Samsung Electronics
Combined average of Moat (AI Resilience), Growth, and Valuation scores.
Moat Score
The world's largest memory chipmaker and a top-three smartphone OEM, but Samsung's moat is diluted by conglomerate breadth: it trailed SK Hynix and Micron in HBM4 qualification timing, trails TSMC by 1-2 process generations in leading-edge foundry, and its consumer ecosystem lacks the switching-cost intensity of Apple's. The DRAM/NAND oligopoly floor and deep process IP remain genuine advantages, but no single segment is best-in-class the way focused peers are.
Samsung's competitive position rests on Memory Oligopoly Scale, Diversified Reach, and Deep Process IP — but breadth cuts against focus:
- Memory Oligopoly, But Not the Pace-Setter: Samsung, SK Hynix, and Micron control ~95% of global DRAM supply, and Samsung remains the largest single producer by volume. But Samsung trailed SK Hynix and Micron in HBM3E/HBM4 qualification with NVIDIA through 2024-2025, only securing NVIDIA HBM4 qualification in Q1 2026 — a costly delay that ceded share of the highest-margin memory segment to faster-moving rivals.
- Foundry: Perpetual Number Two (or Three): Samsung Foundry's 2nm GAA (SF2) process is ramping with Tesla's AI5/AI6 chips as an anchor customer (a ~$16.5B multi-year deal) and a new IBM partnership, but yields still lag TSMC's N2 by a wide margin and the division has posted cumulative losses for years. Foundry diversifies Samsung's semiconductor exposure but is not yet a moat in its own right.
- Consumer Ecosystem Without Apple's Lock-In: Galaxy phones, Watches, Buds, SmartThings, and TVs form a real bundle — SmartThings connects 300M+ devices — but Android's openness means switching to a rival OEM costs little. Galaxy AI (built on Google Gemini) adds feature differentiation but not durable lock-in the way iOS does for Apple.
Moat Verdict
Samsung is a genuine but lagging beneficiary of the AI era: the memory supercycle (DRAM/NAND/HBM4) and Foundry's Tesla AI-chip relationship are direct AI tailwinds, and proprietaryData, regulatoryLockIn, and transactionEmbedding are all intact or strengthening as multi-year HBM contracts and government backing deepen. But Samsung trails SK Hynix and Micron in HBM4 execution and TSMC in foundry yield, and its AI-vulnerable consumer moats (learnedInterfaces, bundling, networkEffects) remain weakened by Android's openness — making Samsung a slower, more diluted AI beneficiary than its focused semiconductor peers.
58.5 resilient · 41.5 vulnerable · 80/20 = 55.1 · = 55
Open a moat to read its note.
One UI and the Galaxy ecosystem create habitual behavior, but Android's openness means switching to a rival OEM (Google, OnePlus, Xiaomi) costs little — no meaningful data or workflow lock-in the way iOS creates for Apple.
Samsung is fundamentally a hardware manufacturer; no proprietary business logic moat comparable to a software platform.
Samsung does not derive competitive advantage from public data access.
Leading-edge DRAM, NAND, and GAA foundry process engineers are among the scarcest technical talent globally, and Samsung's decades-deep Korean R&D bench (Hwaseong, Pyeongtaek) is a genuine advantage — though it competes directly with SK Hynix and TSMC for the same talent pool.
Galaxy phones, Watch, Buds, SmartThings, and TVs form a real cross-device bundle (SmartThings connects 300M+ devices), but attach and lock-in are materially weaker than Apple's ecosystem — most components can be freely mixed with other brands.
Decades of proprietary DRAM/NAND cell-design IP, HBM4 base-die architecture, and GAA transistor process data represent genuine trade secrets that competitors cannot easily replicate.
South Korean government backing (K-chips Act tax credits) plus US CHIPS Act funding for the Taylor, TX fab, reinforced by the Tesla AI5/AI6 foundry agreement tied to US-based advanced manufacturing, gives Samsung durable government-linked positioning in both Korea and the US.
SmartThings connects 300M+ devices, but this reflects installed-base scale rather than true Metcalfe's Law dynamics — Google Home and Apple HomeKit are close substitutes, and the switching cost is modest.
The industry-wide shift to 3-5 year HBM supply contracts now extends to Samsung's NVIDIA-qualified HBM4 volume, and Samsung Foundry's design-ins with Tesla (AI5/AI6) and IBM (2nm) create multi-year customer roadmap embedding similar to TSMC's and Micron's.
Samsung is a component and device supplier, not a system of record for any critical business function; customers dual- and triple-source memory and foundry capacity across Samsung, SK Hynix, Micron, and TSMC.
The largest memory manufacturer by capacity, which lowers commodity DRAM and NAND cost, but trailing SK hynix in HBM limits the advantage where margins are highest.
Galaxy is a global top-two smartphone brand, but it does not hold price the way Apple's does.
Combined average of Moat (AI Resilience), Growth, and Valuation scores.
Moat Score
The world's largest memory chipmaker and a top-three smartphone OEM, but Samsung's moat is diluted by conglomerate breadth: it trailed SK Hynix and Micron in HBM4 qualification timing, trails TSMC by 1-2 process generations in leading-edge foundry, and its consumer ecosystem lacks the switching-cost intensity of Apple's. The DRAM/NAND oligopoly floor and deep process IP remain genuine advantages, but no single segment is best-in-class the way focused peers are.
Growth Score
Samsung's Device Solutions (DS) division is in the steepest part of the AI-driven memory supercycle: Q2 2026 consolidated revenue was a record ₩171.5T (+130% YoY, +28% QoQ) after ₩133.9T in Q1, with operating profit of ₩89.5T. DS contributed ₩127.5T of revenue (from ₩81.7T in Q1) and ₩89.2T of operating profit — nearly the whole company — as DRAM, NAND and HBM prices rose on AI-driven supply tightness. HBM4 sales are guided to more than triple in Q3 and to exceed 60% of HBM revenue in the second half, and Samsung has shipped the first HBM4E samples. After trailing SK Hynix and Micron through 2024-2025, Samsung secured NVIDIA HBM4 qualification in Q1 2026, restoring its position in the highest-margin memory segment. The Device eXperience (DX) division's sales fell 9% QoQ (up YoY), and MX/Networks swung to a ₩0.7T operating loss as memory costs squeezed phones. Samsung Foundry's SF2 (2nm GAA) is in early high-volume production with Tesla's AI5/AI6 chips as the anchor customer under a ~$16.5B multi-year agreement, plus a new IBM 2nm partnership — narrowing (but not closing) years of foundry losses. Mobile (Galaxy S26, foldables) and Consumer Electronics remain low-single-digit growth, mature businesses that dilute the consolidated growth rate versus pure-play memory peers. Since the prior update, Samsung was named alongside SK Hynix and Micron in a June 25, 2026 US class-action lawsuit (N.D. Cal.) alleging DRAM price-fixing since 2022 via a coordinated HBM-driven cutback of DDR3/DDR4 supply — an early-stage suit (no class certified yet) but a new legal overhang shared across the memory oligopoly.
Valuation Score
At ~₩285,500 (September 24, 2026) — nearly 3× the July file's ₩98,000, after Q2 2026 printed a record ₩89.5T operating profit (52% margin) and ₩10,849 of quarterly EPS — Samsung sits between the reset ₩200,000 bear and ₩325,000 base, about 68% of the way from bear to base, yielding a valuation score of 73. The optics are cheap (~5.3× forward earnings) but this is a memory-cycle multiple on what are very likely peak-cycle earnings: the market is discounting the 2027–28 supply response (CXMT's planned capacity adds, SK Hynix and Micron expansions) rather than paying for the current quarter. The ladder is therefore built on P/B and a deliberately low multiple of consensus NTM EPS, not on the brokers' ₩475,000+ targets, which assume the cycle extends through 2028.
The Conglomerate's Dilemma
Samsung's competitive position rests on Memory Oligopoly Scale, Diversified Reach, and Deep Process IP — but breadth cuts against focus:
- Memory Oligopoly, But Not the Pace-Setter: Samsung, SK Hynix, and Micron control ~95% of global DRAM supply, and Samsung remains the largest single producer by volume. But Samsung trailed SK Hynix and Micron in HBM3E/HBM4 qualification with NVIDIA through 2024-2025, only securing NVIDIA HBM4 qualification in Q1 2026 — a costly delay that ceded share of the highest-margin memory segment to faster-moving rivals.
- Foundry: Perpetual Number Two (or Three): Samsung Foundry's 2nm GAA (SF2) process is ramping with Tesla's AI5/AI6 chips as an anchor customer (a ~$16.5B multi-year deal) and a new IBM partnership, but yields still lag TSMC's N2 by a wide margin and the division has posted cumulative losses for years. Foundry diversifies Samsung's semiconductor exposure but is not yet a moat in its own right.
- Consumer Ecosystem Without Apple's Lock-In: Galaxy phones, Watches, Buds, SmartThings, and TVs form a real bundle — SmartThings connects 300M+ devices — but Android's openness means switching to a rival OEM costs little. Galaxy AI (built on Google Gemini) adds feature differentiation but not durable lock-in the way iOS does for Apple.
Moat Verdict
Samsung is a genuine but lagging beneficiary of the AI era: the memory supercycle (DRAM/NAND/HBM4) and Foundry's Tesla AI-chip relationship are direct AI tailwinds, and proprietaryData, regulatoryLockIn, and transactionEmbedding are all intact or strengthening as multi-year HBM contracts and government backing deepen. But Samsung trails SK Hynix and Micron in HBM4 execution and TSMC in foundry yield, and its AI-vulnerable consumer moats (learnedInterfaces, bundling, networkEffects) remain weakened by Android's openness — making Samsung a slower, more diluted AI beneficiary than its focused semiconductor peers.
58.5 resilient · 41.5 vulnerable · 80/20 = 55.1 · = 55
Open a moat to read its note.
One UI and the Galaxy ecosystem create habitual behavior, but Android's openness means switching to a rival OEM (Google, OnePlus, Xiaomi) costs little — no meaningful data or workflow lock-in the way iOS creates for Apple.
Samsung is fundamentally a hardware manufacturer; no proprietary business logic moat comparable to a software platform.
Samsung does not derive competitive advantage from public data access.
Leading-edge DRAM, NAND, and GAA foundry process engineers are among the scarcest technical talent globally, and Samsung's decades-deep Korean R&D bench (Hwaseong, Pyeongtaek) is a genuine advantage — though it competes directly with SK Hynix and TSMC for the same talent pool.
Galaxy phones, Watch, Buds, SmartThings, and TVs form a real cross-device bundle (SmartThings connects 300M+ devices), but attach and lock-in are materially weaker than Apple's ecosystem — most components can be freely mixed with other brands.
Decades of proprietary DRAM/NAND cell-design IP, HBM4 base-die architecture, and GAA transistor process data represent genuine trade secrets that competitors cannot easily replicate.
South Korean government backing (K-chips Act tax credits) plus US CHIPS Act funding for the Taylor, TX fab, reinforced by the Tesla AI5/AI6 foundry agreement tied to US-based advanced manufacturing, gives Samsung durable government-linked positioning in both Korea and the US.
SmartThings connects 300M+ devices, but this reflects installed-base scale rather than true Metcalfe's Law dynamics — Google Home and Apple HomeKit are close substitutes, and the switching cost is modest.
The industry-wide shift to 3-5 year HBM supply contracts now extends to Samsung's NVIDIA-qualified HBM4 volume, and Samsung Foundry's design-ins with Tesla (AI5/AI6) and IBM (2nm) create multi-year customer roadmap embedding similar to TSMC's and Micron's.
Samsung is a component and device supplier, not a system of record for any critical business function; customers dual- and triple-source memory and foundry capacity across Samsung, SK Hynix, Micron, and TSMC.
The largest memory manufacturer by capacity, which lowers commodity DRAM and NAND cost, but trailing SK hynix in HBM limits the advantage where margins are highest.
Galaxy is a global top-two smartphone brand, but it does not hold price the way Apple's does.
Growth Analysis
Growth Drivers
Key Risk
If HBM4 share gains stall against SK Hynix and Micron's multi-year lead, or if the memory cycle reverts in 2H 2027 as hyperscaler capex normalizes, DS division profit could compress sharply — mirroring the 2022-23 downcycle when Samsung's semiconductor division posted large losses. Foundry remains structurally unprofitable outside the Tesla anchor deal, and the June 2026 DRAM price-fixing class action (Samsung, SK Hynix, Micron) is an early-stage but incremental legal overhang shared across the oligopoly.
Score Derivation
77.1 base + 1.3 trajectory − 10 risk = 68
Base 77.1 (10–16% CAGR, midpoint 13%, measured from the peak-cycle 2026 base and diluted by DX) + 1.3 trajectory (memory accelerating — DS revenue ₩81.7T → ₩127.5T QoQ, HBM4 sales guided to more than triple in Q3; foundry and Mobile/CE stable: (1 − 0) / 3 × 4) + 0 margin (stable, kept from earlier today to match MU: operating margin rose from ~43% in Q1 to ~52% in Q2, but that is peak-cycle memory pricing, not a trend that can keep expanding from there) − 10 high risk (a hyperscaler capex pause reverting the memory cycle, graded the same as MU on the same commodity exposure). Foundry moves from accelerating to stable: 2nm wins from cloud and AI customers are still in the design phase and System LSI/Foundry losses widened to ₩2.1T in Q2, so there is no revenue acceleration to point to yet = 68
Memory Supercycle & Foundry Catch-Up
NVIDIA HBM4 qualification secured Q1 2026, after trailing SK Hynix and Micron through 2024-2025
DS division operating margin recovered to ~27% in Q1 2026 as DRAM/NAND ASPs rose sharply on AI-driven supply tightness
Samsung Foundry SF2 (2nm GAA) anchor customer Tesla (AI5/AI6, ~$16.5B multi-year deal); new IBM 2nm partnership announced
Foundry division losses narrowing but not yet profitable; still 1-2 process generations behind TSMC N2 on yield
Named with SK Hynix and Micron in a June 25, 2026 DRAM price-fixing class action (N.D. Cal.); no class certified yet
Price Scenarios (12–24 Months)
Valuation Analysis
Base-case fair value of ~₩325,000 is ~6× consensus NTM EPS (~₩54,000, implied by the ~5.3× forward P/E at ₩285,500) and ~1.8× Yuanta's 2027E book value per share of ₩179,148 — a peak-cycle multiple, not a re-rating to SK Hynix's ~3.5× trailing book. Consensus targets are far higher (36-analyst average ~₩475,850; Yuanta ₩630,000 at 3.5× 2027E book), but those capitalise 2027 operating-profit forecasts of ₩488–590T as if they were durable; that belongs in the bull case. ~₩325,000.
Valuation Multiples
| Forward P/E (NTM) | ~5.3× |
| Q2 2026 EPS (annualised) | ~6.6× |
| Price / 2027E Book | ~1.6× |
| 2027E Operating Profit | ₩488–590T |
| Consensus Target | ~₩475,850 |
On current-cycle earnings Samsung looks absurdly cheap — ~5× forward EPS and ~1.6× 2027E book — but that is what memory stocks look like near an earnings peak. The question is not whether the multiple is low but how much of the ₩488–590T 2027 operating-profit consensus survives the 2027–28 supply response. The ladder capitalises consensus NTM EPS at ~6× (base), a downcycle book multiple of ~1.1× (bear), and a cycle-extends-through-2028 case at ~2.65× 2027E book (bull).
Approximate figures as of September 24, 2026 (price ~₩285,500).
Where We Are vs Targets
Loading live price…
The memory cycle turns in 2027 as CXMT's new capacity and incumbent expansions hit a digesting AI-capex market; earnings fall sharply from the 2026 peak and the stock reverts toward a downcycle book multiple.
- ~1.1× Yuanta's 2027E book value per share of ₩179,148 (≈₩197,000) — a trough-style P/B, and ~3.7× today's consensus NTM EPS of ~₩54,000
- CXMT adds ~100,000 wafers/month of capacity in 2027 and another ~100,000 in 2028, pressuring commodity DRAM pricing where Samsung's exposure is largest
- AI hyperscaler capex digestion in 2H 2027 pulls DRAM/NAND contract prices down from 2026 records; DS operating margin compresses from the 50%+ Q2 2026 level
- HBM4 share gains stall behind SK Hynix and Micron, and Foundry fails to win a second anchor customer beyond Tesla
Consensus NTM earnings (~₩54,000 EPS) broadly delivered as supply stays tight into 2027, capitalised at a peak-cycle ~6× P/E — the market keeps discounting an eventual downturn rather than re-rating Samsung to a structural-growth multiple.
- ~6× consensus NTM EPS of ~₩54,000 (≈₩324,000), equivalent to ~1.8× Yuanta's 2027E BVPS of ₩179,148
- Q3 2026 operating profit lands near the ₩106–112T consensus as DRAM contract-price increases (up to ~20% pushed for Q3) flow through
- 2027 operating profit tracks the lower end of the ₩488–590T broker range as memory supply stays sold out through 2027
- HBM4 share holds and Foundry losses keep narrowing on the Tesla ramp; Mobile stays a drag rather than a driver
The supercycle extends through 2028 as HBM absorbs commodity DRAM capacity and AI inference drives high-capacity memory demand; the market starts paying a structural rather than cyclical multiple.
- ~2.65× Yuanta's 2027E BVPS of ₩179,148 (≈₩475,000), in line with the ~₩475,850 36-analyst average target and below SK Hynix's ~3.5× trailing book
- 2027 operating profit reaches the upper ₩590T (Yuanta) end of forecasts as supply constraints ease later than expected
- HBM4/HBM5 share reaches 30%+ and Samsung Foundry signs a second anchor customer at 2nm
- The conglomerate discount narrows as foreign buying and higher broker targets pull the multiple toward memory pure-play peers