← back to picks

HQ Hsinchu / Changshu · Cayman-inc Reporting TWD Credit synth Aaa/AAA · net cash ~NT$11bn Valuation 2026-06-04 Damodaran FCFF · 1000-iter MC · β-walk FCFF · Dark v3

Cheap at base, only fair at the high-rate bookend · +47% margin of safety

WATCH — a TWD-rate + sustained-AWS-execution bet, not a clean BUY

This is a three-regime valuation, not a point estimate. Intrinsic value NT$1,769/share (post 12% governance haircut; pre-haircut NT$2,011) vs market NT$1,200 — +47% at the methodology-correct 2.31% TWD risk-free (BASE). But the cushion is mostly a rate-and-execution bet: at the normalized 4.30% high-rate bookend intrinsic collapses to NT$1,204 (+0.3%, i.e. fair), and the compounded bear (high rate × β1.58 × 19% margin × 11% growth) prints NT$679 (−43%). The Monte Carlo p5 is NT$1,276 (still +6%) with P(intrinsic<market)=1.6%, but the MC spread does not span the rate-regime risk — the bookend table does.

SectorElectronics (General) · AI-DC BBU + LEVCountry mixUS 70% · DE 17% · CN 7%β / MC σ1.35 levered (walk 1.19/1.35/1.58) · ±NT$335/shGovernanceSimplo-controlled · 12% haircutQuality~40% ROIC · net cash ~NT$11bn · synth AAAIncome~NT$19 dividend · 1.59% gross / 1.26% net yield

Intrinsic / share

NT$1,769.40

post 12% gov · pre NT$2,010.68 · BASE regime

Market / share

NT$1,200.00

2026-06-04 · TWSE · 52wk NT$903–1,555 · highly volatile

Margin of safety

+47.4%

vs intrinsic

Enterprise value

NT$160.78B

79.6% terminal

Cost of equity / debt

8.39% / 2.17%

β 1.35 · CRP 0.27%

Terminal ROIC / g

15.00% / 2.00%

spread ~666bp (ROIC 15.00% vs WACC 8.34%)

What it sells, where it sells

Operating segments

AI-data-center BBUBattery-backup units sold direct to US hyperscalers; 3 of big-4 CSPs adopting, HVDC 400/800V ramp H2'26-2027 — the engine~70%

Light-EV / e-bike packsEuropean e-bike and light-electric-vehicle battery modules; recovering but cyclical (halved once in 2023) — the floor~25%

Industrial / ESS / otherEnergy-storage, telecom (in volume), early AGV/AMR/forklift packs~5%

The whole re-rating rests on the ~70% BBU line, an 18-month-old product sold to a handful of CSPs — so revenue mix is the thesis: the cyclical LEV floor (~25%) has already halved once, and the bet is that BBU standardization is structural rather than one customer's capex cycle.

Country mix (revenue-weighted CRP input)

🇺🇸United States70%

🇩🇪Germany (W. Europe proxy)17%

🇨🇳China7%

🇹🇼Taiwan6%

Revenue is weighted to END demand, not billing site: BBU is invoiced to PSU/integrator partners in TW/CN but the demand is US hyperscaler data-center capex (~70%), so this is a US-AI-capex derivative wearing an Asian billing address — the reported "Asia 58%" line is a billed-to artifact.

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.

An AI-data-center BBU engine stapled to a cyclical e-bike packer. ~70% of TTM revenue (NT$17.1bn, FY2025 +59% YoY) is battery-backup units sold direct to US hyperscalers; the rest is European light-EV/e-bike packs (the recovering-but-mature floor) plus industrial/ESS. The entire re-rating rests on the recent 2024-25 BBU surge — durability is the open question.
Genuine AI-infrastructure beneficiary, but single-customer-class. AI rack power is climbing 6kW → 20-33kW+ → 100kW+; distributed BBU is shifting from optional to standard kit on GB200/GB300-class racks (Infineon endorsement). 3 of the top-4 CSPs have adopted, the 4th expected next-gen; CSP capex was >$300bn in 2025 (+41%) heading to ~$340bn. But end demand is a handful of Tier-1 hyperscalers — AWS-dominant — not a broad TAM.
The margin story is internally contradictory — and that is the crack. Gross margin peaked at 39.4% (FY2024) then fell to 35.9% in FY2025 despite +59% revenue — the opposite of operating leverage. A US foreign broker is Neutral (TP NT$1,330) explicitly on margin-premium normalization vs peers (順達/新盛力) as 5kW BBU commoditizes. We model 22% terminal operating margin (below the current ~23.5%), not expansion.
Exceptional balance sheet, but Simplo-controlled. ~40% ROIC, net cash ~NT$11bn (debt only NT$822M), Altman Z 9.1, asset-light, ~50% dividend payout. Chairman Sung Fu-Hsiang simultaneously chairs Simplo (6121); a May-2026 AGM waived non-compete so directors may run competing battery businesses, atop documented related-party leases and shared cell procurement — transfer-pricing risk that can route value to the parent.
Cash reaches minorities, but a foreign minority bleeds withholding. The ~NT$19 / ~50%-payout dividend is real and growing (controlled-company Condition A met → not a trap), but at a 1.59% gross yield it nets only ~1.26% after 21% Taiwan dividend WHT — so the FCFF intrinsic is a CEILING, not the cash a foreign minority actually realizes.

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 growthY1-5 CAGR vs Electronics (General) median

16% Y1-5Below the +21% consensus; decays to 2.0% terminal → ~NT$50bn (≈3x TTM) by 2036. Respects LEV cyclicality + AWS concentration.

~5% CAGRElectronics (General), global cross-sector median

+large

02Operating marginYear-10 target vs sector median EBIT margin

22.0%BELOW the current ~23.5%; models the commoditization the broker flags, not expansion. Council moved 0.23→0.22.

10.4%Electronics (General) median EBIT margin

+large

03Sales-to-capitalReinvestment efficiency vs sector median

2.5× / 3.0×Asset-light assembler, but council cut growth-phase 3.0→2.5 for working-capital drag on a 3x hardware ramp; mature phase eases to 3.0.

2.38×Electronics (General) industry median

+small

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

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

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

NT$0

05Cost of capital10y WACC — the load-bearing input in a low-rate-currency DCF

8.34%β 1.35 (regression 1.19 shaded ~40% toward sector 1.58) · rf 2.31% (50/50 blend) · ERP+CRP · net-cash D/V

~9.5%sector β 1.58 re-levered → higher WACC (the HIGH/BEAR regime)

regime-defining

Net effect of overrides

The value is built by the growth-and-margin engine (16% Y1-5, 22% terminal margin vs a 10.4% sector median) — but it is GATED by the rate regime. At the BASE 8.34% WACC the intrinsic is NT$1,769 (+47%); at the HIGH 10.32% WACC bookend it falls to NT$1,204 (fair); the compounded BEAR lands NT$679 (−43%). The cushion is substantially a TWD-rate + sustained-AWS-execution bet, which is why this is a WATCH, not a BUY.

+47% BASE

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

AES-KY is two businesses stapled together — a cyclical European e-bike battery packer that is the recovering-but-mature floor, and an AI-data-center battery-backup-unit (BBU) engine sold direct to US hyperscalers that is roughly 70% of revenue and the entire re-rating. The base case rewards the structural BBU volume story (rising CSP capex, BBU becoming standard kit on AI racks, an HVDC mix-up) while disciplining the margin story: we deliberately do not extrapolate the FY2024 39.4% gross-margin peak or the FY2025 +59% revenue boom, because gross margin actually fell to 35.9% while revenue grew 59% — scale is not defending price. Revenue compounds about 16% a year through year five (below the +21% consensus, respecting the cyclical floor and single-customer-class concentration), fading to a TWD-mature rate and roughly tripling the NT$17.1bn base by 2036; operating margin normalizes to 22%, below today's ~23.5%. The decisive feature is that this is a three-regime valuation, not a point estimate: it is cheap at the BASE 2.31% TWD risk-free (+47%), but only fair at the normalized 4.30% high-rate bookend (NT$1,204, +0.3%), and the compounded bear — high rate, sector β 1.58, a 19% margin and 11% growth — prints NT$679, down 43%. So the +47% cushion is substantially a bet that TWD rates stay low AND that one customer-class (AWS-dominant CSPs) keeps the BBU franchise durable and high-margin. That is a position to size for risk, not a clean buy — hence WATCH.

Two debates worth pressure-testing

Is the +47% margin of safety real, or just the low TWD rate flattering the terminal block?

Our view: This is the decisive test, and AES-KY only half-passes it. Terminal value is ~80% of EV, so the rate is load-bearing. At the BASE 2.31% risk-free (WACC 8.34%) intrinsic is NT$1,769 (+47%); at the methodology-correct HIGH 4.30% bookend (WACC 10.32%) it falls to NT$1,204 — essentially today's price. Unlike a robust buy, the cushion does NOT survive a normalized rate. It is cheap if rates stay low and execution holds; it is fair if rates normalize. That asymmetry is why we headline the regime band, not a point target.

Can you model a moat that triples revenue AND holds a premium 22% margin while the product commoditizes?

Our view: The council flagged this as internally contradictory — gross margin fell 39.4%→35.9% while revenue grew +59%. We answer it by splitting the thesis: we believe the VOLUME story (BBU standardization, strong broad-based quarterly evidence) and discipline the MARGIN story (22% terminal, below current, vs a bull 25%). The compounded bear collapses both at once — 19% margin × 11% growth × β1.58 × high rate → NT$679 (−43%). Because the franchise is an 18-month-old single-customer-class product, no track record adjudicates which regime wins; we size for the spread, not the midpoint.

CLAIM 01Revenue compounds ~16% Y1-5 on structural BBU volume, fading to 2.0%.growth_high: 16% · terminal: 2.0% · NT$17.1B → ~NT$50BBBU standardizing on GB200/GB300 racks; 3 of big-4 CSPs adopting, CSP capex >$300bn (+41%). Set below the +21% consensus to respect the LEV cyclical floor and AWS concentration.

CLAIM 02Operating margin NORMALIZES to ~22% by Y10 — not expansion.target_op_margin: 22% (council 0.23→0.22)GM peaked 39.4% (FY24) then fell to 35.9% at +59% revenue. Broker Neutral on margin normalization vs 5kW peers. Still well above the 10.4% sector median on direct-to-CSP + HVDC.

CLAIM 03Reinvestment carries working-capital drag through the 3x hardware ramp.S2C: 2.5× Y1-5 · 3.0× Y6-10Asset-light assembler, but the 2026 capex wave (+20-30%) and a 3x hardware ramp consume inventory/receivables. Council cut the growth-phase 3.0→2.5; mature phase eases back to 3.0.

CLAIM 04Short moat tail — growth and margin converge by Year 10.year_of_convergence: 10 · no extensionCommoditizing low-voltage competition plus single-customer-class concentration argue against a long moat tail. Growth decays linearly Y6-10 to the terminal rate; no extension granted.

CLAIM 05Terminal ROC faded, terminal g ≤ risk-free, zero failure risk.terminal_g: 2.0% · ROC 15% · failure: 0% · gov: 12%Engine-resolved terminal ROC (~53%) is boom-inflated; faded to 15% (still a premium to ~8% WACC). Net cash ~NT$11bn → 0% failure. 12% governance haircut for Simplo-controlled minority + the non-compete waiver.

Where we diverge from sell-side

We headline a regime BAND, not the sell-side's point target. Consensus TP ~NT$1,520 (+~23%); the Neutral broker is at NT$1,330. We agree the BASE intrinsic (NT$1,769) is above price, but we explicitly show it collapses to NT$1,204 at a normalized rate — the call is the spread, not a number.
We believe the volume story and discipline the margin story. Sell-side guides ~+21% revenue and management ~2x in 1-2 years; we set 16% Y1-5 and a 22% terminal margin (below current), modeling the commoditization the GM-vs-revenue contradiction already reveals.
β set to 1.35 by sector-shading, not the 1.19 regression. The backward-looking regression captured a company that was ~half e-bikes over the window; the forward business is a high-beta single-customer AI-capex derivative, so we shade ~40% toward the sector-relevered 1.58 and stress-test 1.58 in the bear.
Governance modeled in the flows + a 12% residual, not a flat discount. 2.4x the parent Simplo's own 5%, reflecting minority-of-a-controlled-sub plus the fresh non-compete waiver; we also cross-check the dividend net of 21% WHT (1.26%) — the FCFF intrinsic is a ceiling for a foreign minority.
We refuse to extrapolate the boom. No 39.4% GM into terminal, no +59% growth into perpetuity, terminal ROC faded from ~53% to 15% — the opposite of the bull's standard-kit-infra-compounder re-rate.

Two-sided case — bear anchors

It is only fair at a normalized rate. Terminal value is ~80% of EV. At the HIGH 4.30% risk-free / 10.32% WACC bookend, intrinsic is NT$1,204 — +0.3% on today's NT$1,200. The +47% BASE cushion is substantially a bet that TWD rates stay low.
The compounded bear is −43%. Stack high rate × sector β 1.58 × a 19% operating margin × 11% growth × an 18% governance haircut (WACC 11.35%) and intrinsic is NT$679. The downside dominates if the franchise commoditizes.
One customer-class is the existential variable. ~70% of revenue and the whole terminal block rest on a handful of CSPs, AWS-dominant, on an 18-month-old product. One or two renewal cycles — not a TAM. CSP in-sourcing breaks the model.
The margin is already eroding. GM fell 39.4%→35.9% while revenue grew +59% — scale is not defending price. If 5kW-tier ASP erosion outruns HVDC, the through-cycle band drifts to low-30s GM / ~21% op margin (the bear margin).
A foreign minority realizes less than the FCFF print. The ~1.59% gross dividend yield nets ~1.26% after 21% Taiwan WHT, and related-party transfer pricing to parent Simplo (newly de-fenced by the non-compete waiver) can route value up. The intrinsic is a ceiling.

Risks to thesis (tail, not bear case)

Rate normalizationHigh

~80% of value is the terminal block. Moving the TWD risk-free from the 2.31% BASE blend to the 4.30% high bookend (WACC 8.34%→10.32%) cuts intrinsic from NT$1,769 to NT$1,204 — the cushion is mostly a rate bet, the dominant risk.

Customer concentration (AWS / CSP)High

~70% of revenue is BBU sold to a handful of hyperscalers, AWS-dominant. Losing one renewal, or CSP in-sourcing, breaks the model — an 18-month-old product with no track record to adjudicate durability.

Margin commoditizationMed

GM already fell 39.4%→35.9% at +59% revenue; 5kW-tier competition (順達/新盛力) is entering. If ASP erosion outruns the HVDC moat-extension, the terminal margin drifts toward the 21% bear, ~NT$300+/share of value.

β / forward risk re-rateMed

The 1.19 regression is backward-looking on a half-e-bike past; the forward business is a high-beta single-customer AI-capex derivative. The β-walk runs 1.19/1.35/1.58 — pinning the sector 1.58 alone takes intrinsic well below base.

Simplo-controlled governanceMed

Chairman runs the parent Simplo (6121); the May-2026 AGM waived non-compete, atop related-party leases and shared cell procurement. Transfer-pricing can route value to the parent — handled as a 12% haircut plus the WHT cash-to-minority cross-check.

LEV cyclicality + TWD/tariffMed

The ~25% light-EV floor already halved once (−33% in 2023); a relapse drags the blend. TWD appreciation compressed FY2025 margin, and US tariff/FX on the data-center supply chain is a live swing factor.

10-year forecast

Revenue compounds ~16% in Y1-5 (below the +21% consensus) and fades to a ~2% TWD-mature rate by Y10, roughly tripling the NT$17.1bn TTM base to ~NT$50bn by 2036; operating margin normalizes to 22% — below today's ~23.5% — rather than expanding, modeling the commoditization the GM-vs-revenue contradiction already reveals.

Monte Carlo distribution

Across 1,000 correlated draws — including a β-walk (1.19/1.35/1.58) and a margin axis — the 5th-percentile outcome is NT$1,276, still ~6% above today's NT$1,200, and only 1.6% of draws land below market. But the MC perturbs the operating inputs around the BASE 2.31% rate; it does NOT span the rate-regime risk, which is the real swing factor (see the regime table above).

Mean NT$1782.52 ± NT$335.09/sh, 1000 iterations (0 failed). P(intrinsic < market NT$1200.00) = 1.6%.

The decision is the regime table, not a point target

Because terminal value is ~80% of enterprise value, the TWD risk-free regime — not any single operating input — moves the verdict. This is the table the council demanded: all three rate regimes plus the compounded bear, post-governance, against today's NT$1,200.

Regime	Risk-free	WACC	β	Intrinsic / sh (post-gov)	Upside vs NT$1,200
LOW — local-bond floor	1.13%	7.16%	1.35	NT$2,141	+78%
BASE — 50/50 blend (headline)	2.31%	8.34%	1.35	NT$1,769	+47%
HIGH — normalized rate	4.30%	10.32%	1.35	NT$1,204	+0.3%
BEAR — compounded (β1.58 · 19% margin · 11% growth · 18% gov)	4.30%	11.35%	1.58	NT$679	−43%

Read it as a spread: cheap if TWD rates stay low AND the BBU franchise holds, fair the moment rates normalize, deeply negative if the product also commoditizes. The Monte Carlo p5 (NT$1,276) is reassuring on the operating axes (P(intrinsic<market)=1.6%) but does NOT span the rate-regime risk — the bookends do. Cash-to-minority cross-check: the ~1.59% gross dividend yield nets ~1.26% after 21% Taiwan WHT, so the FCFF intrinsic above is a CEILING for a foreign minority. Verdict: WATCH — a margin-of-safety call to be sized for risk, not a clean BUY.

Cost of capital build

Risk-free rate	2.31%
Mature-market ERP	4.30%
Levered β	1.35
Weighted CRP	0.27%
Cost of equity	8.39%
Pre-tax cost of debt (synth Aaa/AAA)	2.71%
D / V	~1%
WACC	8.34%

Full year-by-year DCF

Year	Revenue	Op mgn	EBIT	EBIT(1−t)	Reinvest	FCFF	PV
1	NT$19.83B	23.36%	NT$4.63B	NT$3.65B	NT$1.09B	NT$2.56B	NT$2.36B
2	NT$23.00B	23.21%	NT$5.34B	NT$4.21B	NT$1.27B	NT$2.94B	NT$2.50B
3	NT$26.68B	23.06%	NT$6.15B	NT$4.85B	NT$1.47B	NT$3.38B	NT$2.65B
4	NT$30.95B	22.90%	NT$7.09B	NT$5.59B	NT$1.71B	NT$3.88B	NT$2.82B
5	NT$35.90B	22.75%	NT$8.17B	NT$6.44B	NT$1.98B	NT$4.46B	NT$2.99B
6	NT$41.64B	22.60%	NT$9.41B	NT$7.44B	NT$1.91B	NT$5.53B	NT$3.42B
7	NT$48.31B	22.45%	NT$10.85B	NT$8.60B	NT$2.22B	NT$6.38B	NT$3.64B
8	NT$56.03B	22.30%	NT$12.50B	NT$9.94B	NT$2.58B	NT$7.36B	NT$3.88B
9	NT$65.00B	22.15%	NT$14.40B	NT$11.48B	NT$2.99B	NT$8.50B	NT$4.13B
10	NT$75.40B	22.00%	NT$16.59B	NT$13.27B	NT$3.47B	NT$9.80B	NT$4.40B

Methodology & flags

Damodaran FCFF DCF, 10y explicit + perpetuity, in TWD. Risk-free 2.31% (local-currency government bond). Synthetic credit Aaa/AAA. CRP 0.27% from revenue-weighted country mix × Damodaran 2026 CRPs. Terminal ROIC faded to 15.00%; 12% governance haircut applied post-DCF. Monte Carlo: 1000 iterations. Engine v1.0.0 · result: /Users/valentin/Documents/notes_jiliac_labs/Finance/Damodaran/valuations/aes-ky/output/2026-06-04-result.json

R&D cap: OFF · Lease cap: OFF · Failure: OFF · ESO: OFF
Governance haircut: 12% applied post-DCF (NT$2010.68 > NT$1769.40)
Sensitivity tornado: not run

FCFF DCF · dark theme v3. Generated 2026-06-04T17:10:44 from /Users/valentin/Documents/notes_jiliac_labs/Finance/Damodaran/valuations/aes-ky/output/2026-06-04-result.json.