← research bank

Advanced Packaging: Equipment Suppliers Capture the Scarcity Premium

gen-advanced-packaging-bottleneck · conviction — · status open · horizon — · as of 2026-08-07

Equipment suppliers enabling the bottleneck—Besi, ASMPT, Ajinomoto—are the structural winners, not the capacity owners. Besi's 68.7% revenue growth and 29.3% operating margin dwarf ASE Technology's 8.4% growth at 7.9% margin despite ASE sitting $72bn larger. ASMPT grew 45% while upstream of assemblers showing single-digit expansion.
Robust to undisclosed shares. 1 derived input under this thesis; redrawing every supply weight the industry does not publish moves none of them by more than 25%. Computed from evidence at most 15 days old (oldest input: ajinomoto).

Exhibits

Exhibit 1Relative performance, indexed to 100How the names in this thesis have traded against SOXX.
86185285SOXX 22512mo, indexed to 100 at start · dashed = SOXX benchmark

Series available as data/gen-advanced-packaging-bottleneck.csv

Exhibit 2What the conviction is actually made ofEach premise and the number it composes to. A conjunction of plausible premises is far weaker than any of them.
Equipment suppliers enabling advanced packaging capture margin and growth that assemblers and substrate makers cannot, inverting tBE Semiconductor Industries (Besi) †100.0%BE Semiconductor Indust… — revenue growth $… †100.0%BE Semiconductor In… — operating margin 29.… †100.0%ASE Technology Holding Co., Ltd. †100.0%ASE Technology Holding C… — revenue growth… †100.0%ASE Technology Holdi… — operating margin 7.… †100.0%BE Semiconductor Industries (Besi) supplies…100.0%Advanced Packaging Bottleneck82.0%COMPOSED (and)82.0%

† 6 premises marked supporting — shown and arguable, but the conclusion does not depend on them, so they are not multiplied into the composed figure. Citing a filed figure should not cost conviction.

82% if the 2 gates are independent, 82% if they move together. They are claims about one industry, so the truth is between and nobody can say where. Treat this as an ordering device rather than a calibrated probability — the ranking of premises is the information, not the level.

Weakest link: Advanced Packaging Bottleneck at 0.82 — Assumes bottleneck status persists through 2027; easing via Intel/Samsung 2.5D or TSMC CoWoS-L ramps would redistribute pricing power downstream

ASMPT and Ajinomoto exhibit the same inversion at smaller scale, confirming the mechanism operates across multiple equipment and mASMPT †100.0%ASMPT — revenue growth $45 †100.0%Ajinomoto (ABF) †100.0%Unimicron Technology Co… — revenue growth $… †100.0%ASMPT supplies Advanced Packaging Bottleneck100.0%Advanced Packaging Bottleneck78.0%COMPOSED (and)78.0%

† 4 premises marked supporting — shown and arguable, but the conclusion does not depend on them, so they are not multiplied into the composed figure. Citing a filed figure should not cost conviction.

78% if the 2 gates are independent, 78% if they move together. They are claims about one industry, so the truth is between and nobody can say where. Treat this as an ordering device rather than a calibrated probability — the ranking of premises is the information, not the level.

Weakest link: Advanced Packaging Bottleneck at 0.78 — Requires sustained capex urgency; risk is utilization catches supply and tool orders normalize, though 2027 HBM4 and CoWoS-R transitions argue against

The margin structure will persist because each incremental wafer of AI accelerator demand triggers step-function tool purchases whSK Hynix — revenue growth $198 †100.0%Taiwan Semiconducto… — operating margin 56.… †100.0%Advanced Packaging Bottleneck75.0%Advanced Packaging Bottleneck — input-cost… †100.0%COMPOSED (and)75.0%

† 3 premises marked supporting — shown and arguable, but the conclusion does not depend on them, so they are not multiplied into the composed figure. Citing a filed figure should not cost conviction.

One gating premise, so the conclusion is exactly as strong as it. The figure is an ordering device, not a calibrated probability — see how the numbers are made.

Weakest link: Advanced Packaging Bottleneck at 0.75 — Durability assumption: bottleneck must outlast typical 18-month tool delivery and qualification cycles for margin capture to compound; HBM supply easi

The variant

Consensus

The advanced packaging bottleneck enriches the assemblers closest to end demand—TSMC in CoWoS, SK Hynix in HBM—who command pricing power as hyperscalers queue for capacity. Substrate makers and OSATs absorb input cost pressure while racing to add lines. The margin expansion story lives downstream where scarcity meets inelastic AI budgets.

Variant

Equipment suppliers enabling the bottleneck—Besi, ASMPT, Ajinomoto—are the structural winners, not the capacity owners. Besi's 68.7% revenue growth and 29.3% operating margin dwarf ASE Technology's 8.4% growth at 7.9% margin despite ASE sitting $72bn larger. ASMPT grew 45% while upstream of assemblers showing single-digit expansion. The scarcity premium accrues to those gating capacity additions, not running the lines, because every CoWoS expansion multiplies tool orders while hyperscaler demand makes packaging a pass-through cost.

Differentiator

Supply-chain analysis focuses on who ships to Nvidia; it misses that capex desperation transfers margin to toolmakers. The bottleneck doesn't enrich the bottleneck—it enriches whoever sells the drill.

Falsifiers

Open questions

Reasoning chain

Equipment suppliers enabling advanced packaging capture margin and growth that assemblers and substrate makers cannot, inverting the expected value distribution along the supply chain VALID
premises

71% marginal growth accrues upstream of assembly despite 27% of ASE's market cap, suggesting the constraint monetizes at the equipment layer not the capacity layer

ASMPT and Ajinomoto exhibit the same inversion at smaller scale, confirming the mechanism operates across multiple equipment and materials nodes rather than being Besi-specific VALID
premises

Three-tier gradient—tools 45-69%, substrates 16%, assembly 6-8%—maps to distance from the capacity gate, not the demand signal

The margin structure will persist because each incremental wafer of AI accelerator demand triggers step-function tool purchases while packaging remains a fixed-percentage pass-through cost that assemblers cannot retain VALID
premises

Hyperscaler willingness to pay doesn't flow to assemblers because toolmakers sit between demand and supply expansion, capturing the desperation premium each capacity addition represents

Write-up

Pre-filled skeleton: gen-advanced-packaging-bottleneck.md