← Back to blog
Essay No. 039  ·  AI Infrastructure  ·  Melbourne, Australia
AI Infrastructure Kulicke & Soffa KLIC Wire Bonding Advanced Packaging TCB Hybrid Bonding Vertical Wire Memory HBM Power Semi OSAT TSMC CoWoS SoIC

The Back-End Bottleneck.Original analysisNot investment advice

Why Kulicke & Soffa’s wire-bonder cycle is becoming an AI packaging, memory, and power-assembly story.
PM
Pugalenthi Magendran
April 2026  ·  Melbourne, Australia
12 min read

KLIC’s 2021 story was a wire-bonder supercycle. The 2026 story is a back-end interconnect test. Wire bonding still matters because the world still needs huge volumes of cheap, reliable packages. But the future is moving toward vertical wire, thermo-compression bonding, hybrid bonding, advanced dispensing, power modules, and memory interconnect. KLIC is not automatically an AI packaging winner, but it sits in the part of the semiconductor stack that is becoming harder to ignore: the back end, where chips become systems.

In 2021, Kulicke & Soffa looked like a wire-bonder supercycle. Long lead times. OSAT capex. Trailing-edge capacity. Mini-LED upside. Management under-guiding. Investors still treating the company like the next downturn was one quarter away.

The uploaded SemiAnalysis article argued that Wall Street was missing the deeper pattern. KLIC was cyclical, yes. But there was a secular force inside the cycle: automotive, IoT, edge, 5G, and trailing-edge silicon content were all increasing, and most of those chips still needed wire bonding.1

That was the old thesis. In 2026, the better version is larger.

KLIC is not only a wire-bonder company. It is a back-end interconnect company trying to move from old-school bonding into the next assembly stack.

Key idea

Kulicke & Soffa is a back-end interconnect company trying to turn wire-bonding leadership into a broader assembly platform. The 2021 thesis was right that trailing-edge silicon and OSAT capacity created a secular force inside a cyclical wire-bonder market. But the 2026 update is bigger: AI, memory bandwidth, chiplets, power modules, and advanced packaging are pulling KLIC toward vertical wire, fluxless TCB, hybrid bonding, advanced dispensing, and power assembly. The risk is that wire bonding remains cyclical, hybrid bonding is fiercely competitive, and KLIC must prove that its advanced solutions can become large, durable businesses rather than temporary cycle beneficiaries.


I. The 2021 thesis was right, but incomplete

In August 2021, Dylan Patel published a SemiAnalysis piece arguing that KLIC’s cycle was being mispriced. Channel checks showed wire-bonder lead times of 42–52 weeks and later above 52 weeks for main business categories. ASE / Amkor / Chinese OSAT capex was supportive, and the piece increased its estimate of new trailing-edge capacity from ~200,000 WPM to ~270,000 WPM. Mini-LED was running better than expected, and a future micro-LED opportunity was implied. The piece called KLIC’s “advanced packaging” business positive while noting it was not true 2.5D / 3D, and argued management was sandbagging guidance. The deeper insight was "secular within cyclical" — automotive, IoT, edge, 5G, and trailing-edge silicon content kept supporting wire-bonder demand even as wire bonding slowly lost share of total package type.1

2021 thesis

KLIC was not just riding a temporary equipment shortage. It was riding the fact that automotive, IoT, edge, 5G, and trailing-edge chips still needed enormous low-cost assembly capacity.

Diagram · Secular within cyclical
Cyclical

Wire-bonder cycle

OSAT capex booms, lead times stretch, orders ramp, and then digestion arrives. Wire bonders ride the cycle as long as cheap-package volume keeps growing.

Secular

Trailing-edge content

Automotive, IoT, edge, 5G, industrial, and mature-node logic keep adding silicon content. Each cycle starts from a higher base of low-cost assembly demand.1

A simplified, original split. The 2021 framing remains useful in 2026 as the bedrock under the more advanced packaging story.

II. Chips do not ship as naked dies

The front end gets attention: EUV, leading-edge nodes, transistor scaling, wafer fabs, GPUs, HBM. But the back end turns dies into usable products: attach, bond, interconnect, encapsulate, protect, test, assemble, and connect to substrates, modules, and systems.

Diagram · Front end vs back end
Front end

Make the die

  • EUV / advanced nodes
  • Transistor scaling
  • Wafer fabs
  • Process integration
  • Yield learning
Back end

Make it a system

  • Attach & bond
  • Interconnect & encapsulate
  • Substrate / module integration
  • Power & thermal paths
  • Test & reliability
A simplified, original split. Front end and back end answer different questions; both decide whether the chip ships.
Diagram · Wafer to system
01
Wafer fab
front end
02
Die
singulation
03
Attach
die / substrate
04
Bond
wire / TCB / hybrid
05
Package
encapsulate / mold
06
Module
power / memory
07
System
board / rack
A simplified, original 7-step flow. KLIC sits across the bond / package / module steps, with growing relevance to system-level packaging.

The most advanced die in the world is still useless until the back end turns it into a system.


III. Wire bonding still matters

Wire bonding is old, but not dead. It remains cheap, mature, reliable, high-throughput, widely qualified, and good enough for huge categories of trailing-edge chips — automotive controllers, industrial chips, sensors, analog, power devices, IoT, commodity memory, consumer electronics, and mature-node logic.

Wire bonding is not the future of every package. But it remains the present of enormous package volume.


IV. The back end is becoming strategic

TSMC’s 2026 North America Technology Symposium materials describe a packaging roadmap with 5.5-reticle CoWoS in production, 14-reticle CoWoS planned for 2028 integrating roughly 10 large compute dies and 20 HBM stacks, a 40-reticle SoW-X System-on-Wafer expected in 2029, plus SoIC 3D stacking and COUPE co-packaged optics.7 That is not all KLIC-addressable, but it is the macro proof: the semiconductor frontier is moving from die-level scaling to package-level integration.

The industry is no longer only asking whether it can make a better die. It is asking whether it can connect many dies, memories, substrates, optical engines, and power paths into one reliable system.


V. The rebound is real

K&S’s Q2 FY2026 results frame revenue at ~$242.6M, gross margin at ~49.3%, net income at ~$35.1M, and non-GAAP EPS at ~$0.79, with demand described as stronger than expected across general semiconductor, memory, automotive, and industrial end markets. Q3 FY2026 revenue guidance is roughly $310M ± $20M with non-GAAP EPS guidance of roughly $1.00 ± 10%.3

Card · K&S Q2 FY2026, simplified
~$242.6M
Q2 FY2026 revenue3
~49.3%
Gross margin3
~$0.79
Non-GAAP EPS3
~$310M
Q3 FY2026 revenue guide (midpoint)3
Figures as reported by K&S. Treated as company disclosures; this essay is not investment advice and contains no price targets or EPS forecasts.

A beat-and-raise quarter proves demand. It does not by itself prove strategic transition.


VI. TCB is the real option

K&S’s Q2 FY2026 release describes raising fiscal 2026 capex from about $12M to about $22M to expand TCB production, with the additional capacity supporting up to about $400M in annual TCB system sales.3 The FY2025 10-K positions APTURA as ultra-fine-pitch, fluxless, direct-copper thermo-compression bonding for chiplet-based advanced packages.2

Diagram · TCB, simplified
Input

Die + substrate

Two surfaces with very fine-pitch interconnect pads. Cleanliness and alignment matter.

Output

Bonded interconnect

Heat and pressure form fine-pitch interconnects. Fluxless direct-copper TCB enables dense, low-resistance copper-to-copper connections for chiplets.2

A simplified, original visual. Wire bonding is KLIC’s installed-base machine; TCB is its advanced-packaging bridge.

The bonding ladder runs from cheap-and-mature to dense-and-strategic. KLIC plays across it.

Diagram · Bonding evolution ladder
Rung 01

Wire bonding

Cheap, mature, high-throughput. Huge low-cost package volume.

Rung 02

Vertical wire

Denser interconnect for stacked memory.5

Rung 03

TCB

Fine-pitch, fluxless direct-copper bonds for chiplets.2

Rung 04

Hybrid bonding

Direct wafer / die bonding at very fine pitch.8

A simplified, original ladder. Density and process complexity rise from left to right; competitive intensity rises with them.

Wire bonding is KLIC’s installed-base machine. TCB is its advanced-packaging bridge.


VII. Memory is becoming a bonding problem

K&S’s March 2026 memory portfolio expansion covers Ball Bonding, Vertical Wire, Advanced TCB, and future Hybrid Bonding, with ProMEM described as delivering up to ~20% higher throughput, Vertical Wire designed to increase interconnect density and reduce package footprint for stacked memory, and TCB expected to grow about ~70% sequentially in fiscal 2026. The company references HBM, high-bandwidth flash, and high-density DRAM as target categories.5

Diagram · Memory interconnect roadmap, simplified
Volume

Ball Bonding

commodity / mature memory5
Denser

Vertical Wire

stacked memory footprint5
Advanced

TCB

fine-pitch, fluxless copper5
Future

Hybrid Bonding

HBM4-class density58
A simplified, original roadmap. KLIC plays across all four rungs, but the value migrates rightward as memory density and bandwidth rise.

KLIC’s memory opportunity is not only more NAND stacks. It is the fight to keep bonding relevant as memory becomes denser and more bandwidth-driven.


VIII. Vertical wire, explained simply

Traditional wire bonding connects pads with looped wires. Vertical wire aims to create more direct vertical interconnect structures, helping reduce footprint and increase interconnect density in stacked packages. It is a K&S solution aimed at stacked memory and denser packaging, not a universal replacement for HBM hybrid bonding.5

Vertical wire is KLIC trying to stretch wire-bonding DNA into denser memory packages.


IX. Hybrid bonding is the hard wall

Reuters reporting describes Besi receiving hybrid-bonding orders from two leading memory producers for HBM4 applications and follow-on orders from a leading Asian foundry for logic applications, alongside Applied Materials taking a roughly 9% stake in Besi, framed as recognition of hybrid bonding’s strategic importance.89

Competitive wall

KLIC can participate. KLIC does not automatically own hybrid bonding.

Hybrid bonding is being contested by larger, better-capitalised players. K&S talks about future hybrid bonding in its memory portfolio, but the early HBM4 momentum and the Applied / Besi stake make clear that the hybrid-bonding category does not belong to any single supplier yet.589

KLIC can participate in the advanced-packaging transition, but it is not automatically the winner of the hybrid-bonding era.


X. Power semiconductors add another layer

K&S’s March 2026 ASTERION-TW announcement frames an ultrasonic terminal welding system for power-module manufacturing, targeting renewable energy, transportation, and data-center markets, with capability to bond copper terminals up to ~2mm thick and weld-placement repeatability of ±40 microns.6

Diagram · Power assembly map
Data center

Rack PSUs + 800 VDC

AI factory power6
EVs

Traction / OBC / BMS

power modules6
Renewables

Solar / wind

inverter modules6
Industrial

Motors / HVDC

drives, transmission6
A simplified, original map. K&S does not need to make power semiconductors; it needs to help assemble them.

If electrification increases power-module complexity, power assembly becomes part of the same back-end bottleneck.


XI. Mini-LED was the earlier non-semi option

The 2021 SemiAnalysis piece described mini-LED running better than expected, with internal estimates roughly $75M (2021), $100M (2022), and $150M (2023), and a future micro-LED opportunity mentioned as a possible adjacency.1 Mini-LED was an early example of K&S applying precision assembly / bonding knowledge outside the classic semiconductor cycle. In the 2026 framing it is no longer the centre of the story; TCB, memory, and power assembly are.

Mini-LED showed KLIC could move process knowledge into adjacent assembly markets. TCB, memory, and power modules are the more important 2026 tests.


XII. China exposure cuts both ways

K&S’s FY2025 10-K reflects significant exposure to customers headquartered in or shipping to China across mature-node OSAT, packaging, and assembly demand, alongside the standard export-control, trade-policy, and cyclicality risk language any back-end-equipment supplier carries.2

China is both demand engine and risk vector.


XIII. The bear case is real

K&S’s own 10-K language acknowledges that alternative packaging technologies — hybrid bonding, TCB, flip-chip bonding, wafer-level packaging — could reduce demand for wire-bonding equipment over time, and that a majority of revenue still depends on wire bonding. The filing discusses cyclicality, order cancellations, deferred purchases, inventory, and market-downturn risk.2

Reading the dependence

KLIC can participate without capturing enough value.

The risk is not that KLIC misses the back-end transition entirely. The risk is that it participates without capturing enough value — advanced solutions ramp slowly, wire-bonder share slips faster than expected, hybrid bonding stays in stronger hands, and the mix never improves enough to lift the business out of equipment-cycle behaviour.2


XIV. Competitive landscape

Bonding and advanced-packaging equipment competition is intensifying. Besi sits at the centre of hybrid bonding. Applied Materials took a roughly 9% stake in Besi. ASMPT competes broadly in assembly and packaging equipment. Shinkawa, Yamaha Robotics, and Canon Machinery compete in adjacent categories. Disco overlaps in dicing and grinding. Tokyo Electron, Lam, and Applied Materials touch front-end-related advanced-packaging processes. OSATs have internal capabilities. China’s domestic equipment suppliers add another vector.89

The back end is attractive because it is becoming strategic. That also means stronger competitors are moving in.


XV. The actual 2026 thesis

The correct claim is not “KLIC is just a wire-bonder company.” That is too old.

2026 thesis

Kulicke & Soffa is a back-end interconnect company trying to turn wire-bonding leadership into a broader assembly platform. The 2021 thesis was right that trailing-edge silicon and OSAT capacity created a secular force inside a cyclical wire-bonder market. But the 2026 update is bigger: AI, memory bandwidth, chiplets, power modules, and advanced packaging are pulling KLIC toward vertical wire, fluxless TCB, hybrid bonding, advanced dispensing, and power assembly. The risk is that wire bonding remains cyclical, hybrid bonding is fiercely competitive, and KLIC must prove that its advanced solutions can become large, durable businesses rather than temporary cycle beneficiaries.

Diagram · KLIC product stack, simplified
01
Wire bonding ball + wedge, mature volume engine2
Installed base
02
Vertical wire denser stacked memory packages5
Memory bridge
03
APTURA TCB fluxless direct-copper for chiplets23
Advanced bridge
04
Hybrid bonding (future) HBM-class density, contested category58
Aspiration
05
Power assembly ASTERION-TW terminal welding6
New leg
A simplified, original stack. The platform thesis depends on rows 2–5 becoming large enough to change the mix, not just being announced.

XVI. What could break the thesis?

KLIC can participate in advanced packaging but still remain mostly a cyclical back-end equipment supplier.

Bear case · what could break the thesis
  1. Share slip. Wire bonding continues losing share of total package value over time.2
  2. Cycle dominates mix. KLIC remains mostly cyclical despite advanced-packaging efforts.
  3. TCB shortfall. TCB demand does not reach the implied capacity target.3
  4. Hybrid bonding concentration. Besi and larger equipment vendors dominate the category.8
  5. Trials ≠ volume. Advanced-packaging trials do not convert into durable production volume.
  6. Memory lumpiness. Memory capex is lumpy and can compress mix improvements.
  7. China softness. China demand slows or becomes more price-competitive.2
  8. Export-control disruption. Policy moves can disrupt China revenue paths.
  9. OSAT pause. OSAT capex pauses after a strong cycle.
  10. Power assembly slow ramp. ASTERION-TW takes longer to scale than expected.
  11. Order cancellations. Customers delay or cancel under macro stress.2
  12. Margin pressure. Advanced solutions show lower margins than expected.

XVII. What could break the bear case?

KLIC does not need to win every advanced-packaging layer. It needs to convert enough bonding expertise into TCB, vertical wire, memory, and power assembly to become more than a wire-bonder cycle stock.

Bull case · what could break the bear
  1. Bonding depth. Decades of bonding-process knowledge.1
  2. Wire bonding necessary. Required for massive low-cost package volume.
  3. TCB tailwind. Chiplet and advanced-memory growth pulls TCB demand.3
  4. Vertical Wire defence. Helps KLIC extend memory packaging relevance.5
  5. Power module growth. AI data centers, EVs, renewables, industrial.6
  6. Customer reach. Long OSAT / IDM relationships ease new-tool adoption.2
  7. Adjacency expansion. Advanced dispensing and panel-level tools broaden the platform.
  8. Mix shift. Even partial advanced-packaging share improves business quality.
  9. Back end strategic. Package-level integration is mainstreaming.7

KLIC does not need to win every advanced-packaging layer. It needs to convert enough bonding expertise into TCB, vertical wire, memory, and power assembly to become more than a wire-bonder cycle stock.


XVIII. What to watch

What to watch
  • Q3 FY2026 revenue vs guidance.3
  • TCB order growth.3
  • TCB capacity ramp pace.3
  • Whether TCB reaches meaningful annual sales.
  • APTURA adoption.2
  • Vertical Wire customer adoption.5
  • ProMEM throughput adoption.5
  • Hybrid bonding product progress.5
  • Besi HBM4 momentum.8
  • Power assembly revenue.6
  • ASTERION-TW design wins.6
  • China revenue share.2
  • OSAT capex trend.
  • Memory capex trend.
  • HBM packaging path.8
  • Advanced packaging capex at TSMC / Samsung / Intel.7
  • Wire-bonder share of total K&S revenue.2
  • Gross margin by product mix.
  • Backlog quality and cancellation risk.2
  • Advanced dispensing and panel-level progress.

Glossary

A short reference for the vocabulary used above. Definitions are simplified.

Glossary
Back end
Semiconductor packaging, assembly, bonding, and test processes after wafer fabrication.
OSAT
Outsourced semiconductor assembly and test company.
Wire bonding
Using fine wires to connect chip pads to a package.
Ball bonding
Common wire-bonding method using a ball-shaped bond.
Wedge bonding
Wire-bonding method often used in power and specialty packages.
TCB
Thermo-compression bonding, using heat and pressure for fine-pitch interconnects.
Fluxless bonding
Bonding without flux chemicals; helpful for cleaner fine-pitch interconnects.
Direct-copper bonding
Copper-to-copper interconnect method.
Hybrid bonding
Direct wafer / die bonding with very fine interconnect pitch.
Vertical Wire
K&S interconnect approach for denser stacked memory packages.
HBM
High-bandwidth memory used near AI accelerators.
CoWoS
TSMC advanced packaging technology for AI / HPC chips.
SoIC
TSMC 3D stacking technology.
Power module
Packaged power-semiconductor system used in EVs, renewables, industrial systems, and data centers.
Trailing-edge node
Older semiconductor process node, often 28nm and above.
Capex
Capital expenditure.
Backlog
Customer orders not yet fulfilled.

XIX. The back-end bottleneck

KLIC’s 2021 story was a wire-bonder supercycle.

The 2026 story is a back-end interconnect test.

Wire bonding still matters because the world still needs huge volumes of cheap, reliable packages. But the future is moving toward vertical wire, TCB, hybrid bonding, advanced dispensing, power modules, and memory interconnect.

KLIC is not automatically an AI packaging winner. But it sits in the exact part of the semiconductor stack that is becoming harder to ignore: the back end, where chips become systems.

The bear case is real. Wire-bonder dependence is high. Hybrid bonding is contested by Besi and larger equipment vendors. China exposure cuts both ways. Advanced-solution revenue has to grow before wire-bonder share loss becomes a serious problem. None of those concerns disappear because the back end is fashionable.

The bull case is also real. The packaging frontier is moving from die-level to package-level to rack-level integration, and bonding sits underneath every step of that transition. KLIC has decades of bonding-process depth, customer relationships across OSATs and IDMs, an expanding memory portfolio, an advancing TCB roadmap, and a new power-assembly leg. None of that guarantees outcome. All of it raises the floor.

The 2026 question is straightforward: how much of the back-end strategic premium can KLIC actually capture? Watch TCB, hybrid bonding, vertical wire, and power assembly the way the 2021 piece watched OSAT capex. That is how the answer arrives.

That is the back-end bottleneck.


1 Patel, D. (Aug 2021). Kulicke & Soffa Industries, $KLIC, Crushes Earnings But Continues To Sandbag, $6 EPS ’21, $6.75 EPS ’22. SemiAnalysis. Historical anchor for the 2021 framing — 42–52 week wire-bonder lead times moving above 52 weeks, ASE / Amkor / Chinese OSAT capex, the trailing-edge capacity estimate moving from ~200,000 WPM to ~270,000 WPM, mini-LED / micro-LED discussion, management sandbagging, "secular within cyclical" framing, and automotive / IoT / edge / 5G / trailing-edge silicon content backing wire-bonder demand. Used as inspiration only. No content, structure, or charts reproduced. The 2021 EPS values referenced in the source title are not reused as forecasts in this 2026 essay.

2 Kulicke & Soffa Industries. Investor relations and SEC filings. Source for FY2025 10-K language used in this essay, including segment framing, wire-bonding dependence, APTURA TCB / vertical-wire positioning, advanced-packaging strategy, China exposure and trade-policy risk, cyclicality / cancellation / inventory language, and the disclosure that alternative packaging technologies could reduce demand for wire-bonding equipment over time.

3 Kulicke & Soffa (May 2026). Q2 FY2026 results. Source for Q2 FY2026 revenue ~$242.6M, gross margin ~49.3%, net income ~$35.1M, non-GAAP EPS ~$0.79, Q3 FY2026 revenue guidance ~$310M ± $20M, non-GAAP EPS guidance ~$1.00 ± 10%, capex raised from ~$12M to ~$22M to expand TCB production, and the framing that the added capacity supports up to ~$400M in annual TCB system sales.

4 Kulicke & Soffa. Q2 FY2026 earnings presentation (Investor Relations). Product-mix, demand-driver, and TCB / memory / power-assembly framing used to support the discussion in this essay.

5 Kulicke & Soffa (Mar 2026). Memory solutions portfolio expansion. Source for Ball Bonding / Vertical Wire / Advanced TCB / future Hybrid Bonding framing, ProMEM ~20% throughput claim, Vertical Wire stacked-memory framing, TCB ~70% sequential growth expectation in FY2026, and references to HBM, high-bandwidth flash, and high-density DRAM.

6 Kulicke & Soffa (Mar 2026). ASTERION-TW launch. Source for ultrasonic terminal welding for power module manufacturing, renewable energy / transportation / data-center targeting, copper terminals up to ~2mm thick, and ±40 micron weld-placement repeatability framing.

7 TSMC (2026). 2026 North America Technology Symposium. CoWoS at 5.5R today, 14R by 2028 (~10 dies + 20 HBM stacks), 40R SoW-X by 2029, SoIC, and COUPE positioning. Used as macro proof that package-level integration is becoming the AI packaging frontier.

8 Reuters. Reporting on Besi hybrid-bonding orders, including HBM4-class memory engagements and follow-on logic orders, used as competitive context for the hybrid-bonding category.

9 Reuters. Reporting on Applied Materials taking a ~9% stake in Besi, framed as strategic recognition of hybrid bonding’s importance in advanced packaging.

10 SEMI and credible OSAT / IDM disclosures (ASE, Amkor, and others) are referenced in this essay only at the level the cited K&S, SemiAnalysis, and Reuters materials already disclose. No specific OSAT capex numbers are restated without primary-source backing.

Further reading
*   *   *

This is Essay No. 039. The topics: intelligence, AI, systems, knowledge, and the questions underneath the questions everyone else is asking. If you read this far and disagreed with any part of it, write to me. I read everything.

Pugalenthi Magendran