Long-Term Thesis
Long-Term Thesis
1. Long-Term Thesis in One Page
The long-term thesis is that ASMPT becomes a focused advanced-packaging franchise — anchored on TCB process-of-record at TSMC CoWoS and across multiple HBM customers — that compounds through the AI hardware build-out and carries that qualification edge into hybrid bonding before HBM5 ramps around 2028-29. That is not the business the consolidated FY2025 P&L shows. The 5-to-10-year case works only if three things hold simultaneously: (a) SEMI isolates from the dilutive SMT business so the moat shows up in group economics, (b) sustained R&D intensity above 15% on a rising SEMI revenue base closes the hybrid-bonding gap to BESI within 24-36 months, and (c) the C2S TCB qualification at the leading foundry survives one generation transition without re-qualification. With all three, ASMPT can compound revenue 8-10% per year on structurally higher gross margin than FY2025's 37.8%. Break any one and ASMPT remains a cyclical broad-line equipment maker that earns its cost of capital but does not compound — closer to KLIC than to BESI.
Thesis Strength (1-5)
Durability 5-10 yr (1-5)
Reinvestment Runway (1-5)
Evidence Confidence (1-5)
Scoring (1=weakest, 5=strongest): Thesis Strength = Medium (3), Durability = Medium (3), Reinvestment Runway = High (4), Evidence Confidence = Medium (3).
The single most load-bearing fact for the long-term thesis is that the moat lives in only one half of the company. SEMI ran a 45.5% gross margin in 1H 2025 — within striking distance of BESI's pure-play 63% — while SMT ran 32% on commoditising mid-cycle economics. The 5-to-10-year question is not "is ASMPT a good business?" — it is "will management actually separate the good business from the average one, and can it defend the good business through the next packaging-technology transition?"
2. The 5-to-10-Year Underwriting Map
These are the seven things that have to be true for ASMPT to outcompound its peer set through the back-end equipment cycle. Each row is paired with the evidence that exists today, the structural reason it could last, and the single fact pattern that would break it.
Driver #2 — hybrid bonding — matters more than any other row. Drivers 1, 3, 4, 5, 6, 7 can all hold and the franchise still de-rates to a follower if BESI extends its 24-month hybrid-bonding lead into a generational moat at HBM5. Conversely, every other driver could disappoint by 20% and the thesis still compounds at low-double-digits provided ASMPT lands hybrid bonding before the HBM5 transition. Two decades of industry history: being on the right product at the wrong node compresses multiples for half a cycle; the inverse compounds for two.
3. Compounding Path
ASMPT revenue has compounded at roughly 0.6% per year over the last decade (FY2015 HK$13.0B → FY2025 HK$13.7B) because two full cycles cancelled out. The forward question is whether the FY2026-2030 path looks like FY2015-2025 or like a structurally re-based AP franchise. The realistic 5-to-10-year base case lifts revenue to HK$22-28B with mix shifting decisively to SEMI / AP and gross margins recovering toward the historical 40-44% band as TCB and hybrid bonding dominate SEMI.
Three observations decide the compounding math. Gross margin discipline survived a 43% peak-to-trough revenue decline — group GM held at 37.8% in FY2025 versus 41-44% at cycle peaks. That floor makes the base-case 40-44% steady-state credible if mix shifts toward SEMI. The balance sheet is the optionality — net cash actually grew from HK$2.1B (FY2021 peak) to HK$3.3B (FY2025 trough), letting R&D rise through the downturn (HK$1.92B FY2024 vs HK$1.65B FY2021). ASMPT can fund the hybrid-bonding race from operations alone for three to four years even at trough margins. Through-cycle ROE has averaged 13-14% over the decade — respectable, below the 18-22% of a true compounder. A BESI-style multiple only arrives if SEMI separates and standalone returns approach 20%+.
4. Durability and Moat Tests
Five durability tests that decide whether the moat is structural or rented. At least one is competitive (drives whether ASMPT keeps customers), at least one is financial (drives whether the economics hold), and the rest cover the surrounding ecosystem the moat depends on.
Tests 1 and 2 are binary — they decide wide moat vs rented. Test 3 is the financial proof that the moat is converting to margin, not just share. Tests 4 and 5 cover the surrounding system — service tail and management discipline — that holds the moat across years. The combination matters more than any single one, but losing test 2 (hybrid bonding) by FY2028 dominates the long-run rating: the other four can be passing while the underlying technology base is being obsoleted.
5. Management and Capital Allocation Over a Cycle
The capital allocation pattern over the last decade is the cleanest part of the long-term thesis and the part most at risk in the next 24 months. Through the FY2023-25 trough, management did three things that mattered: raised R&D into the downturn (HK$1.92B FY2024 vs HK$1.65B at the FY2021 peak), preserved net cash (HK$2.1B → HK$3.3B), and pruned the portfolio (NEXX to Applied Materials, AAMI monetised for HK$862M, AEC Shenzhen liquidated, SMT under Strategic Options review). AAMI proceeds came back to shareholders as a HK$0.79 special dividend — a clean disposal-and-return pattern minorities want to see. Dilution is modest: share count rose 411.2M (FY2021) to 416.7M (FY2025) — 0.27% per year, SBC ran 1% of revenue, trustee bought 353k shares on-market to feed the employee scheme rather than issuing new shares.
The 5-to-10-year question is whether this discipline survives the CEO succession. Robin Ng announced retirement February 2026 with no internal successor named; the chairman was promoted May 2025 after an 18-year INED stint; executive directors combined own 0.28%; the CEO has been a net seller of HK$64M over the last twelve months with zero offsetting buys. That last fact is the single largest governance flag for a long-duration owner — the people closest to the order book are monetising into the re-rating, not buying alongside it. ASM International's 24.65% stake acts as a technically literate strategic anchor and has, so far, supported focus over diversification.
The most reassuring pattern is rising R&D through the trough. The most concerning is the executive ownership profile combined with sustained insider selling — and the unresolved CEO succession into a transition year where the SMT review and hybrid-bonding race both need decisive capital-allocation leadership. Long-duration owners should weight the next 12 months — succession announcement, SMT review outcome, first capital-markets day under the new CEO — as the largest single piece of the 5-to-10-year rating still in motion.
6. Failure Modes
Five thesis-breakers, in order of severity. Each is a specific, observable event — not a generic "execution risk."
The two Critical failure modes — BESI hybrid-bonding extension and TSMC re-qualifying a second-source — are not independent. Both stem from the same dynamic: ASMPT”s lead is at the current generation of advanced packaging, not the next. Either individually would re-rate the SEMI franchise from pure-play AP to broad-line equipment economics; together they would invalidate the long-term thesis and the equity would re-price toward KLIC-adjacent multiples. Both signal well in advance via R&D output (hybrid-bonding customer count, named C2S wins) — neither is the kind of risk that crystallises overnight.
7. What To Watch Over Years, Not Just Quarters
These are the five multi-year observations that update the long-term thesis. Each has a specific metric, a time horizon, and a validation / refutation pattern. None of them is a single-quarter print.
The long-term thesis updates most on a named hybrid-bonding production-line POR at an HBM or foundry customer before BESI's lead extends past 36 months — that single observation, more than any cyclical revenue print, decides whether ASMPT compounds as a focused advanced-packaging franchise through HBM5 or whether the moat unwinds at the exact node where AI volume peaks.