Business

Know the Business

Figures converted from Hong Kong dollars (HKD) at historical FX rates — see data/company.json.fx_rates. Ratios, margins, multiples, share counts and dates are unitless and unchanged.

ASMPT is two businesses stapled together: a high-IP back-end packaging franchise (SEMI) riding the largest demand pull of its history through Thermo-Compression Bonding (TCB) for AI memory and logic, and a mature surface-mount placement business (SMT) whose margins look more like an industrial automation peer than a semicap one. The consolidated P&L hides this dispersion — a 38% group gross margin and 7% segment-profit margin are the weighted average of a structurally great SEMI engine and a slow-grower formally put up for review. The market is most likely under-pricing the SEMI franchise (still expanding TCB installed base past 500 tools) and over-extrapolating SMT weakness; the gap closes if management actually separates the two.

1. How This Business Actually Works

Revenue is a capital-equipment sale to chipmakers and electronics assemblers, with a long-life service annuity attached. A foundry, OSAT, memory IDM or Tier-1 EMS places a multi-tool order; ASMPT builds and ships the bonder, placement head or molding system over 4–9 months; revenue is recognised on shipment. After installation the tool throws off 7–12 years of spares, upgrades and service. The economics are set on day one of the order: which SKU, which customer, which package.

Three properties of these economics matter more than anything else on the P&L.

Margin is set by mix, not efficiency. SEMI gross margin runs in the mid-40s when TCB ships heavily; it drops into the high-30s when wire bond and mainstream die bond dominate. The 215 bps decline in group gross margin from FY2024 (40.0%) to FY2025 (37.8%) is SEMI's sale of a one-off deposition tool plus an SMT product mix headwind, both disclosed — not an "operations" problem.

Process-of-record is the moat, not market share. A TCB tool qualified into HBM3E 12-high at SK hynix or into TSMC's CoWoS C2S line is locked in for the life of that package. ASMPT is sole supplier for chip-to-substrate at the leading foundry's OSAT partner, with 500+ TCB tools installed globally — the largest base in the industry. That stickiness, not the headline 9% share of back-end equipment, is the asset.

The cycle hits working capital before margins. A back-end downturn shows up first in advance payments (fell from 18% of FY2024 H2 backlog to single digits at the Q1 2024 trough) and inventory days (200+ at the FY2023 trough). FCF turned negative in FY2025 even with operating profit rising, because receivables ballooned $135M as TCB shipments concentrated late in the year — a working-capital story, not a quality-of-earnings story.

2. The Playing Field

The right peer set for ASMPT is not "semiconductor equipment" — it is back-end advanced-packaging equipment, where five companies actually compete for the same customer orders and where economics differ wildly by purity of mix.

FX as of 2026-05-15: EUR/USD ~1.175, KRW/USD ~1450, JPY/USD ~155, HKD/USD ~7.78. FY2025 revenues converted at period-end rates. Multiples are approximate, scrubbed to one decimal.

The market pays for purity of advanced packaging, not size. BESI, Hanmi and Disco trade at 25x–50x EV/sales because every dollar of their revenue is in an AI-adjacent, high-margin product line. ASMPT trades at 5x EV/sales because the market is averaging a SEMI engine that deserves a BESI-like multiple with an SMT engine that deserves a TOWA-like multiple — possibly less.

Margin gap is structural, not operational. BESI's 63% GM and ASMPT's 38% are two mix profiles, not two management teams. BESI sells effectively only TCB and hybrid bonders. ASMPT sells those plus $816M of SMT, plus mainstream wire bonders. SEMI alone runs ~45% GM in 1H 2025 — close to K&S and TOWA, still well below BESI.

ASMPT's edge in the peer set is breadth and customer diversity, not unit economics. Top-5 customers are ~16% of revenue (FY2025 AR), against ~70%+ at BESI and Hanmi. That cushions the cycle but caps the per-dollar valuation.

3. Is This Business Cyclical?

Yes — deeply, but with a structural growth engine layered on top of the cycle. Demand is geared to OSAT factory utilisation (which moves with end-market chip demand) and to memory and foundry capex. When OSAT utilisation falls below ~70%, tool orders stop. The 11-quarter back-end downturn from late 2022 to early 2025 saw ASMPT continuing-operations revenue fall from $2,814M (FY2021) to $1,607M (FY2024) — a 43% peak-to-trough decline. The current upturn is driven almost entirely by TCB for AI, with mainstream still lagging.

Peaks and troughs are violent. Net margin compressed from 16% (FY2017 peak) to 4% (FY2019 trough), recovered to 14% (FY2021 super-cycle), then fell to 2.8% (FY2024). ROIC swings from the high-20s to low single digits across a typical 5-year cycle. Anchor on FY2024–25 as "normal" and the valuation reads wrong on the way up; anchor on FY2021 and it reads wrong on the way down.

This cycle is bifurcated for the first time. SEMI grew +22% in FY2025 while SMT fell 1% — and inside SEMI, TCB revenue is up roughly +50% on bookings while mainstream wire-bond is still soft. The Q2 2025 book-to-bill split: SEMI 0.85 (TCB orders pause; AP order flow lumpy), SMT 1.47 (China EV / AI server bookings). "ASMPT cycle = OSAT utilisation" now misses half the picture; the other half is HBM/AI capex.

Working capital is the real cycle indicator. Operating cash flow collapsed from $131M (FY2024) to $31M (FY2025) despite higher reported profit, almost entirely because receivables grew $135M as TCB shipments concentrated into late-year. Track inventory days and trade receivables before headline revenue.

4. The Metrics That Actually Matter

These are the five numbers that explain whether ASMPT is creating or destroying value at any point in the cycle. Headline EPS is not one of them.

Headline metrics that don't tell the story: P/E (distorted by JV disposal gain of $143M in FY2025), reported gross margin at the group level (blended), and revenue YoY (averages a +22% SEMI with a -1% SMT). Use the segment cuts.

5. What Is This Business Worth?

Value here is two engines, not one. SEMI is a high-IP, AI-cycle-leveraged advanced-packaging franchise that the market values at 20x–35x EV/sales standalone (BESI, Hanmi). SMT is a mature, mid-cycle, low-margin placement business whose peers (TOWA, Yamaha) trade at 2x–4x. ASMPT sits at ~5x blended because the market is unsure which engine drives the stock. The right lens is sum-of-the-parts — and the SOTP only becomes a priced SOTP if management actually separates them, which they have formally opened the door to.

The right comparable for SEMI is BESI and Hanmi, not "back-end equipment". On 1H 2025 segment economics, SEMI runs ~45% gross margin, ~10% segment-profit margin, and roughly half of revenue in AP. BESI runs 63% GM and 29% net margin because BESI is all AP. Each percentage-point of AP mix shift in SEMI is worth roughly one BESI-multiple notch.

SMT is the optionality. Whether SMT is worth $1.0B or $2.0B as a separate entity is less important than whether management surfaces that value. The Strategic Options review (Jan 2026) is the trigger. Retain without margin improvement and the blended discount persists; sale or spin lets the SEMI engine carry the AP multiple it has been hiding behind SMT.

Cycle-adjusted earnings matter more than current-year. FY2025 reported net profit of $116M is inflated by a $143M one-off JV disposal gain (FY2024 was depressed by $13M restructuring). Through-cycle EPS over the last 10 years is closer to $0.45; cycle-peak could approach $0.77. A P/E debate on current-year reported EPS is misleading.

6. What I'd Tell a Young Analyst

Watch three things, in order. First, quarterly SEMI book-to-bill — a single quarter above 1.0 confirms TCB has resumed widening the order book; a print below 0.85 says AI/HBM capex is digesting. Second, the SMT Strategic Options outcome — sale, spin, retention, or partnership; the largest near-term re-rating catalyst the company has. Third, the working-capital line — receivables growth outpacing revenue growth (as in FY2025) is a yellow flag even when bookings look strong.

What the market is most likely getting wrong. Two things in opposite directions. It is over-extrapolating SMT weakness — Q2 2025 SMT book-to-bill of 1.47, driven by China EV and AI server bookings, is the highest in three years and inconsistent with the "SMT is just industrial drag" narrative. And it is under-pricing the optionality of separation — the consolidated stub is not the right basis for valuation if the parts are about to be unbundled.

What would change the thesis. Loss of POR at a major HBM customer (especially if Hanmi expands beyond SK hynix into Micron or Samsung) would compress the SEMI franchise. A new BIS rule extending advanced-packaging export controls to mainland China memory packaging (currently below threshold) would hit ASMPT's largest geographic segment directly. A "retain as-is" outcome from the SMT review would lock in the conglomerate discount for the cycle.