We use cookies to understand how our site is used so we can improve it. Necessary cookies are always on. You can accept all, reject all, or choose what to allow. Learn more
Semiconductors / OSATThe packaging boom went one way. It went the other.
Semiconductors / OSAT
The packaging boom went one way. It went the other.
Semiconductor back-end is in the middle of the largest value migration in its history: as AI and high-performance computing pull the money toward advanced packaging — 2.5D/3D, fan-out, chiplet integration — the OSAT industry is splitting into a fast, profitable frontier and a slow, commoditizing floor. This is a study of that split, read through Korea's LB Semicon, a pure-play in the one corner the migration bypassed: display-driver-IC packaging. It is the cleanest case of what the wrong side of a value migration looks like on a P&L — a recovering top line, a margin gone negative, a written-down fab, and a recapitalization to fund a pivot larger rivals reached years ago.
An advanced semiconductor package — the layer where chips are bonded, stacked and tested after they leave the fab. This is the back-end, the OSAT industry, and it is where the value of a chip increasingly settles: AI accelerators need their compute and memory dies integrated in a single package, and that integration is now worth more than the assembly it replaced. This study lives in the gap between the two — the advanced frontier where the money went, and the display-driver niche where LB Semicon stayed.
Semiconductor back-end — the assembly, packaging and testing a chip undergoes after it leaves the fab — is in the middle of the largest value migration in its history. For decades, packaging was the cheap, commoditized end of the chip: bond the die, encase it, test it, ship it. Then artificial intelligence rewrote the economics. An AI accelerator is no longer a single die; it is a cluster of compute and high-bandwidth-memory chips that have to be integrated, stacked and wired together at the package level — and that integration, called advanced packaging, has become one of the highest-value steps in the entire chain. The outsourced-assembly-and-test industry, OSAT for short, is splitting in two as a result: a fast, profitable frontier of 2.5D/3D, fan-out and chiplet work pulled by AI and high-performance computing, and a slow, commoditizing floor of conventional assembly where price is the only competition. This is a study of that split — read through Korea's LB Semicon, a pure-play stranded on the wrong side of it: a packager of display-driver ICs, the one corner of OSAT the migration left behind.
The headline numbers make OSAT look uniformly healthy, and that is the trap. The outsourced-assembly-and-test market is large and growing — but the aggregate hides the split underneath it. The relevant question is not how fast OSAT grows; it is where inside OSAT the growth concentrates. In 2024, by one house's count, roughly 68 percent of the devices OSAT players packaged were destined for advanced formats — flip-chip, fan-out — rather than conventional assembly, and the share is still climbing as AI demand spreads. The value, in other words, is rushing to one end of the market. LB Semicon sits at the other: it bumps, films and tests the driver chips that run display panels, a mature, panel-cycle-dependent niche with no connection to the AI supercycle. This is a study of that migration — the frontier where the money went, the floor where it did not, and the diversification divide between them — read through the company that shows, more clearly than any forecast, what the floor looks like on a P&L.
Share of OSAT-packaged devices destined for advanced formats (flip-chip, fan-out) in 2024, reflecting AI/HPC and chiplet demand (Market Growth Reports). The OSAT market itself is ~$46–73bn in 2025 across houses — the low cluster ~$44–47bn (Research and Markets, Mordor, Market Research Future) on a narrower service definition, up to ~$73bn on a broader one (Market Growth Reports) — growing ~5–11% to ~$77–98bn by 2030–35. Packaging is ~76.8% of OSAT revenue (Mordor). The aggregate is healthy; the composition is migrating to advanced packaging, the part LB Semicon does not serve. Sources: Mordor Intelligence; Research and Markets; Market Research Future; Market Growth Reports.
Where the value migrated
Strip OSAT into its layers and the gradient is stark. The advanced-packaging slice — 2.5D and 3D stacking, silicon interposers, fan-out wafer-level packages — is the engine. Yole Group, the recognized specialist in the field, sizes it at roughly US$42–46 billion in 2024 and projects it to US$79.4 billion by 2030, a ~9.6 percent compound rate; secondary houses reach further, with Market Data Forecast putting advanced packaging near US$140 billion by 2034 at an ~11.8 percent rate. Inside that, the 2.5D/3D platform is the fastest of all — Yole expects its revenue and wafer shipments each to expand more than 20 percent a year, the go-to architecture for AI accelerators. Set against the conventional-OSAT aggregate growing ~5–11 percent, and against the display-driver packaging that LB Semicon does, which tracks the low-single-digit panel cycle, the message is unambiguous: every dollar of OSAT growth is being claimed at the advanced-packaging end. The OSAT winners — ASE, Amkor, JCET globally, Hana Micron in Korea — are the houses that built capacity there. The laggards are the ones left on the commodity floor.
LB Semicon's own filings reach for the right language: they describe OSAT as moving "from simple assembly to system integration," the advanced-packaging narrative in miniature. But aspiration is not capacity. The plant and the revenue are anchored in display-driver bumping, chip-on-film and test — demand that follows LCD and OLED panel volumes, not the AI/HPC supercycle. And the niche is structurally hard to earn in: it is mature and cyclical; the panel industry has migrated to China, where local packaging capacity is rising and competing on price; and the LCD base that underpins large-area driver ICs is itself in slow decline, with Omdia expecting the global display-driver-IC market to fall about 1 percent in 2025 before stabilizing in 2026. The clearest way to see what that position does to a business is to watch it land on a single company's margin.
Revenue recovered. The margin went negative — and kept falling.
Read the two lines against each other. Through 2019–2022 this was a ~9–13 percent operating-margin business — a healthy display-driver packager. Then it broke. From 2023 the margin went negative and, crucially, kept deepening — −3.0 percent, then −4.2 percent, then −8.3 percent — even as revenue recovered from the trough and reached ₩479.8 billion in 2025. That divergence is the whole diagnosis. A margin that falls while volume returns is not a volume cycle; it is pricing. In FY2025 the gross margin itself turned negative — cost of sales of ₩483.6 billion exceeded revenue of ₩479.8 billion, meaning the company sold packaging below what it cost to make — and a roughly ₩122 billion impairment of its production assets (a ₩112.6 billion property-and-equipment write-down plus intangibles and leases) produced a net loss of ₩150.9 billion. A negative gross margin and a fab write-down are the market's way of saying the assets cannot earn their cost of capital in their current use. That is a position problem, not a cost problem.
The frontier where the value went. Advanced packaging — 2.5D and 3D die stacking, silicon interposers, fan-out for AI and high-performance computing — is the fast, profitable end of OSAT, growing toward ~US$79bn by 2030 (Yole) with the 2.5D/3D platform expanding more than 20 percent a year. It is also the end of the market LB Semicon does not serve. The gap between this picture and the display-driver floor is the migration this study is about.
The commodity floor
Display-driver-IC packaging is the floor of the OSAT market, and even the scaled incumbents run thin there. The merchant specialists are Taiwanese — ChipMOS and Chipbond — and their economics tell the story. ChipMOS treats display-driver work as its lower-margin commodity line and says so plainly, managing toward a richer mix as memory recovers; its display-driver-and-gold-bump slice was a little over half of revenue and shrinking, and through the 2025 down-cycle its consolidated gross margin compressed to about 6.6 percent in the second quarter, down more than seven points year-on-year on weak utilization. Chipbond, the world's largest backend capacity for panel driver ICs, ran a ~22.5 percent full-year gross margin in 2024 — a five-year low. The point is not a single number; it is the direction. Display-driver packaging is a fixed-cost-heavy business whose margin swings violently with utilization and whose structural ceiling is set by mature panel demand. If the scaled incumbents earn thin and volatile at the floor, a sub-scale Korean pure-play below their scale earns less — which is exactly what LB Semicon's negative gross margin shows.
Even the scaled incumbents run thin at the floor
~6.6% → ~22.5%
Display-driver packaging is structurally low-margin and utilization-sensitive. ChipMOS — which calls display-driver work its commodity line versus memory — saw consolidated gross margin compress to ~6.6% in 2Q25 (down >7pts YoY) in the down-cycle; Chipbond, the largest backend capacity for panel driver ICs, ran ~22.5% full-year gross margin in 2024, a five-year low. Both are scaled merchant incumbents; a sub-scale Korean DDI pure-play earns less, as LB Semicon's negative FY2025 gross margin shows. Sources: ChipMOS Technologies (PR Newswire / SEC 6-K, 2Q25); Chipbond financial data via Investing.com / Digitimes.
The niche the migration bypassed. LB Semicon bumps, films and tests the driver chips that run LCD and OLED panels — demand that tracks the panel cycle, not the AI supercycle. With LCD maturing, the panel industry migrating to China, and the global display-driver-IC market set to slip ~1% in 2025 before stabilizing (Omdia), this is the structural floor of OSAT — and the corner where the company's revenue and capacity are anchored.
The diversification divide
The cleanest evidence for the thesis sits next door. Korea has two listed OSAT houses of comparable heritage, and they have diverged completely — not because of the cycle, which they shared, but because of product mix. Hana Micron diversified into memory-module packaging, system-in-package and advanced/AI-adjacent work, rode the memory and HBM cycle alongside SK hynix and Samsung, and grew to ₩1,251 billion in FY2024 at a positive, rising ~8.5 percent operating margin. LB Semicon stayed a display-driver pure-play and fell to a −8.3 percent margin with a fab impairment. Same country, same industry, same cycle — the variable that separated them is where in the OSAT market each was positioned. Globally the same hierarchy holds: the scaled leaders are advanced-packaging houses (ASE alone is ~45 percent of the top-ten merchant OSAT by 2024 revenue, with Amkor and JCET behind), driving the heterogeneous-integration build-out for AI; the thin-margin layer is merchant display-driver work; and a sub-scale specialist concentrated entirely in that layer sits at the most exposed point on the map.
Company
Latest-FY revenue
Op. margin
Mix / position
ASE Holdings (TW)
~US$18.5bn
profitable
Top-3 OSAT — advanced/HI leader
Hana Micron (KR)
~₩1,251bn
+8.5%
Diversified — memory / SiP / advanced
SFA Semicon (KR)
~₩400bn
low-single-digit
Memory-led packaging — cyclical
ChipMOS (TW)
~US$0.8bn
thin / volatile
DDI + memory incumbent
LB Semicon (KR)
~₩480bn
−8.3%
DDI display-driver pure-play
The diversification divide among OSAT houses, by latest-year revenue and operating margin. The variable is product mix: memory/SiP/advanced exposure earns a positive margin; a display-driver pure-play does not. Revenue converted at period FX; figures approximate and scope-sensitive. Sources: DART (Hana Micron 067310 FY2024; LB Semicon 061970 FY2025); TrendForce (ASE 2024 revenue/share); company filings; stockanalysis.com; Craft.co.
The fab the market repriced. In FY2025 LB Semicon wrote down ~₩122bn of its production assets (a ₩112.6bn property-and-equipment impairment plus intangibles and leases) — a balance-sheet admission that capacity dedicated to commodity display-driver packaging cannot earn its cost of capital in that use. The write-down, on top of an operating loss, drove a ₩150.9bn net loss and a multi-instrument recapitalization to fund a pivot toward higher-value, non-DDI backend work.
The case in point: distress and the pivot
LB Semicon is worth studying because it is the migration lived as a balance sheet. Faced with the loss and the impairment, the company executed a full multi-instrument recapitalization across 2025–26 — to be read as financing distress and a possible pivot, not as a market signal. In October 2025 it issued a ₩60 billion subordinated perpetual convertible bond (filed twice on DART — once as a convertible, once as an equity-recognized capital security, because a perpetual counts partly as equity), subscribed by a private-equity-fund syndicate led by UAMCO-KB Credit; in December 2025 it disposed of treasury shares twice for operating capital; and in May 2026 it decided a ₩49.8 billion rights issue. The founding Koo family remains the largest shareholder at ~27.2 percent, and the credit fund enters as a convertible holder — a sponsor-involved capital reset, not yet a common-equity stake. The instruments together are the toolkit of a company stabilizing a damaged balance sheet and, decisively, funding a move off the floor: the rights-issue facility spend is earmarked for non-DDI bumping and backend capacity — SoC, power management, image sensors, power semiconductor — with revenue impact guided from around 2028.
Financing the write-down — and a pivot
₩60bn + ₩49.8bn
The 2025–26 recapitalization: a ₩60bn subordinated perpetual convertible bond (Oct-2025, filed under two DART headings; subscribed by a PEF syndicate led by UAMCO-KB Credit; conversion from Oct-2026), two treasury-share disposals for operating capital (Dec-2025), and a ₩49.8bn rights issue (decided May-2026; ₩30bn facility + ₩19.8bn working capital). The facility capex targets non-DDI / advanced backend (SoC, PMIC, CIS, power semiconductor), with revenue impact guided from ~2028. Read as financing distress and a pivot, not an investment signal. Sources: DART rcept_no 20251020000284 / 20251020000285; 20251208000318 / 20251224000627; 20260515002719.
Stabilization is not a turnaround. The recapitalization answers the solvency question — the company can fund itself through the trough — but not the strategy question, which is whether the capital buys a credible move into where OSAT value has gone. That move runs into markets where ASE, Amkor and Hana Micron are already entrenched, and the group concentration cuts the wrong way: the LB Group affiliate LB Lusem (KOSDAQ 376290), a global top-three player in chip-on-film driver packaging, carries the same display-driver exposure rather than hedging it. The honest read is that LB Semicon's base case is stabilization from distress, not growth from strength — a solvency floor whose upside is entirely conditional on a pivot into a contested market, and whose downside is funding survival in a commodity niche. The market environment that decides which sorts into three regimes.
Where the market goes to 2030
Project the OSAT market forward and two things hold at once. The advanced-packaging frontier keeps compounding — AI accelerators spreading from data centers into consumer and automotive, 2.5D/3D and fan-out claiming the growth and the margin — while the display-driver floor stays mature, cyclical and price-eroded, gated by panel volumes and Chinese capacity. For a pure-play that starts on the floor, the forward path is not set by the market's average; it is set by whether it can move. The environment sorts into three regimes, defined by two variables the company does not fully control: how far and how fast a panel-cycle recovery lifts display-driver utilization, and — decisively — whether the recapitalization converts into non-DDI and advanced capacity that earns a real margin before the commodity niche erodes it further.
Regime
What would have to be true in the market
Where revenue lands by 2030
The pivot lands
Panel cycle recovers and the recapitalization funds a real move into non-DDI / SiP / advanced packaging that earns margin; capacity re-rates from distressed to growing.
~₩790bn
Slow stabilization
A gradual panel recovery lifts revenue; margins crawl toward break-even as a partial non-DDI pivot offsets DDI commoditization. Solvency restored, niche not escaped.
~₩575bn
The squeeze wins
DDI oversupply and Chinese capacity keep gross margins negative, the pivot stalls, the panel cycle stays soft; the recapitalization funds survival, not growth.
~₩450bn
Three market-environment regimes for a distressed display-driver OSAT pivoting toward advanced/non-DDI packaging, to 2030 — analytical constructs, not a company forecast, target or valuation. Each is defined by the depth of a panel-cycle recovery and whether the recapitalization funds a credible pivot; the revenue endpoints are illustrative, anchored to the company's trajectory and blended OSAT-segment CAGRs.
“A growing industry is no help to a company in the one corner the growth bypassed. In a value migration, where a packager sits — the advanced frontier or the commodity floor — decides its fate more than the cycle ever will; and the only route off the floor is to move before it erodes the assets that would fund the move.”
— Nathan Research Group, Semiconductor Packaging Series N°01
The pattern is not unique to packaging. Any industry where the headline market grows while value migrates to one end — memory tiers, foundry nodes, optical components — punishes the player concentrated at the commoditizing floor and rewards the one that moved up in time. LB Semicon is a clean case because the divergence is unusually legible: a Korean peer that diversified into memory and advanced work sits profitably above break-even, while the display-driver pure-play sits alone below it, selling below cost and writing down its own fab. Its recapitalization stabilizes the balance sheet, but a return requires the harder thing — converting capital into a position where the value actually went, against rivals who got there first. Our full brief on the OSAT and advanced-packaging market — the sizing and its scope traps, the migration mechanics, the diversification divide, the competitive map and the pivot outlook — is available to download with this article.
Working With Nathan Research Group
Partner With Nathan Research Group
Public filings and market forecasts establish the shape of the OSAT migration — the basket sizes, the advanced-packaging CAGRs, the diversification divide between the Korean OSAT houses. What they cannot show is whether LB Semicon's recapitalization actually funds a credible pivot — which non-DDI or advanced markets, with which fabless customers and what capex — how deep and how soon a panel-cycle recovery lifts display-driver utilization, or what the PEF credit holder and the founding LB Group intend for the asset. That intelligence lives with the packaging engineers, fabless sourcing teams and restructuring specialists who built, ran and competed with these lines — and reaching them, compliantly, is what we do.
Korea’s first dedicated expert network — Seoul, since 2013
Who we put in the room
OSAT packaging & test operations (DDI / bumping / COF)
Engineers who ran DDI bumping, chip-on-film and test lines — on how far a commodity display-driver packager can actually redeploy capacity toward non-DDI work, and how long a pivot really takes.
Advanced packaging & AI supply chain
Specialists in 2.5D/3D, fan-out and chiplet integration — where OSAT capex and margin are flowing, and how high the barriers are for a sub-scale entrant against ASE, Amkor and JCET.
Korean OSAT peers (Hana Micron / SFA Semicon / LB Lusem)
Operators who read the diversification divide from the inside — why memory and SiP exposure made Hana Micron profitable while a DDI pure-play and its COF affiliate were squeezed.
LB Semicon consolidated revenue and operating margin, FY2019–FY2025 (DART, rcept_no 20260320000806). Revenue recovered from the 2023 trough (₩416.9bn → ₩479.8bn) yet operating margin fell deeper each year — from a healthy ~9–13% through 2022 to −8.3% in 2025, with a negative gross margin (the company sold display-driver packaging below cost) and a ~₩122bn fab impairment that drove a −₩150.9bn net loss. Shown as evidence of the commoditization dynamic, not as a forecast.
Display-panel cycle & China OSAT capacity
Analysts who track LCD/OLED panel volumes and rising mainland-China driver-IC packaging capacity — the two forces that set DDI utilization and price at the commodity floor.
Taiwan DDI incumbents (ChipMOS / Chipbond)
Insiders on the scaled merchant DDI packagers — how thin the display-driver line runs through a down-cycle, and where mix-shift toward memory and higher-value work is actually possible.
Korean restructuring, PEF & capital markets
Restructuring and sponsor specialists who read a multi-instrument recapitalization — the perpetual convertible, the treasury disposals, the rights issue, and what a credit-fund holder's entry signals for control.
How an engagement works
1Scope
We turn your thesis into a precise expert profile and question set, mapped to the decisions you need to close.
2Source & vet
We screen and compliance-clear each expert — relevance, recency, and the absence of conflicts — before any call.
3Convene & synthesize
We arrange interviews on your timeline and, where useful, deliver written synthesis tied back to your questions.
If your team is evaluating LB Semicon, the OSAT and advanced-packaging market, or the broader semiconductor sector, tell us the decision you’re trying to make.
A direct reply from a partner, not an intake form.
Based solely on public sources. Nathan Research Group does not request or facilitate material non-public information, and runs every engagement through a documented compliance protocol.
