverse mix shift; far below the motherboard-era 12-15% that has "gone structurally"

22.48%US Computers/Peripherals (branded/IP-rich — wrong comp for an assembler)

−NT$98

03Sales-to-capitalReinvestment efficiency vs sector median

4.0×Asset-light, ~100%-prepayment, zero-debt model — the real engine of value

3.62×US Computers/Peripherals cross-section

+NT$10

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

2.0%TWD long-run inflation; held ≤ the 2.65% risk-free (structural MB decline argues the low end)

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

NT$0

05Cost of capital10y WACC vs WACC implied by AI-skewed cohort β

8.50%β 1.25 (3-source company regression) · rf 2.65% (65/35 blend) · ERP+CRP · zero debt

~9.5%Global Computers/Peripherals β_u 1.468 (AI-server-heavy) → higher WACC

+NT$12

Net effect of overrides

Overrides net negatively vs the branded-cohort defaults — deliberately. The work is in refusing the 22.48% sector margin (this is a contract-style assembler, not an IP-rich brand) and refusing to extrapolate the VGA-inflated 2025 revenue. The asset-light sales-to-capital and the lower-than-AI-cohort β add value back, but the honest stub is a flat, thin-margin business whose worth comes from the balance sheet — plus the marked Rack stake in the bridge.

−NT$100

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

ASRock is no longer one company, so we did not value it as one. We valued the wholly-owned stub — the old ASRock of consumer motherboards, graphics cards, the small industrial unit, and peripherals — and carried its 46.2% stake in the separately-listed AI-server business, ASRock Rack, as a marked-to-market holding rather than blending the two together. The stub is not a growth story: its biggest and historically richest leg, motherboards, is shrinking fast (units guided down 37% in 2026), and its graphics-card leg had an artificially strong 2025 from orders pulled forward ahead of a memory-price spike, which now reverses. So the honest base case has stub revenue falling in the near term and then roughly flat, with the operating margin settling near 7% rather than the dead motherboard-era teens. What gives the stub any value is its balance sheet — asset-light, zero debt, paid by customers up front — which lets even a flat, thin-margin business earn above its cost of capital. Add the marked Rack stake and the net cash, subtract a small non-Rack minority and a 13% governance haircut for Pegatron's control, and you land at NT$258.7 against a market price of NT$262.5. That is fairly valued: the intrinsic clears the price only if you accept the 1.13% low-rate bookend, so this is a WATCH, not a buy.

Two debates worth pressure-testing

Should ASRock be valued as one consolidated company, or split?

Our view: Split. Consolidating a fast-growing, low-margin AI-server engine (ASRock Rack, ~52% of consolidated revenue) with a declining high-margin board core produces a mix-shift artifact that describes neither business — a blended ~5.74% OM and ~3.5% growth that is an average of two opposite stories. We value the stub on its own economics (cyclical, partly declining, ~7% OM) and carry the 46.2% Rack stake at its NT$13.35bn market value in the equity bridge. The decomposition: stub operations ~NT$164/sh (55%), the Rack stake ~NT$108/sh (36%), net cash ~NT$29/sh (10%), less a small non-Rack minority −NT$4/sh → NT$297/sh pre-governance.

Is the −1.4% margin of safety a real "fair value", or a low-rate illusion?

Our view: It is genuinely fair, and rate-fragile. At the methodology-correct 2.65% TWD risk-free the intrinsic is NT$258.7 — 1.4% below spot. It clears the price only at the suppressed 1.13% local-bond floor (NT$292.7, +11.5%); at a normalized 4.3% rate it falls to NT$231.0 (−12.0%). The Monte Carlo agrees the call is a coin-flip-against: p50 NT$256 sits just under the NT$263 price and P(intrinsic<market) is 61.4%. No margin of safety to underwrite a buy → WATCH.

CLAIM 01Stub revenue falls near-term then stabilises — roughly flat over the decade.growth_high: −1% · terminal: 2.0% · NT$23.0B → ~NT$23-24BFY25 +36% was a VGA pull-forward, not a trend. Guided MB units −37% and VGA volume −30% in 2026 swamp the small industrial(+24.5%)/peripherals growers → FY26 down ~10-20%, then flat.

CLAIM 02Year-10 operating margin ~7% — below the FY25 stub level, far below the motherboard era.target_op_margin: 7% from Y5FY25 stub OM 7.78% (above the blended 5.74% once low-margin Rack is removed), but the highest-margin leg shrinks while VGA/industrial are lower-margin → mid of a 6.5-8% band. Not the 22.48% branded-cohort comp.

CLAIM 03The balance sheet, not the P&L, creates the value — sales-to-capital ~4.0.S2C: 4.0× Y1-5 and Y6-10Asset turnover ~2x, receivables turnover 16.5x (near-prepayment), zero debt. Above the 3.62 US cross-section; not pushed higher given the Beitou property drag and the mix away from highest-turn MB.

CLAIM 04Margin reaches the ~7% plateau by Year 5; growth decays to risk-free by Year 10.year_of_convergence: 5Mature cyclical, no S-curve. The MB-decline and VGA-reversal digestion window resolves within five years; thereafter growth fades linearly up to the 2.0% terminal.

CLAIM 05No failure risk; terminal ROC faded to 0.15; 13% governance haircut.terminal_g: 2.0% · failure: 0% · gov haircut: 13%Net cash, zero debt, 24-yr dividend record → failure 0. Engine's auto-ROC (~0.224) overstates a perpetual moat for a declining assembler → faded to 0.15. Pegatron control + property drag + minority routing → 13% residual.

Where we diverge from sell-side

We split the company; the sell-side blends it. The three covering analysts carry an ~NT$301 consolidated target. Our SOTP says the value is real but already in the price: the stub is cyclical and declining, and the AI growth lives in a stake we mark, not consolidate — so we do not capitalise Rack's growth twice.
We refuse to extrapolate the 2025 revenue peak. The +36% stub jump was a VGA pull-forward; the base case models FY26 DOWN ~10-20%, not continued growth.
We refuse the branded-cohort margin. The 22.48% US Computers/Peripherals EBIT margin is the wrong comp for a contract-style assembler with no pricing-power moat; we hold ~7% terminal OM.
β set to 1.25 from the 3-source company regression, not the 1.468 cohort. The global cohort is AI-server-heavy (high β) while this stub explicitly excludes the AI-server business — the low-β residual is the right read.
Governance is in the residual at 13%, not bolted on as a flat "Taiwan discount." The flows already carry ROC fade, the marked stake net of look-through, and the small non-Rack minority; the haircut is leakage/dilution under Pegatron control plus the May-2026 property signal.

Two-sided case — bear anchors

Rate normalization breaks the thin cushion. At a normalized 4.3% TWD risk-free the intrinsic falls to NT$231.0 — −12.0% below spot. ~67% of stub value is the terminal block, so the answer is rate-sensitive and there is no margin of safety to absorb it.
The motherboard decline could outrun the small growers. If MB units keep falling past the guided −37% and industrial/peripherals cannot offset, stub OM fades toward 6% and the terminal excess return disappears — the stress regime.
The marked Rack stake is itself a market price. ~36% of pre-gov value is the 46.2% Rack holding at NT$13.35bn. If 7711.TW de-rates (AI-server margin pressure, customer-credit risk), the cross-holding marks down 1-for-1.
Governance/capital-allocation drift. The NT$2.4bn Beitou property purchase routes cash into real estate, not buyback-and-cancel or minority buy-in; mild dilution (+1.28% share count) compounds the control discount.
The whole thing is fairly valued. Unlike a trough/peak debate, here the Monte Carlo p50 (NT$256) sits just under spot and P(intrinsic<market)=61.4% — the base case simply does not offer an edge at today's price.

Risks to thesis (tail, not bear case)

Rate normalizationHigh

~67% of stub value is the terminal block. The risk-free is at the methodology 2.65% blend; a move to a normalized 4.3% takes the intrinsic to NT$231 (−12.0%). With no base-case margin of safety, rates move the verdict.

Motherboard decline acceleratesMed

Units already guided −37% in 2026, the steepest of the DIY big four. If the secular slide outruns the small industrial/peripherals growers, stub OM fades toward 6% and the terminal excess return erodes.

VGA pull-forward reversal deeper than modelledMed

FY25 was inflated by orders pulled forward ahead of ~100% memory-price spikes. If the 2026 reversal is sharper than the −30% volume assumed, near-term revenue drops more than ~10-20%.

Rack stake mark-downMed

~36% of pre-gov value is the marked 46.2% ASRock Rack stake (NT$13.35bn). AI-server margin pressure or customer-credit shocks at 7711.TW mark the cross-holding down 1-for-1.

Pegatron control / capital allocationMed

46.3% control, live related-party channels, weaker ROC minority protection, and a NT$2.4bn non-core property purchase routing cash away from minorities — captured in the 13% governance haircut but a live drift risk.

TWD appreciationLow

FX drag drove a ~97% collapse in non-operating income in FY2025; a persistently strong TWD compresses reported NT$ revenue and margin even at flat USD volumes.

10-year forecast

Stub-only forecast (ASRock Rack excluded — it enters as the marked cross-holding). Revenue is held roughly flat as the motherboard decline and VGA pull-forward reversal are absorbed; the operating margin settles to its ~7% structural plateau by Year 5.

Monte Carlo distribution

The 1,000-draw distribution straddles the price: p5 NT$226, p50 NT$256, p95 NT$292, mean NT$258 ± NT$20. With spot at NT$263, P(intrinsic<market)=61.4% — the base case offers no edge at today's price.

Mean NT$257.82 ± NT$20.27/sh, 1000 iterations (0 failed). P(intrinsic < market NT$262.50) = 61.4%.

SUM-OF-THE-PARTS & THREE RISK-FREE REGIMES

This is not a single-company blend. The valuation decomposes the NT$297.4 pre-governance intrinsic into the wholly-owned stub, the marked AI-server stake, net cash, and the non-Rack minority — then a 13% governance haircut yields NT$258.7. The decision number is the 2.65% base regime; the bookends show the answer clears spot only at the suppressed low-rate floor.

Stub operations (MB + VGA + Industrial + peripherals)+NT$164 · 55%

ASRock Rack stake (46.2% of 7711.TW, marked to market)+NT$108 · 36%

Net cash (zero debt)+NT$29 · 10%

Non-Rack subsidiary minority−NT$4 · −1%

= NT$297.4 pre-governance → −13% governance haircut → NT$258.7 intrinsic (vs NT$262.5 spot).

Risk-free regime	TWD rf	Intrinsic (post-gov)	vs spot NT$262.5
Low (suppressed local bond)	1.13%	NT$292.7	+11.5%
Base (65/35 blend) — decision	2.65%	NT$258.7	−1.4%
High (normalized)	4.3%	NT$231.0	−12.0%

Clears spot only at the low-rate bookend — a rate bet, not a margin of safety. Combined with a Monte Carlo p50 of NT$256 sitting just below price and P(intrinsic<market)=61.4%, the verdict is WATCH.

Cost of capital build

Risk-free rate	2.65%
Mature-market ERP	4.32%
Levered β	1.25
Weighted CRP	0.45%
Cost of equity	8.50%
Pre-tax cost of debt (synth Aaa/AAA)	3.05%
D / V	~0%
WACC	8.50%

Full year-by-year DCF

Year	Revenue	Op mgn	EBIT	EBIT(1−t)	Reinvest	FCFF	PV
1	NT$22.73B	7.62%	NT$1.73B	NT$1.45B	NT$-57M	NT$1.51B	NT$1.39B
2	NT$22.51B	7.47%	NT$1.68B	NT$1.40B	NT$-57M	NT$1.46B	NT$1.24B
3	NT$22.28B	7.31%	NT$1.63B	NT$1.36B	NT$-56M	NT$1.42B	NT$1.11B
4	NT$22.06B	7.16%	NT$1.58B	NT$1.32B	NT$-56M	NT$1.37B	NT$992M
5	NT$21.84B	7.00%	NT$1.53B	NT$1.28B	NT$-55M	NT$1.33B	NT$886M
6	NT$21.75B	7.00%	NT$1.52B	NT$1.26B	NT$-22M	NT$1.28B	NT$790M
7	NT$21.79B	7.00%	NT$1.53B	NT$1.25B	NT$11M	NT$1.24B	NT$709M
8	NT$21.97B	7.00%	NT$1.54B	NT$1.25B	NT$44M	NT$1.21B	NT$643M
9	NT$22.28B	7.00%	NT$1.56B	NT$1.26B	NT$77M	NT$1.18B	NT$587M
10	NT$22.72B	7.00%	NT$1.59B	NT$1.27B	NT$111M	NT$1.16B	NT$541M

Methodology & flags

Damodaran FCFF DCF, 10y explicit + perpetuity, in TWD. Risk-free 2.65% (local-currency government bond). Synthetic credit Aaa/AAA. CRP 0.45% from revenue-weighted country mix × Damodaran 2026 CRPs. Terminal ROIC faded to 15.00%; 13% governance haircut applied post-DCF. Monte Carlo: 1000 iterations. Engine v1.0.0 · result: /Users/valentin/Documents/notes_jiliac_labs/Finance/Damodaran/valuations/asrock/output/2026-06-02-result.json

R&D cap: OFF · Lease cap: OFF · Failure: OFF · ESO: OFF
Governance haircut: 13% applied post-DCF (NT$297.39 > NT$258.73)
Sensitivity tornado: not run

FCFF DCF · dark theme v3. Generated 2026-06-02T07:16:35 from /Users/valentin/Documents/notes_jiliac_labs/Finance/Damodaran/valuations/asrock/output/2026-06-02-result.json.