Haesung DS (195870) — 投资建议报告

Forward P/E

10.8x

P/B (TTM)

1.22x

P/S (TTM)

1.41x

EV/EBITDA

9.7x

Dividend Yield

1.54%

Beta

2.12

1. Executive Summary 投资概要

Investment Thesis

Haesung DS is a global #2 semiconductor lead frame manufacturer transitioning from a deep cyclical trough into a multi-engine recovery driven by automotive lead frame pricing power, DDR5 package substrate utilization normalization, and emerging AI heat spreader optionality.

Founded1984

Employees1,244

CEOYoung-Sik Choi

HQSeoul, South Korea

ISINKR7195870001

52-Week Range₩21,650 – ₩108,500

Product Portfolio

Product Line	Est. Revenue Share	Key Customers
Lead Frame (리드프레임)	~75–77%	IDMs, OSATs (Infineon, STMicro, ASE)
Package Substrate (패키지기판)	~23–28%	Samsung Electronics, SK Hynix
Tape Substrate (테이프기판)	<3%	Financial, Gov't ID, Telecom
Graphene (그래핀)	~0% (R&D)	—

Shares estimated by Kyobo Securities / SK Securities [Research]

Three-axis re-rating story (Korea Invest Insights):
Axis 1: Automotive lead-frame recovery with pricing power
Axis 2: DDR5 substrate utilization 30% → 70% in 2026E
Axis 3: AI heat-spreader second-source entry (emerging)

2. Investment Thesis 核心论点

Thesis 1: Lead Frame — Global #2 with Structural Pricing Power Recovery

强 (Strong)

Market position: Global #2 behind Mitsui High-Tec in semiconductor lead frames. Automotive-qualified products create a high barrier to substitution (AEC-Q100, <5ppm defect rates, long-cycle qualification).
Pricing power: Industry-wide lead frame price increases of 10–30% in 2025 (SK Securities). Raw material pass-through (Cu, Au, Ag, Pd) with ~1Q lag protects margins.
Volume recovery: 80% export ratio, serving 80+ semiconductor companies globally. Automotive demand recovery driving lead frame volume growth. Q4 2025 OP +248% YoY confirms the inflection.
Key monitoring: 2Q26 OPM as "verification gate" for raw material pass-through normalization.

Risk: Commoditization pressure on standard lead frames; raw material pass-through lag impacted 1Q26 margins (OPM 5.8% vs 12.1% in 4Q25).

Thesis 2: Package Substrate — The High-Leverage Turnaround

强 (Strong)

Swing factor: DDR5 utilization ~30% in H1 2025 → ~70% in 2026E (KB Securities). Substrate is a high operating leverage business — incremental volume flows disproportionately to profit.
Structural tailwind: High-end substrate makers shift capacity to AI FC-BGA, leaving mid/low-end memory substrate demand underserved. Haesung DS's R2R cost advantage on DDR4/DDR5 is a key differentiator.
Consensus estimate: FY2026E revenue ~₩840B (+28.6% YoY), OP ~105.7B (+127% YoY).

Risk: Extreme customer concentration (Samsung+SK Hynix >80% of substrate revenue). No HBM substrate capability — missing the fastest-growing AI memory segment.

Thesis 3: AI Heat Spreader — Asymmetric Optionality

中 (Medium)

Emerging business: IC heat spreader for AI accelerators. Samples qualifying in 2Q26 with 2H26 revenue potential.
TAM: ~US$567M (2024), 9.7% CAGR, ~85% market concentration (Shinko, Honeywell, Jentech, I-Chiun).
Revenue: Base case ₩5–20B in 2026E, ₩20–50B in 2027E. Small for P&L but significant for valuation re-rating (from "auto cyclical" to "AI thermal supply chain").

Risk: Qualification samples still in progress. Commercialization track record in new products is unproven.

3. Financial Trajectory 财务轨迹

Consolidated P&L (KRW Million, K-IFRS)

Line Item	FY2021	FY2022	FY2023	FY2024	FY2025	FY2026E
매출액 (Revenue)	655,373	839,358	672,245	602,994	653,397	840,000
YoY Growth %	—	+28.1%	−19.9%	−10.3%	+8.4%	+28.6%
매출원가 (COGS)	519,718	574,126	511,524	491,557	555,593	—
매출총이익 (Gross Profit)	135,655	265,232	160,722	111,437	97,803	—
Gross Margin %	20.7%	31.6%	23.9%	18.5%	15.0%	~18.5%
판매비와관리비 (SG&A)	49,325	60,816	58,210	54,567	51,312	—
영업이익 (Operating Profit)	86,330	204,416	102,512	56,870	46,491	105,700
OP Margin %	13.2%	24.4%	15.3%	9.4%	7.1%	~12.6%
영업외손익 (Non-op, Net)	3,190	894	3,350	10,082	−13,908	—
법인세차감전순이익 (Pre-tax Profit)	89,520	205,310	105,862	66,952	32,583	—
법인세비용 (Income Tax)	18,469	45,910	21,436	8,255	8,763	—
Effective Tax Rate %	20.6%	22.4%	20.2%	12.3%	26.9%	—
당기순이익 (Net Profit)	71,052	159,400	84,426	58,697	23,821	~85,900
지배주주순이익 (attr. to parent)	71,262	159,400	84,426	58,697	23,821	~85,900
기본주당이익 (Basic EPS, ₩)	₩4,192	₩9,376	₩4,966	₩3,453	₩1,401	~₩4,527

FY2021–2025: OpenDART API structured data. FY2026E: Consensus estimates.

Revenue & OP Trend (KRW Billion)

4. Business Deep Dive 业务深度分析

Segment 1: Lead Frame (리드프레임) — ~75–77% of Revenue

Global #2 semiconductor lead frame manufacturer behind Mitsui High-Tec.

D1 — Business Essence 业务本质

Lead frames are the "metal skeleton" inside a semiconductor package: supporting the die, routing electrical signals to the PCB, carrying current, and dissipating heat. Haesung DS produces lead frames for automotive, mobile, PC, industrial, power management, and consumer electronics applications. Customers include IDMs (Infineon, STMicro) and OSATs (ASE, Amkor). [Research]

D2 — Market Position & Moat 市场地位

Global rank: #2 worldwide (behind Mitsui High-Tec, Japan). FY2025 revenue ~₩493B (~US$340M) [Research]
Moat: Automotive lead frames require long-cycle AEC-Q100 qualification (traceability, <5ppm defect rates) — hard to substitute on price alone [Research]
80% export ratio across 80+ semiconductor companies; history traceable to Samsung Techwin's MDS division (spun off 2014) [Research]
Global competitors (Mitsui, Shinko, ASM Pacific) raising prices 10–30% in 2025 — Haesung DS enjoys pricing power for first time since pandemic [SK Securities]

D3 — Profit Driver Decomposition 利润驱动拆解

Volume: Global automotive semiconductor demand (ICE + EV). Recovery driven by EV market demand and customer inventory stabilization [Research]
Price: Raw material pass-through (Cu, Ag, Au, Pd) with ~1Q lag. Industry price increases 10–30% give pricing power [SK Securities]
Mix: Higher-value automotive lead frames carry better margins vs consumer/PC. EV transition drives higher pin-count, finer-pitch demand [Research]
OPM: Lead Frame historically has higher margins than Package Substrate, though specific OPM undisclosed

D4 — Demand Structure 需求结构

End Market	Exposure	Trend
Automotive	Major (~50%+)	Recovery; EV transition drives higher-value frames
Mobile	Moderate	Mature, cyclical
PC / Consumer	Moderate	Mature, inventory-driven
Industrial / Power	Small growing	Steady

Customer inventory normalized since mid-2025; safety stock orders expected to expand in 2026 [Kyobo Securities]

D5 — Capacity & Mid-Term Plan 产能与战略

Investing in AI-driven autonomous manufacturing (AX partnership with INTERX for quality inspection AI) [Company newsroom 2026]
R&D for heat-spreader / heat-slug technology leveraging existing metal processing know-how [SK Securities]
No large M&A; organic CapEx focused on substrate capacity expansion
The Changwon plant expansion (₩388B total) was primarily for substrate capacity, but the timing aligned poorly with the semiconductor downcycle

D6 — Competition & Risk 竞争与风险

Competitor	Country	Position
Mitsui High-Tec	Japan	#1 global leader
Haesung DS	Korea	#2 global
Shinko Electric	Japan	Major player
ASM Pacific Tech	Hong Kong	Major player
I-Chiun	Taiwan	Niche / high precision

Risks: Commoditization; raw material cost volatility (Cu/Au/Ag/Pd inflated 1Q26 OPM to 5.8% vs 12.1% in 4Q25); auto cycle; long-term substitution by advanced packaging [Research]

D7 — Governance & Capital Allocation 治理与资本配置

Dividend: ₩900/share (~1.54% yield), stable payout. No buyback history.
Debt/Equity: ~41% (from 388B capex financing). Interest expense doubled to ₩6.7B in FY2025.
ROE: ~6.2% (TTM) — depressed by the earnings trough, consensus expects recovery to ~15.5% in 3 years.
No activist investor pressure. Management has not signaled major governance changes.
Korean Value-Up Program (밸류업) could incentivize improved shareholder returns — monitor for potential catalyst.

D7 — Link to Consolidated Financials

[Model] Revenue +28.1% FY2022 (super-cycle peak) → −19.9% FY2023 → −10.3% FY2024 → +8.4% FY2025. Lead frame volume recovery is a key FY2025 rebound driver. Gross margin compressed from 31.6% to 15.0% partly due to unfavorable product mix shift. OPM fell from 24.4% to 7.1% largely driven by lead frame price competition and volume decline.

Segment 2: Package Substrate (패키지기판) — ~23–28% of Revenue

Mid/low-end memory package substrates for DDR4/DDR5 using Reel-to-Reel (R2R) process.

D1 — Business Essence 业务本质

Package substrates are small circuit boards between the memory die (DRAM) and motherboard. Haesung DS focuses on memory package substrates (DDR4/DDR5) using a Reel-to-Reel (R2R) continuous lamination process — different from the panel-by-panel approach used by high-end substrate makers. Key customers: Samsung Electronics, SK Hynix. [Research]

D2 — Market Position & Moat 市场地位

Niche: Mid/low-end memory substrates ONLY — NOT high-layer-count AI FC-BGA (Samsung Electro-Mechanics, Daeduck, Ibiden, AT&S territory) [Research]
R2R advantage: Cost-and-yield advantages on high-volume DDR4/DDR5 vs panel-based competitors [Korea Invest Insights]
Structural tailwind: High-end makers shifting to AI FC-BGA leave mid/low-end demand underserved — Haesung DS benefits from this capacity shift [SK Securities]
New customer wins in FY2026: "Domestic A (D1b Prime), B (D1b), China C (D1y), US D (D5)" — diversifying beyond Samsung/Hynix [SK Securities]

D3 — Profit Driver Decomposition 利润驱动拆解

#1 driver: Utilization rate. Substrate is a high operating leverage business. 3Q25 still slight loss, 4Q25 turned profitable as DDR5 demand accelerated [SK Securities]
DDR5 transition: Higher ASP and margin than DDR4. Utilization ~30% in 1H25 → ~70% in 2026E [KB Securities]
Operating leverage: Once fixed costs covered, incremental revenue flows disproportionately to OP. Normalized OPM ~10–15% [Research]

D4 — Demand Structure 需求结构

End Market	Exposure	Trend
Memory (DRAM)	>90%	DDR5 volume ramp in 2026
AI / Server	Indirect (via memory)	Growing as AI servers consume more DRAM
PC / Mobile	Indirect	Mature

Direct linkage to memory cycle: Haesung DS's substrate business is a high-beta proxy for the Korean memory industry. FY2023–2024 showed how painful the downcycle can be. [Model]

D5 — Capacity & Mid-Term Plan 产能与战略

No disclosed large-scale substrate CapEx expansion — growth comes from utilization improvement, not capacity addition
R&D and CapEx directed toward new customer qualifications (Chinese, US memory makers) rather than technology leapfrogging
AI heat-spreader initiative suggests management looking for higher-value opportunities beyond memory substrates [Research]
The ₩388B Changwon plant (completed FY2023–2025) was primarily for substrate capacity — the depreciation drag is now mostly recognized

D6 — Competition & Risk 竞争与风险

Competitor	Country	Position
Samsung Electro-Mech.	Korea	High-end FC-BGA (shifting away from mid/low)
Daeduck Electronics	Korea	Mid-range PCB/substrate
LG Innotek	Korea	High-end substrate
Ibiden	Japan	Premium FC-BGA
AT&S	Austria	Premium IC substrates

Risks: Extreme customer concentration (Samsung+SK Hynix >80%); memory cycle dependency; technology tier risk; NO HBM exposure — missing the fastest-growing AI memory segment [Research]

D7 — Governance & Capital Allocation 治理与资本配置

No separately disclosed substrate CapEx — growth is utilization-driven.
New customer wins (China C, US D) suggest a deliberate diversification strategy away from Samsung/Hynix concentration.
The AI heat-spreader initiative signals management is seeking higher-value-add opportunities beyond memory substrates.
Capacity expansion timing (₩388B during the downcycle) raises questions about capital allocation discipline.

D7 — Link to Consolidated Financials

[Model] 31.6% → 15.0% GP margin compression is significantly explained by substrate margin deterioration (DDR4 overhang, low utilization). FY2025 recovery (+8.4% revenue) partly driven by substrate stabilization. Consensus FY2026E revenue ~₩840B with OPM ~12.6% driven primarily by substrate turnaround.

Revenue Composition (FY2025 Estimate)

Segment Risk & Operating Leverage Profile

Metric	Lead Frame	Package Substrate
Revenue Share	~75–77%	~23–28%
Operating Leverage	Medium	High
Pricing Power	Improving (10–30% hikes)	Limited (Tier-2 supplier)
Cyclicality	Medium (auto anchors)	High (memory-linked)
Customer Concentration	Diversified (80+ cos)	Extreme (2 customers)
FY2026 Recovery Driver	Volume + price	Utilization 30% → 70%

Key insight: Package Substrate has HIGHER operating leverage but also HIGHER cyclicality. When utilization normalizes, it drives disproportionate profit recovery. The FY2022→FY2025 margin collapse was primarily from substrate mix deterioration (~40%→23% of revenue) and underutilization.

FY2022 → FY2025 Operating Profit Bridge (KRW Billion)

5. Segment Structure 业务结构

Profit Bridge Attribution (FY2022→FY2025)

Profit Bridge Item	Impact	Driver
Lead Frame volume decline	Major	Global semiconductor downcycle
Lead Frame margin compression	Moderate	Price competition, raw material cost
Package Substrate utilization collapse	Major	DDR4→DDR5 transition, memory CapEx cuts
Package Substrate margin loss	Major	Low utilization → negative operating leverage
FX losses (non-op)	Notable	₩13.9B non-op loss in FY2025

Value Chain Positioning

Raw Materials (Cu, Au, Ag, Pd)
        ↓
   Haesung DS Manufacturing
   ├── Lead Frame (stamping/etching)→ IDM/OSAT
   ├── Package Substrate (R2R)     → Samsung/SK Hynix
   ├── Tape Substrate              → Smart Card Makers
   └── Graphene (CVD)              → R&D only

Key Differentiators:

Dimension	Advantage	Disadvantage
Lead Frame	#2 global scale, auto qual	Not #1 (Mitsui)
Substrate	R2R cost advantage	No HBM / FC-BGA
Customers	Deep ties to Korean giants	Over-dependence
New tech	AI heat spreader	Graphene unproven

Emerging Business: AI Heat Spreader / Heat Slug

Dimension	Detail
Product	IC package heat spreader for AI accelerators — leverages lead frame metal processing know-how
TAM	~US$567M (2024), 9.7% CAGR → ~US$682M (2026). ~85% concentrated (Shinko, Honeywell, Jentech, I-Chiun)
Haesung DS role	Second-source entrant; samples qualifying 2Q26, potential 2H26 revenue
Revenue potential	Base: ₩5–20B in 2026E, ₩20–50B in 2027E
Re-rating mechanism	Multiple expansion from "auto cyclical" → "AI thermal supply chain" classification

6. Anomaly Analysis 异常分析

Anomaly #1: Revenue Volatility — Massive Swings High Severity

Observation: Revenue swung from ₩655B (FY2021) → ₩839B (+28.1%, FY2022) → ₩672B (−19.9%, FY2023) → ₩603B (−10.3%, FY2024) → ₩653B (+8.4%, FY2025).

Phase	Revenue Δ	Key Driver
FY2022 Surge	+28.1%	Memory super-cycle peak. Samsung/SK Hynix ramping DDR4 at peak pricing. Auto lead frames grew on post-COVID demand.
FY2023 Collapse	−19.9%	Semiconductor downcycle. Samsung DS record losses (~10-15TN KRW). IT lead frames −20% YoY. Auto LF +11% (bright spot).
FY2024 Decline	−10.3%	DDR4→DDR5 order vacuum. Slow DDR5 qualification. Package substrate revenue −44% from FY2022 peak (to ₩168.6B).
FY2025 Recovery	+8.4%	DDR5 ramp gaining. Mix shifted: LF 77% vs Substrate 23% (vs historical 60/40). Q4 rev ₩179.9B (+23.9% YoY).

Root Cause: Direct function of the semiconductor memory cycle (DDR4 peak → crash → DDR5 ramp). The structural shift in revenue composition (substrates falling from ~40% to ~23%) mechanically amplified the cycle's impact on the P&L.

Anomaly #2: Gross Margin Collapse — 31.6% → 15.0% Critical Severity

Observation: Gross margin nearly halved from 31.6% (FY2022) to 15.0% (FY2025) in just 3 years.

Factor	Impact	Detail
Product Mix Deterioration	🔴 High	High-margin substrates (est. 30-35% GP) fell from ~40%→~23% of revenue. Low-margin LF (15-20% GP) rose to 77%.
Underutilization	🔴 High	Substrate utilization collapsed during FY2023-2024. Fixed costs (₩388B Changwon) spread over lower revenue.
DDR4 Price Erosion	🟡 Medium	DDR4 substrate ASPs declined as end-of-life product. No DDR5 qualification to offset.
COGS Inflexibility	🟡 Medium	COGS −10.9% (FY22→FY24) vs Rev −28.2%. COGS +13% in FY2025 despite Rev +8.4% (Changwon depreciation).

Root Cause: Structural mix shift (substrates 40%→23% of revenue) mechanically compressed GP margins. Underutilization of new Changwon capacity added fixed-cost pressure. As a Tier-2 supplier, Haesung DS has limited ability to pass through costs during downturns.

Anomaly #3: Operating Profit Plunge — ₩204.4B → ₩46.5B (−77.3%) Critical Severity

Observation: OP collapsed from ₩204.4B (FY2022, 24.4% margin) to ₩46.5B (FY2025, 7.1% margin).

Year	Revenue Δ	OP Δ	Incremental Margin
FY2023	−₩167.1B	−₩101.9B	−61% (negative operating leverage)
FY2024	−₩69.3B	−₩45.6B	−66%
FY2025	+₩50.4B	−₩10.4B	Revenue grew but OP still fell

FY2025 Paradox: Revenue +8.4% but OP −18.3% due to: (1) GP margin continued contracting (18.5%→15.0%), (2) Depreciation accelerated (₩39.0B→₩57.7B, +48%), (3) R&D elevated at ₩18.5B.

Recovery Signal: Q4 2025 (prelim.) Rev ₩179.9B (+23.9% YoY), OP ₩21.8B (+247.8% YoY, OPM 12.1%) — suggests H1 2025 was the trough.

Root Cause: High fixed costs (depreciation ₩39B→₩58B, R&D ₩18.5B, SG&A ₩51-61B) created negative operating leverage. When revenue falls, most of the decline flows through to OP. The new Changwon plant's depreciation hit during the worst part of the cycle.

Anomaly #4: EPS Collapse — ₩9,376 → ₩1,401 (−85.1%) Critical Severity

Observation: Basic EPS dropped from ₩9,376 (FY2022) to ₩1,401 (FY2025).

Component	FY2022	FY2025	Change
Net Profit (KRW M)	159,400	23,821	−85.1%
Shares Outstanding (M)	17.00	17.00	~flat (no dilution)
Basic EPS (KRW)	₩9,376	₩1,401	−85.1%

NP decline (−85%) worse than OP decline (−77%) due to: (1) Non-Operating Loss −₩13,908M in FY2025 vs +₩10,082M in FY2024 = −₩24.0B swing; (2) Effective tax rate jumped from 12.3% to 26.9%.

Valuation context: Trailing PER ~6.8x at trough. Forward PER ~10.8x. Consensus EPS recovery: ~₩2,263 (FY2026E) → ~₩5,543 (FY2027E).

Root Cause: Purely from net profit decline — no share dilution. The non-operating loss (−₩13.9B) and higher tax rate (26.9% vs 12.3%) amplified the EPS decline beyond the OP decline.

Anomaly #5: FY2025 Non-Operating Loss −₩13,908M High Severity

Observation: Non-op swung from +₩10,082M (FY2024) to −₩13,908M (FY2025), a −₩24.0B deterioration.

Line Item	FY2024 (₩M)	FY2025 (₩M)	Change	Driver
Interest Expense	−3,178	−6,736	−3,558	Debt for Changwon capex (₩388B)
Interest Income	+2,875	+1,679	−1,196	Lower cash / rates
Net Interest	−304	−5,057	−4,754	Debt doubled
FX Gains/(Losses)	+11,690	−1,180	−12,870	KRW appreciation in 2025
Asset Writedown	−1,560	−7,390	−5,830	DDR4 equipment impairment
Other Non-Op	+237	−209	−446	—
Total Non-Op	+10,082	−13,908	−23,990

Root Cause — Three factors:
(1) FX swing −₩12.9B (54% of deterioration): Export revenues in USD with limited hedging. FY2024 weak KRW → large FX gains. FY2025 KRW strength → FX losses.
(2) Interest doubling −₩4.8B: ₩388B capex financed through debt. Interest expense doubled from ₩3.2B to ₩6.7B.
(3) Asset writedowns −₩5.8B: ₩7.4B impairment on DDR4 equipment (underutilized due to delayed DDR5 transition).
This is the financial manifestation of the capex cycle timing misalignment with the industry cycle.

7. Profit Quality 利润质量

Key Quality Metrics

Metric	FY2022	FY2025	Assessment
Gross Margin	31.6%	15.0%	🔴 Deteriorating
OP / Pre-tax Ratio	99.6%	142.7%	🟡 Non-op noise
Effective Tax Rate	22.4%	26.9%	🟡 Normalizing
ROE	~30%	~6.2%	🔴 Cycle-depressed
Payout Ratio	~6%	~64%	🟡 High payout on trough
FCF	Positive	−₩18.1B TTM	🔴 CapEx-heavy
Net Cash Position	Positive	Debt/Equity 41%	🟡 Leverage increased

Quality Summary

Haesung DS's profit quality is structurally tied to the semiconductor cycle. Key observations:

GP margin compression is structurally driven by product mix — substrates now only 23% of revenue vs 40% historically. This is partly permanent (capacity shift to higher-end substrates by competitors).
Non-operating noise is high due to FX exposure. FY2024 had unsustainable +₩11.7B FX gains; FY2025 had −₩1.2B losses. This creates EPS volatility that doesn't reflect core business quality.
FCF negative during the trough reflects the CapEx-heavy business model. The ₩388B Changwon investment strained the balance sheet (debt/equity 41%).
Dividend coverage is stretched — ₩900/share on ₩1,401 EPS implies 64% payout ratio. In FY2022, the payout was only ~6%.
Recovery catalysts: Consensus expects ROE to recover to ~15.5% within 3 years. The key is whether FY2026E OPM of 12.6% is achievable — this requires both substrate utilization recovery AND lead frame margin stabilization.

8. Valuation 估值分析

Peer Comparison

Company	Fwd P/E	P/B	EV/EBITDA
Haesung DS (195870)	10.8x	1.22x	9.7x
Mitsui High-Tec	~14x	~1.8x	~8x
Shinko Electric	~18x	~2.1x	~10x
ASM Pacific Tech	~22x	~2.5x	~15x
Samsung Electro-Mech.	~15x	~1.1x	~8x
KOSPI Semi Peer Avg.	~16x	~1.5x	~10x

Haesung DS trades at a ~33% discount to peer average forward P/E.

Scenario Analysis (FY2026E EPS ~₩4,527)

保守 Conservative

₩45,000

+12%

P/E 10x

基准 Base Case

₩67,900

+69%

P/E 15x

目标 Target

₩90,500

+125%

P/E 20x

Current: ₩40,150. Downside support at 10x P/E. Analyst target: ₩90,650 (+125.8% upside).

9. Catalysts 股价催化剂

2026年8月18日

2Q26 Earnings Release — The "verification gate." After 1Q26 raw material pass-through lag compressed margins, 2Q26 will show if pricing normalization is on track. Consensus expects OPM recovery toward ~10%.

2026年7–9月

AI Heat Spreader Sample Qualification — Qualifying heat-slug samples with a major AI accelerator customer. Successful qualification would be a significant re-rating catalyst, potentially shifting classification from "auto cyclical" to "AI thermal."

2026年下半年

DDR5 Utilization Ramp to ~70% — Each 10pp utilization improvement adds ~₩12–15B to annual OP. Consensus FY2026E OP of ₩105.7B implies substantial substrate recovery.

2026年

Korea Value-Up Program — Government initiative for better shareholder returns. Haesung DS pays ₩900/share (1.54% yield). Any buyback or dividend increase is a positive signal.

2026–2027年

Sustained Lead Frame Price Increases — Industry-wide 10–30% price hikes (SK Securities). If sustained, this structurally improves LF margins representing ~75% of revenue.

10. Risks 风险因素

Risk Factor	Severity	Description
Semiconductor Cycle	🔴 High	Revenue and profits tightly correlate with memory cycle. Any renewed slowdown in DDR5/HBM demand delays recovery. FY2023–2024 demonstrated the downcycle severity.
Customer Concentration	🔴 High	Substrate business depends >80% on Samsung Electronics and SK Hynix. Loss of any key customer would be materially damaging.
No HBM Capability	🟡 Med-High	Only DDR5 substrates, not HBM — the fastest-growing AI memory segment. This limits participation in the AI memory growth story.
Debt & Interest	🟡 Medium	₩388B capex doubled interest expense to ₩6.7B. D/E ~41%. Slower recovery worsens debt service burden.
FX Volatility	🟡 Medium	USD-denominated export revenues with limited hedging. FY2025 non-op loss −₩13.9B primarily from KRW appreciation.
Tech Substitution	🟡 Medium	Advanced packaging reduces lead frame content in high-end chips. Auto ICs remain LF-heavy but structural decline risk exists.
Korea Discount	🟢 Low-Med	Korean semi component stocks trade at persistent discount to global peers. At 10.8x forward P/E, discount may limit multiple expansion.

11. Conclusion 投资结论

Rating: 買入 (BUY) · Target: ₩90,650 (+125.8%) · 12-Month Horizon

Haesung DS offers a compelling cyclical recovery play with asymmetric AI optionality. The company has passed the trough of the deepest semiconductor downcycle in a decade, and is now entering a multi-engine recovery driven by lead frame pricing power (#2 globally, 10-30% price hikes), DDR5 substrate utilization normalization (30%→70%), and emerging AI heat spreader entry. At 10.8x forward P/E — a ~33% discount to peer average — the market prices in recovery but not the full substrate turnaround or AI re-rating potential.

Global #2 lead frame position with structural auto-qualification moat
Industry-wide price increases 10-30% — pricing power returning
DDR5 utilization 30%→70% creates powerful operating leverage in 2026E
Q4 2025 OP +248% YoY confirms cyclical inflection
AI heat spreader samples in qualification — asymmetric re-rating catalyst
Consensus FY2026E OPM 12.6% from 7.1% trough — substantial recovery
Valuation at 10.8x fwd P/E — 33% discount to peer average
Extreme customer concentration in substrate (Samsung+SK Hynix >80%)
No HBM substrate capability — missing fastest-growing AI memory
Debt-funded ₩388B capex — interest expense doubled; adds earnings risk
FX volatility: −₩12.9B swing in non-op income in FY2025
Korea Discount may limit multiple expansion
Graphene R&D (~₩18.5B/yr) produces no revenue — cash drain

해성디에스 (Haesung DS)