Samsung Electronics: Device Solutions Supplied 99.7% of Second-Quarter Operating Profit, and at 230,500 Won the Shares Already Trade Above the 185,000 Won Conservative Value

The three-year shortage claim needs a larger discount. Samsung expects memory supply constraints to persist through 2028 and become more severe in 2027. Near-term tightness is plausible. HBM consumes more wafer area than ordinary DRAM per unit of usable bit output, sophisticated packaging remains constrained, and hyperscalers are still increasing AI infrastructure spending. The problem begins after that. SK hynix plans major additions at M15X and Yongin; Micron expects new Idaho DRAM output from around mid-2027; Samsung itself is adding capacity; and Chinese producer CXMT is expanding. Equipment spending among the leading producers is increasing sharply. Industry research supports tight DRAM conditions in 2027 but sees NAND loosening earlier, which already contradicts a uniform shortage thesis. The bear case should assume that supply responds successfully in 2027–28, compressing conventional-memory prices while AI demand remains healthy. It does not require an AI collapse.

Foundry presents a separate question. Second-quarter earnings improved on HBM base-die demand and U.S. customer orders. Samsung expects double-digit year-on-year foundry revenue growth in the second half, an advanced-node revenue mix above 50%, and mass production of second-generation 2-nanometre mobile products. These are encouraging indicators, but they do not yet prove that Samsung has built a broad external-customer franchise. TSMC still controls roughly 70% or more of outsourced foundry revenue, while Samsung’s share has been in the high-single digits. Much of Samsung’s near-term improvement appears linked to internal products, HBM base dies and a small number of large U.S. programmes. The durable inflection will arrive only when external customers use Samsung for repeat generations, not merely when utilisation rises because one internal or adjacent product fills a fab.

Outside semiconductors, the quarter was weak. The Device eXperience division reported KRW 48.0 trillion of revenue and a KRW 0.8 trillion operating loss. Mobile eXperience and Networks produced KRW 33.2 trillion of revenue and a KRW 0.7 trillion loss, compared with KRW 2.8 trillion of profit in the first quarter. Visual Display and Digital Appliances were slightly loss-making, while Samsung Display and Harman contributed only KRW 0.7 trillion and KRW 0.4 trillion respectively. This matters because the conventional argument for Samsung’s conglomerate structure is that smartphones, displays and consumer electronics stabilise semiconductor volatility. In the latest quarter, those businesses offered little stabilisation.

The balance sheet, however, remains exceptional. At the second-quarter end Samsung held roughly KRW 167.6 trillion of net cash. Quarterly operating cash flow was KRW 105.1 trillion against KRW 14.1 trillion of purchases of property, plant and equipment, producing about KRW 91 trillion of simplified free cash flow before intangible investment. This is peak-cycle cash generation rather than a sustainable quarterly run rate, but it gives Samsung room to fund advanced fabs, absorb a pricing downturn and return capital without relying on external finance.

Shareholder returns have improved but remain less decisive than the balance sheet permits. The FY2024–26 framework promises regular annual dividends of KRW 9.8 trillion and distribution of 50% of free cash flow over the policy period, subject to investment requirements. In 2025 Samsung paid KRW 11.1 trillion in dividends and repurchased KRW 8.2 trillion of shares. Some repurchases have been cancelled; other treasury shares have funded employee compensation, so gross buyback announcements should not be treated as equivalent to permanent share-count reduction. Investors are increasingly asking why a company with such a large net-cash balance and record semiconductor earnings retains a relatively modest ordinary-share dividend yield.

Governance remains part of the Korean discount. The board has an independent majority, but economic control lies within the wider Samsung group’s family-linked ownership structure rather than through a simple majority stake in Samsung Electronics itself. Related-party links, the role of Samsung Life and Samsung C&T, inheritance and succession considerations, and differing interests among ordinary shareholders, preferred shareholders and group affiliates complicate capital allocation. Korea’s “value-up” reforms and Samsung’s March 2026 value-enhancement plan may narrow the discount, but voluntary reform cannot replace a clear capital-return formula. Labour relations also remain a live operating issue: a large union threatened an extended strike in 2026 before reaching a settlement, while legal and internal disputes over the agreement continued.

The share price illustrates how quickly expectations have moved. The 005930 ordinary shares closed at KRW 230,500 on August 6, down 23.0% from the prior report’s KRW 299,500 reference and 38.5% below the latest 52-week high of KRW 374,500. The current price now sits within the previous report’s KRW 170,000–240,000 base range, rather than near the top of its bull range. That movement does not validate the old valuation automatically. The earnings base has risen so dramatically that the same price now represents a different expectation set. Conversely, the recent drop does not establish a bargain because the market is already discounting some normalisation from an extraordinary second-quarter print.

The qualitative portrait is a company in transition at a cyclical peak. Samsung has moved from being the HBM laggard toward becoming a credible second supplier for the newest generation, while its foundry is showing the first signs of improvement after years of weak utilisation and yield concerns. The transition is incomplete because neither HBM leadership nor foundry externalisation is proven. The current earnings record is real, cash-backed and strategically useful; it is also too concentrated and too cyclical to extrapolate.

Vertical history and financial evolution

Samsung Electronics began in Suwon in 1969 as Samsung Electronics Industry, at a time when South Korea was pursuing export-led industrialisation and domestic electronics production depended heavily on imported technology. Its early products were black-and-white televisions and household appliances. The original business solved a national industrial problem as much as a consumer problem: Korea needed manufacturing scale, foreign technology transfer and export capacity in products then dominated by Japanese and Western companies. Samsung’s early model was assembly-led and cost-driven. The present model, by contrast, is vertically integrated across component research, capital-intensive fabrication, product design, software, branding and global distribution.

The company listed on the Korean exchange on June 11, 1975 and adopted the Samsung Electronics name in 1984. Accessible contemporary records establish the listing date but do not provide a reliably reconcilable IPO price, proceeds figure or valuation on a modern post-split basis; those figures are excluded rather than reconstructed from secondary anecdotes. The listing initially represented a domestic industrial manufacturer, not the global semiconductor and smartphone franchise investors recognise today.

The first durable strategic turn came with semiconductors. Samsung entered memory when incumbents in the United States and Japan had deeper process knowledge, stronger patent positions and established customers. The decision required accepting years of heavy capital spending and learning-curve losses. Samsung’s eventual advantage was not a single invention. It was institutional willingness to build large fabs through downturns, compress process-development cycles and combine government-supported industrial growth, imported know-how, disciplined manufacturing and group financing. That capability later allowed Samsung to take share as Japanese memory producers retrenched and U.S. producers consolidated.

A useful division of Samsung’s history contains six stages:

The table compresses the chronology; the economic story lies in how each stage changed the company’s decision-making.

During the 1983–1997 semiconductor-building phase, Samsung chose scale over near-term returns. Memory manufacturing has a steep learning curve, and the cost per usable bit falls when a producer improves yield, shrinks process geometry and spreads fixed fab costs over larger output. A company that cuts capital spending sharply in a downturn can protect one year’s cash flow while losing the process position needed for the next upturn. Samsung repeatedly chose the opposite. The Asian financial crisis then forced Korean conglomerates to confront debt and portfolio complexity, but Samsung Electronics emerged with semiconductors and consumer electronics at its core.

From the late 1990s through 2012, the company’s vertical integration became a commercial weapon. Samsung could sell components to competitors while using related technologies internally. LCD scale supported televisions; memory and application processors supported mobile devices; distribution and marketing turned hardware expertise into a global brand. The company was no longer merely a low-cost producer. It occupied several layers of the electronics value chain and could move capital between them.

The 2013–2020 period showed both the power and the limits of that structure. Galaxy smartphones generated large cash flows, but Android commoditisation and competition from Apple at the premium end and Chinese vendors in the mass market reduced handset differentiation. OLED displays became a stronger component franchise. Memory cycles delivered exceptional profits, particularly during the 2017–18 shortage. Samsung also bought Harman for about USD 8 billion, acquiring automotive electronics and connected-car exposure rather than trying to build those customer relationships organically. The company spun off its printing-solutions business in 2016 and sold it to HP, a rational exit from a subscale activity that no longer fit the core portfolio.

Capital-market actions during this period reflected a maturing company. Samsung approved a 50-for-1 stock split in 2018, improving accessibility for domestic retail investors without changing enterprise value. The 2017 financial year had KRW 239.6 trillion of revenue, KRW 53.6 trillion of operating profit and KRW 42.2 trillion of net income, figures that captured the previous memory supercycle. The stock split lowered the nominal share price and makes unadjusted historical per-share comparisons misleading.

The governance narrative worsened in parallel. The imprisonment and later pardon of de facto leader Lee Jae-yong, disputes around group succession, and scrutiny of affiliate transactions reinforced the market’s view that Samsung Electronics’ cash might serve wider group objectives rather than only the interests of its listed shareholders. Those concerns never prevented the company from investing effectively in memory or displays, but they affected the multiple investors were willing to pay for its cash and operating assets.

The 2021–2025 stage exposed a strategic execution problem. Artificial-intelligence accelerators increased demand for high-bandwidth memory, but SK hynix established an early qualification and packaging lead. Samsung retained broad DRAM scale yet did not translate that scale into first-mover HBM leadership. Foundry also struggled to narrow the gap with TSMC. Public yield estimates varied widely, customer concentration remained high and utilisation weakened outside a few advanced products. The memory downturn then pushed group operating profit to KRW 6.6 trillion in 2023, down from KRW 43.4 trillion in 2022, despite Samsung’s continued spending. Revenue fell to KRW 258.9 trillion in 2023 before recovering to KRW 300.9 trillion in 2024 and KRW 333.6 trillion in 2025.

The financial record demonstrates cyclicality more clearly than any narrative label:

Sources: Samsung AGM disclosure and consolidated financial-performance archive.

Revenue grew by only about 39% between 2017 and 2025, while operating profit moved through a range from KRW 6.6 trillion to more than KRW 53 trillion. The main variable was not unit growth alone. Memory pricing, mix, inventory adjustments and utilisation converted relatively modest revenue changes into enormous earnings swings. This pattern warns against using a single-year price-to-earnings ratio.

The recent recovery also changed the internal mix. In 2023, Device eXperience represented about 60% of divisional sales and Device Solutions 24%. By 2025 those shares were approximately 52% and 36%. In the second quarter of 2026, Device Solutions revenue before eliminations was more than two and a half times Device eXperience revenue, and its operating profit dwarfed every other segment.

Cash-flow conversion has been stronger than reported net income over a full cycle, partly because depreciation on past fab investment is large and working capital releases cash in recoveries. In 2025, operating cash flow was KRW 85.3 trillion against KRW 45.2 trillion of net income. Purchases of property, plant and equipment were KRW 47.5 trillion and intangible investment was KRW 4.6 trillion, leaving conventional free cash flow of roughly KRW 33.2 trillion. In 2024, operating cash flow was KRW 73.0 trillion and property, plant and equipment purchases were KRW 51.4 trillion.

Across 2021–25, Samsung’s aggregate operating-cash-flow-to-net-income ratio was approximately 1.7 times, using the audited annual cash-flow statements. That does not mean accounting profit is understated by 70%. Depreciation is a real economic cost in a company whose semiconductor tools become obsolete quickly, and working-capital movements reverse over time. The ratio instead shows that reported earnings have generally been cash-backed, while free cash flow remains highly sensitive to how much of capital expenditure is treated as necessary to maintain competitive capacity.

Samsung depreciated approximately KRW 43.6 trillion in 2025, close to its KRW 47.5 trillion property-and-equipment spending. Semiconductor production equipment commonly carries useful lives around five years in the company’s accounting. A defensible 2025 maintenance-capital estimate is KRW 35–40 trillion, with the remaining KRW 8–13 trillion treated as growth or strategic expansion. On that basis, 2025 owner earnings were approximately KRW 45–50 trillion, calculated as operating cash flow less estimated maintenance capital. The result is close to reported net income, so the owner-earnings adjustment does not exceed the assignment’s 30% threshold.

The distinction becomes harder in 2026. HBM, advanced DRAM and 2-nanometre foundry expansion are simultaneously maintenance and growth spending. They expand capacity, but a producer that does not make those investments may lose the right to compete. The valuation below does not add back all “growth capex.” It normalises cash earnings after a capital charge sufficient to keep Samsung technologically relevant.

Inventory was KRW 52.6 trillion at the end of 2025, little changed from KRW 51.8 trillion a year earlier, with an allowance of KRW 5.8 trillion. The balance did not show an uncontrolled build entering 2026, but memory inventory can lose value rapidly when contract prices fall. Inventory days and valuation allowances should be read alongside price trends rather than in isolation.

Equity-method investments are material to the balance sheet but not the present earnings thesis. Associates and joint ventures had a carrying value of about KRW 13.8 trillion and contributed approximately KRW 0.7 trillion of 2025 profit. Stakes include Samsung Electro-Mechanics, Samsung SDS, Samsung Biologics, Samsung Bioepis and Samsung SDI. The valuation treats those holdings as non-operating assets rather than combining affiliate revenue with Samsung Electronics’ operations.

Minority interests require a similar adjustment. Non-controlling interests represented about KRW 12.0 trillion of the group’s KRW 436.3 trillion of year-end 2025 equity and received approximately KRW 0.9 trillion of consolidated profit. The valuation uses earnings attributable to Samsung Electronics shareholders and deducts non-controlling interests from enterprise value. Samsung Display is consolidated; its sales and profit are not added again as an affiliate asset.

The balance sheet remains the most stable element of the vertical record. Samsung has repeatedly accumulated cash during upcycles, spent aggressively through downturns and avoided dependence on leverage. Interest expense was only about KRW 0.6 trillion in 2025, while the group ended the second quarter of 2026 with approximately KRW 167.6 trillion of net cash. This capital advantage is a moat because semiconductor downturns often force weaker firms to delay node transitions precisely when investment is most productive.

Returns on capital are less consistently superior than the balance sheet suggests. Peak memory margins produce very high returns, but foundry losses, underutilised fabs and consumer-electronics capital dilute them. The 2023 operating trough pushed group returns toward low single digits despite enormous accumulated assets. A full-cycle view supports a mid-teens return on tangible operating capital for the better businesses, not a permanent peak-cycle return for the consolidated group.

The share-price history mirrors those shifts. The 2016–18 rise priced a memory supercycle and stronger shareholder returns. The 2018–19 decline priced inventory correction and trade uncertainty. The 2020–21 advance combined pandemic electronics demand, memory expectations and retail liquidity. The 2022–23 fall tracked memory oversupply and a collapse in operating profit. During 2024–25, Samsung underperformed the pure HBM beneficiaries because investors assigned more value to SK hynix’s proven Nvidia position than to Samsung’s larger but less effective memory footprint.

In 2026 the market first marked Samsung upward as HBM4 qualification improved, memory prices surged and operating-profit estimates rose. The 005930 ordinary line reached a 52-week high of KRW 374,500 before falling to KRW 230,500 by August 6. The same session brought sharp declines across Korean semiconductor shares as investors questioned whether record AI-memory earnings had already been capitalised. Samsung’s common line had briefly exceeded KRW 2,100 trillion in market value earlier in the year, illustrating how much of the rerating rested on continued record profits.

The market has assigned Samsung several identities over time: export manufacturer, consumer-electronics champion, smartphone growth company, memory cyclical, cash-rich conglomerate and AI-memory catch-up story. Its present valuation combines the last three. That combination deserves a higher multiple than a conventional commodity producer if HBM4 share is sustained, but a lower multiple than a pure structural compounder because conventional DRAM, NAND and foundry still require enormous reinvestment and remain exposed to cycles.

Business model, moat, industry and cycle

Samsung’s operating machine contains four economically different systems.

Device Solutions includes Memory, System LSI and Foundry. Memory sells DRAM, NAND, enterprise SSDs and HBM. Conventional memory economics depend on bit shipments, contract prices, yields, utilisation and industry supply discipline. HBM adds co-design, qualification, advanced packaging and long-term allocation agreements. System LSI designs application processors, image sensors and other logic products. Foundry manufactures chips designed by internal and external customers.

Device eXperience includes smartphones, tablets, wearables, networks equipment, televisions, appliances and related services. Its gross margins are structurally lower than peak memory margins, product cycles are shorter, and brand and distribution matter more than wafer economics. Samsung Display manufactures mobile and other panels, including OLED. Harman supplies automotive electronics, connected-car systems and audio products.

The 2025 segment structure showed a relatively balanced revenue portfolio:

Segment revenue includes internal transactions, so the sum exceeds consolidated revenue.

The second quarter of 2026 bore little resemblance to that balance:

Segment revenue includes intercompany transactions and cannot be summed directly to consolidated sales.

The business reason behind the change is pricing leverage. Semiconductor fabs carry high fixed costs. Once utilisation is high and selling prices rise, incremental revenue can flow through at very high margins because the depreciation, engineering and facility costs already exist. The reverse is also true. When prices fall and fabs operate below capacity, profit can disappear much faster than revenue. Device eXperience has less dramatic operating leverage but faces marketing, component and channel costs that cannot be reduced instantly when a phone launch or appliance cycle weakens.

Memory’s cost structure consists of fab depreciation, process equipment, clean-room infrastructure, wafers, chemicals, energy, engineering labour, testing and packaging. Variable costs per chip are low relative to the capital base. Yield improvement matters as much as nominal capacity. A producer with a 10-percentage-point yield advantage can obtain considerably more saleable dies from the same equipment and floor space.

HBM changes that equation in two ways. It consumes more DRAM wafer capacity per usable bit than conventional products, tightening the broader market when producers shift mix toward HBM. It also adds through-silicon vias, stacking, bonding, testing and customer-specific base-die requirements. Packaging capacity and final-stack yield become bottlenecks. That complexity supports higher prices, but it creates more points at which execution can fail.

Samsung’s strongest moat is capital-backed manufacturing learning. The company can spend tens of trillions of won annually through the cycle, operate at enormous scale and absorb years of suboptimal returns in a new node. The net-cash position reinforces that advantage. New entrants cannot replicate it by buying equipment alone; they need process recipes, yield data, suppliers, engineers and customer trust accumulated over generations.

Its second moat is vertical integration across memory, foundry, logic design, displays and finished devices. Integration allows Samsung to test new components internally, coordinate an HBM stack with its base die and use smartphones as an anchor customer for displays, image sensors and application processors. The HBM4 generation may finally show the strategic value of combining memory and advanced logic manufacturing.

Vertical integration is not self-executing. It did not prevent Samsung from falling behind SK hynix in HBM3E, and it did not persuade leading fabless designers to treat Samsung Foundry as interchangeable with TSMC. Organisational complexity can slow decisions, encourage internal sourcing without market validation and blur accountability between divisions. Integration is a real moat when it shortens qualification and lowers total system cost. It becomes a marketing moat when internal volume hides weak external demand.

The third moat is global consumer distribution and brand. Samsung can launch phones, televisions and appliances across almost every major market, negotiate with carriers and retailers, and support a broad product range. That channel scale remains difficult to reproduce. The weakness is that Android limits software differentiation. Apple controls hardware, operating system, services and premium ecosystem economics more tightly. Chinese smartphone vendors can compete aggressively on specifications and price. Samsung’s device franchise protects scale and cash generation more effectively than it creates rising margins.

OLED display technology and manufacturing form a fourth moat. Samsung Display has longstanding expertise in mobile OLED and customer relationships with premium-device producers. The economics are better than those of commoditised LCD, though Chinese panel makers continue to add capacity and narrow performance gaps.

Foundry has the assets of a moat but has not yet proved the customer behaviour associated with one. A true foundry moat consists of process-design kits, electronic-design-automation support, intellectual-property libraries, packaging, predictable yield, delivery reliability and a broad ecosystem that lowers a customer’s design risk. TSMC’s advantage lies in this accumulated system. Samsung’s advanced-node equipment and transistor architecture are necessary but insufficient.

Industry structure supports unusually high memory profitability when supply is disciplined. DRAM has consolidated around Samsung, SK hynix and Micron, with CXMT emerging in China. NAND is less concentrated and includes Kioxia, Western Digital’s former flash operations, Micron, SK hynix/Solidigm and Chinese suppliers. DRAM’s three-incumbent structure generally enables better supply discipline than historical memory markets, but capital responses still occur when margins become extraordinary.

HBM is currently even more concentrated because qualification and packaging raise barriers. SK hynix holds the strongest incumbent position, Samsung is the recovering challenger, and Micron has become a credible third source. Customers want multiple suppliers to reduce dependency and improve bargaining power, which creates an opening for Samsung even without technological leadership. The profit pool concentrates with suppliers that secure high-volume accelerator qualifications and can deliver packaged stacks at acceptable yield.

Downstream bargaining power is mixed. Nvidia has enormous influence because its accelerator platforms determine HBM specifications and allocation. Large cloud customers can negotiate long-term commitments but cannot instantly create qualified supply. In conventional DRAM and NAND, large server and device customers have greater ability to switch among suppliers, making price the dominant variable when supply is abundant.

The current industry upcycle combines four overlapping cycles. The first is a conventional memory inventory and pricing cycle. The second is an AI infrastructure capital-expenditure cycle. The third is a technology iteration cycle from HBM3E to HBM4 and HBM4E. The fourth is a fabrication and packaging capacity cycle. These cycles will not peak simultaneously.

AI demand is the secular component. HBM content per accelerator and aggregate accelerator shipments can rise for years even if annual growth slows. Conventional-memory pricing is the cyclical component. Server DRAM, enterprise SSDs and high-end NAND benefit from AI infrastructure, but their supply is more fungible and additions arrive faster than new HBM packaging ecosystems.

Samsung is presently in the steep part of the pricing upcycle and the early part of its own HBM share recovery. That combination produces exceptional earnings momentum. It also places the company near the point where competitors have the strongest incentive to add capacity. The 70% Device Solutions operating margin is evidence of scarcity, not a sensible mid-cycle assumption.

Management’s shortage guidance through 2028 rests on legitimate constraints. HBM migration reduces conventional DRAM bit output, advanced packaging remains tight, and AI-server shipments and hyperscaler capital spending are growing. Industry estimates point to roughly 31% AI-server shipment growth and sharply higher capital spending by major cloud providers in 2026.

The supply response is already visible. SK hynix is expanding M15X and plans initial production at Yongin; Micron’s Idaho fab is expected to contribute from around mid-2027; Samsung is preparing additional capacity; and CXMT’s projects could materially increase Chinese wafer starts. Semiconductor-equipment spending by the major memory producers is rising. TrendForce expects DRAM to stay tight into 2027 but sees NAND conditions easing sooner. The evidence supports a split conclusion: scarcity through much of 2027 is plausible, while a uniform shortage through 2028 carries a low confidence level.

Government policy and geopolitics shape both demand and supply. The United States and allied countries subsidise domestic semiconductor fabrication while restricting advanced equipment and technology transfers to China. Samsung operates important manufacturing capacity in China, including NAND production, and changes in validated-end-user treatment or equipment licensing can raise costs, limit upgrades or reduce asset flexibility. Samsung and SK hynix have explored greater use of Chinese tools in local fabs after U.S. policy changes, indicating that export controls are becoming a recurring operating constraint rather than a one-time headline.

Chinese policy creates the opposite force. Subsidies and state-backed capital support domestic memory and foundry capacity. CXMT may not immediately match the most advanced HBM products, but it can displace commodity DRAM demand inside China and pressure global pricing at older nodes. SMIC plays a similar role in mature-node logic. The risk to Samsung runs in two directions: export controls constrain its Chinese assets while Chinese competitors add supply in products with fewer restrictions.

Korean policy affects valuation more directly. The government’s corporate-value-up programme seeks better disclosure, stronger shareholder returns and a narrower governance discount. Initial participation was voluntary and disappointed investors who expected binding reform. Samsung’s 2026 value-enhancement plan and larger investment commitments respond to the initiative, but the decisive test is whether excess cash reaches shareholders through cancelled shares and a durable dividend formula.

The verdict on management is mixed. Samsung has demonstrated extraordinary long-run capital allocation into memory, OLED and manufacturing scale. The Harman acquisition created a credible automotive-electronics position without destabilising the balance sheet. The company’s willingness to invest through the 2023 semiconductor trough was strategically correct.

Execution credibility is weaker in the areas now carrying the valuation. HBM qualification took longer than competitors’, and foundry guidance has repeatedly preceded proof of broad external adoption. Current management claims should receive different weights: disclosed revenue, profit and cash are high-confidence; signed long-term agreements provide useful volume visibility; multi-year shortage forecasts and undisclosed yield claims receive heavy discounts.

The ownership structure also limits alignment. Samsung Electronics does not use a U.S.-style dual-class structure, and the listed preferred shares have similar economic rights but no ordinary voting rights. Control is nevertheless exercised through a network of family-linked stakes and affiliates rather than proportional direct ownership. The common shares deserve some voting premium over the preferred line. Yet the 25.3% closing-price discount of the preferred shares on August 6 is larger than differences in dividend rights alone would imply. It reflects liquidity, index inclusion, voting value and investor segmentation.

The board comprises three executive and five independent directors. That formal majority is positive, yet board composition has not removed concerns about succession, affiliate transactions and cash deployment. Management credibility is consequently higher in engineering and balance-sheet stewardship than in minority-shareholder alignment.

Horizontal competition and current fundamentals

Samsung has ample competitors, but no single peer matches its portfolio. The correct horizontal analysis uses three primary reference companies and several secondary ones. SK hynix is the closest HBM and DRAM competitor. Micron is the global memory peer with a simpler corporate structure and U.S. capital-market valuation. TSMC is the foundry benchmark. Apple and Chinese Android vendors frame Samsung’s device economics, while Nvidia is both a customer-platform reference and the most important allocator of HBM demand.

SK hynix became the specialist that converted early HBM co-development into qualification leadership. Customers choose it because it established a record of shipping high-volume HBM stacks into leading accelerator platforms and because its packaging partnership and process execution reduced customer risk. Its business is less diversified than Samsung’s and therefore more exposed to memory pricing, but that concentration became an advantage when HBM was the industry’s most valuable profit pool. The market rewarded proof rather than theoretical integration.

Samsung became the scale integrator. Customers choose it for supply depth, manufacturing breadth, the possibility of combining DRAM and custom base dies, and the security of a second qualified source. Its HBM case is strongest when customers want to diversify away from SK hynix and when Samsung’s foundry capability improves stack performance or cost. Customers avoid or limit Samsung when qualification uncertainty threatens accelerator launch schedules.

Micron became the focused third supplier. Its smaller overall memory scale can make rapid allocation gains economically meaningful. It reported fiscal third-quarter 2026 revenue of USD 41.46 billion, operating cash flow of USD 25.39 billion and GAAP net income of USD 28.24 billion, illustrating how extreme the memory-pricing environment had become. Micron’s simpler reporting gives investors clearer exposure to the cycle, though it lacks Samsung’s consumer franchise and balance-sheet scale.

TSMC became the risk-minimisation platform for fabless semiconductor design. Customers accept premium pricing because its ecosystem, yield consistency, process-design tools, capacity planning and packaging reduce the probability that a multibillion-dollar chip programme misses its launch. Samsung Foundry competes on transistor technology, price, capacity alternatives and the strategic desire for a second source. The gap is measured less by laboratory node labels than by repeat external-customer tape-outs and commercially viable yield.

The approximate competitive cross-section is:

HBM shares are estimates and vary by quarter, output basis and source methodology; foundry shares also vary depending on whether the dataset covers pure-play revenue or total manufacturing.

The numbers reveal why broad “semiconductor peer” multiples can mislead. SK hynix and Micron have purer exposure to the present memory profit pool. TSMC has superior foundry quality and lower commodity exposure. Samsung owns more assets, more consumer franchises and more net cash, but several of those operations earn lower returns. A conglomerate discount can be economically rational even without governance concerns because the portfolio mixes businesses that deserve different multiples.

HBM share estimates require care. One Reuters-cited dataset placed SK hynix at 53%, Samsung at 35% and Micron at 11% in the third quarter of 2025. Other estimates showed Samsung materially lower earlier in that year. TrendForce’s 2026 output forecast put SK hynix around 50% and Samsung around 28%. Differences arise from bit output versus revenue, shipment timing and the treatment of inventory. The robust conclusion is narrower: SK hynix remains the leader, Samsung has recovered from a weak 2025 position, and Micron remains a meaningful third supplier.

Samsung’s Nvidia qualification appears genuine. Independent reporting in June 2026 said HBM4 passed qualification and commercial supply was approved. This changes the probability distribution materially because Samsung can now participate in the Vera Rubin generation rather than remaining a perpetual “nearly qualified” supplier. SK hynix still entered the allocation process with earlier customer experience, and public allocation estimates generally leave it with the largest share.

Pricing visibility has improved through longer supply agreements. Industry reports describe three-to-five-year HBM arrangements and large portions of 2026–27 capacity already committed. These contracts reduce volume uncertainty but do not necessarily fix price for their full duration; renegotiation clauses, specification changes and generation transitions can reset economics. The valuation treats them as backlog-like evidence, not bond-like revenue.

Yield is the largest unresolved HBM variable. Samsung disclosed performance specifications for HBM4 and discussed qualification, but it did not disclose a customer-verified mass-production yield. SK hynix has publicly indicated an objective of bringing HBM4 yield toward HBM3E levels. Without comparable audited data, any precise Samsung-versus-SK yield estimate is speculative. A valuation that assumes parity immediately would outrun the evidence.

Foundry evidence is similarly mixed. Samsung’s second-quarter results improved because of HBM base dies and U.S. customer demand. Management expects second-half revenue to grow at a double-digit rate, advanced-node mix to exceed 50%, and second-generation 2-nanometre mobile products to enter mass production. Reports of major U.S. programmes, including Tesla-related production and discussions with Qualcomm, give Samsung credible opportunities to diversify beyond internal Exynos and HBM-related work.

Yet public 2-nanometre yield estimates range from about 40% to around 60%, while estimates for TSMC are often higher. Such a wide range indicates that the market lacks dependable data. The report does not use a precise foundry yield assumption. It requires observable evidence: sustained external revenue, utilisation above breakeven, repeat customer generations and an end to material divisional losses.

TSMC’s lead is structural today. Its foundry share exceeds 70% in several recent datasets, and advanced nodes comprise a large portion of wafer revenue. Samsung can still win specific customers that want geographic or supplier diversification, negotiate aggressive economics or benefit from its gate-all-around process. A few contracts can improve Samsung’s utilisation sharply because its external revenue base is much smaller. That operating recovery should not be confused with ecosystem parity.

The last four quarters show how rapidly the consolidated company changed:

Sources: Samsung quarterly disclosures and contemporaneous reporting.

The base effect is visible in the table. Device Solutions profit rose from KRW 0.4 trillion to KRW 89.2 trillion in one year. The year-ago quarter carried the combined weight of poor utilisation, HBM delays and restrictions affecting China-related operations. The current quarter contains the reverse: record selling prices, strong server demand, richer HBM mix and high utilisation.

Profit quality was strong in cash terms. Consolidated net income was approximately KRW 71.6 trillion, and operating cash flow reached KRW 105.1 trillion. Simplified free cash flow before intangible investment was about KRW 91 trillion. The operating-cash-flow-to-net-income ratio of roughly 1.47 times indicates that the record earnings were not merely accruals.

The quality issue lies in sustainability rather than conversion. A 70% Device Solutions margin is above any defensible mid-cycle assumption. It embeds shortage pricing, favourable product mix and high fab utilisation. Even if HBM remains structurally attractive, conventional memory prices will not need to collapse for margins to normalise materially.

Mobile weakened sharply. Mobile eXperience and Networks revenue declined to KRW 33.2 trillion from KRW 38.1 trillion sequentially, while operating profit moved from KRW 2.8 trillion to a KRW 0.7 trillion loss. Product timing, component costs and marketing affected the quarter, but the result also shows that Samsung’s handset franchise no longer supplies the stable high-margin counterweight it did during Galaxy’s strongest years.

Samsung Display remained profitable at KRW 0.7 trillion, and Harman earned KRW 0.4 trillion. These businesses are useful diversifiers and possess attractive customer positions, but together they contributed only about 1.2% of consolidated operating profit in the quarter.

Analyst estimates rose sharply after the preliminary and final results, but the share-price reaction shows that estimate upgrades alone no longer guarantee positive returns. Semiconductor equities fell on August 6 even after record industry profits because investors had begun to question the duration of AI-memory scarcity and the amount of leverage embedded in expectations. SK hynix experienced a similar pattern: record results did not prevent a negative market reaction when shipment timing and expectations became more demanding.

The market is trading five variables. The largest is memory pricing. The second is Samsung’s HBM4 share recovery. The third is the duration of AI infrastructure spending. The fourth is whether foundry losses turn into sustainable external-customer earnings. The fifth is capital return from the growing net-cash balance.

The first three are already well represented in the stock. Even after the decline from KRW 374,500, the ordinary shares trade at a substantial premium to 2025 book value and at a mid-teens multiple of this report’s normalised earnings estimate. The market is not pricing the second-quarter profit as permanent, but it is pricing a material structural HBM franchise and an extended period of above-normal memory profitability.

The bull case rests on concrete evidence. Samsung has passed a critical HBM4 qualification, management expects a rapid sequential HBM4 ramp, long-term supply agreements cover a large portion of capacity, and the balance sheet can fund both memory and foundry expansion. If Samsung’s HBM share moves above 30% while HBM economics remain better than conventional DRAM, normalised earnings power is higher than in previous cycles.

The bear case does not require technological failure. Capacity additions can normalise conventional-memory prices in 2027–28; SK hynix can remain the preferred supplier for the most valuable allocations; Micron can retain a meaningful share; and foundry can remain structurally subscale. Under that path, Samsung still earns money and remains financially strong, but current peak earnings fall toward a much lower mid-cycle level.

The most important disagreement is the split between cyclical and structural profit. Bulls implicitly classify a large portion of the current Device Solutions profit as durable because AI changes memory content and supply requirements. Bears classify most of it as scarcity rent that will attract capacity. This report places HBM structural profit above the historical cycle but treats the majority of second-quarter margin as temporary.

Valuation, risks and tracking

The current price produces three incompatible-looking valuation statistics. Against 2025 ordinary EPS of KRW 6,605, the stock trades at about 34.9 times earnings. Against annualised second-quarter 2026 EPS of roughly KRW 43,400, it trades at only about 5.3 times. Against estimated mid-cycle EPS of KRW 14,000–16,000, it trades at approximately 14.4–16.5 times. The first statistic capitalises a pre-boom year, the second capitalises a record quarter, and the third is the economically relevant starting point.

Price-to-book analysis also needs class discipline. At 2025 year-end, consolidated equity attributable to owners was approximately KRW 424 trillion. That equity supports both ordinary and preferred economic claims, not only the ordinary line. Dividing it by the combined outstanding economic share count produces book value around KRW 64,000 per share before the large 2026 profit accumulation. The current ordinary price was therefore about 3.6 times trailing book, while a post-earnings forward book value would be materially higher. A trailing P/B comparison understates the new equity base but still shows that the shares no longer trade like a low-return Korean conglomerate.

Samsung historically moves between approximately one times trough book value and materially above two times when memory earnings, technology confidence and shareholder returns align. The present effective multiple is elevated relative to ordinary cycle troughs but lower than the 2026 speculative high. The centre can remain above earlier decades if HBM makes a larger proportion of memory profit less commodity-like and if governance reforms improve cash returns. Neither condition is fully proven.

Peer valuation offers context, not an answer. SK hynix earns a HBM leadership premium but carries greater memory concentration. Micron offers cleaner cycle exposure and more transparent segment economics. TSMC deserves a higher-quality multiple because its customer ecosystem, return on capital and earnings visibility are superior. Samsung deserves a balance-sheet premium to weaker memory producers and a conglomerate discount to TSMC. A valuation derived by averaging those peers would mix incompatible economics.

The absolute valuation begins with cash-flow passthrough. Over 2021–25, Samsung generated roughly KRW 1.7 of operating cash flow for every KRW 1 of net income. After deducting estimated maintenance capital, 2025 owner earnings of KRW 45–50 trillion were close to reported earnings. The scenarios therefore use the owner-earnings basis, but it does not produce a radically different valuation from normalised net income.

The report separates traditional memory, structural HBM, logic/foundry and the non-semiconductor businesses. It then capitalises normalised owner earnings and adds net non-operating assets after minority interests. The calculations use the combined economic share count of ordinary and preferred shares, because both classes participate in earnings, while the final price signal applies to the more liquid voting ordinary line.

This is valuation-scenario analysis within a research framework, not investment advice.

The conservative case assumes conventional-memory prices normalise during 2027, NAND loosens earlier than DRAM, Samsung’s HBM share remains in the mid-to-high 20s, and foundry stays loss-making. Owner earnings remain above 2025 because HBM contributes a larger structural profit pool and the balance sheet generates income, but they fall far below the annualised second-quarter level.

The base case assumes Samsung retains roughly 28–32% HBM share, HBM4 and HBM4E carry margins above conventional DRAM, traditional memory settles at healthy rather than shortage-level pricing, and foundry reaches modest profitability through base dies and a limited set of external customers. Device eXperience returns to a normal mid-single-digit margin, while Samsung Display and Harman continue to contribute.

The optimistic case assumes Samsung becomes a durable co-leader in HBM, its share moves into the low-to-mid 30s, the industry remains supply constrained through much of 2028 and foundry wins repeat 2-nanometre customers. The 16.5-times multiple reflects improved earnings quality but remains below a pure structural-growth semiconductor multiple because memory and foundry remain capital-intensive.

The most fragile base-case assumption is HBM operating profit. Reducing that contribution to 70% of the base estimate removes approximately KRW 12.6 trillion of pre-tax operating profit and roughly KRW 10 trillion of owner earnings. At the same 15-times multiple, base value falls by about KRW 23,000 per economic share, from roughly KRW 240,000 to about KRW 217,000. That result would place the current price above revised base value.

The conservative fair value of KRW 185,000 is below the KRW 230,500 current price, so the margin of safety relative to the conservative case is zero. A purchase zone with at least a 20% discount to conservative value begins below KRW 148,000.

If normalised earnings remain flat for three years and the share price finishes near its current level, the expected return consists largely of dividends. The ordinary annual dividend of KRW 1,668 for 2025 equates to a yield of about 0.72% at KRW 230,500, far below South Korea’s approximately 4.16% ten-year government-bond yield on August 6. Under a flat-earnings, flat-multiple outcome, there is no margin of safety at this buy price.

This is not a simple “good company, bad price” case. The price is close to base value rather than plainly excessive, and a strong HBM outcome could create substantial upside. The absence of conservative-case protection makes waiting rational for a buyer who requires a margin of safety. An existing holder faces a different decision, because selling also gives up the chance that Samsung converts HBM4 qualification into durable share.

The margin-of-safety sufficiency verdict is none.

The market’s expectation gap will be decided by operating variables rather than the next headline profit percentage. The most important near-term data are HBM4 revenue mix, customer allocation, Device Solutions margin, conventional DRAM contract pricing, foundry losses and capital-return policy. A third-quarter profit beat driven only by another conventional-memory price increase would be less valuable than a smaller beat accompanied by independently visible HBM share gains and foundry external revenue.

The first permanent-loss risk is supply normalisation. Its probability is high and its impact medium to high. The transmission path begins with wafer additions, packaging expansion and improved yields. Contract DRAM and NAND prices then fall even while bit demand grows. Device Solutions margin could move from 70% toward 25–35%, reducing group owner earnings and compressing the multiple as the stock loses its shortage premium. Observable indicators include equipment orders, wafer-start guidance, customer inventories and sequential contract-price changes.

The second risk is an HBM allocation reversal. Its probability is medium and its impact high. Nvidia or another major accelerator customer could allocate HBM4E and later-generation volume primarily to SK hynix and Micron because of yield, thermals, power or packaging reliability. Samsung would then sell more conventional DRAM into a looser market, reducing both revenue mix and valuation quality. The early indicators are Samsung HBM share below 25%, repeated qualification delays, lower-than-guided HBM4 mix and customer-specific shipment deferrals.

The third risk is a false foundry turn. Its probability is medium to high and its impact medium. HBM base-die orders and one or two U.S. contracts can raise utilisation without creating a broad ecosystem. If 2-nanometre yield remains below commercial targets, external customers will continue to concentrate leading designs at TSMC. Samsung then carries depreciation on advanced fabs without adequate revenue, eroding returns on capital and consuming cash that could otherwise be returned to shareholders. The observable evidence is foundry breakeven, repeat-generation customer wins and utilisation after the initial HBM base-die ramp.

The fourth risk is capital allocation and governance. Its probability is medium and its impact medium to high. The company can produce enormous cash while leaving minority shareholders with a low yield and ambiguous treatment of treasury shares. If capital is committed to low-return foundry capacity or group objectives without a stronger cancellation and dividend policy, the Korean conglomerate discount can persist even as operating earnings rise. The indicators are net cash, cumulative share cancellation, dividend-policy changes and related-party transactions.

The fifth risk is geopolitical constraint on Chinese operations. Its probability is medium and its impact high under an adverse policy change. Restrictions on advanced equipment could impair upgrades at Chinese fabs, force inefficient tooling substitutions or create asset write-downs. The valuation effect would extend beyond lost output because investors would apply a higher discount to all China-dependent semiconductor assets.

The sixth risk is labour disruption. Its probability is low to medium after the 2026 settlement, but impact could be medium if a dispute reaches advanced memory production or delays customer shipments. The threatened 18-day strike was suspended after agreement, but challenges to the settlement and arguments over incentive distribution remained. The main indicators are union participation, production-line staffing and whether bargaining moves from compensation to operational action.

Positive catalysts include confirmation that HBM4 exceeds 60% of second-half HBM revenue; independent evidence that Samsung HBM share moves above 30%; foundry breakeven supported by external rather than internal customers; a binding capital-return policy with large-scale share cancellation; and evidence that new capacity starts later or yields less than expected.

Negative catalysts include sequential deterioration in DRAM contract pricing; an HBM4E qualification delay; Samsung HBM share remaining below 25%; foundry losses widening despite higher revenue; net cash falling below KRW 100 trillion without corresponding shareholder distributions; or labour and export-control developments that interrupt production.

The tracking dashboard converts those risks into observable thresholds:

Samsung had not formally announced the third-quarter reporting date by the research base date. Its established calendar and current IR event schedule imply a release in late October 2026, likely around October 29–30; this remains an estimate until the company posts the notice.

HBM share matters because it tests whether Samsung’s recovery is structural. The revenue mix tests whether management’s Q3 ramp appears in actual sales. Device Solutions margin identifies how much scarcity rent remains. Foundry profit tests whether utilisation is economically productive. Cash flow and net cash show whether accounting earnings survive capex. The preferred discount supplies a daily market signal on governance, liquidity and common-share voting value.

Cross-synthesis, key data, uncertainties and sources

Looking vertically, Samsung has proved one capability more convincingly than any other: it can convert capital, engineering depth and manufacturing repetition into scale leadership after entering from behind. That capability built its DRAM franchise, televisions, mobile displays and global device distribution. It survived multiple memory cycles because the balance sheet allowed it to invest when industry returns looked unattractive.

Past success did not come from one source. Korea’s export-led industrial policy provided an enabling environment; global electronics demand created the opportunity; memory cycles supplied periods of extraordinary profit; and management’s willingness to make long-duration capital commitments converted those tailwinds into share. Luck mattered in the timing of competitors’ retreats, but luck does not explain decades of process execution.

The same success factors remain present, with two qualifications. Capital strength is intact and larger than ever. Manufacturing breadth remains difficult to replicate. Execution has become less consistent at the frontier. Samsung’s HBM3E delay and foundry gap show that scale cannot substitute for customer-specific qualification, ecosystem quality and organisational speed.

Horizontally, Samsung’s real advantage over SK hynix is breadth and integration. Its disadvantage is that SK hynix has turned HBM specialisation into customer trust and allocation leadership. Against Micron, Samsung has greater scale, more internal demand and stronger net cash, but Micron offers investors cleaner exposure and can move faster as a focused supplier. Against TSMC, Samsung possesses advanced equipment and a credible process roadmap but lacks the same external-customer ecosystem and yield reputation.

The HBM weakness looks partly temporary. Passing HBM4 qualification and beginning commercial shipments represent more than narrative. Samsung is again inside the leading platform’s supply chain. The structural concern lies in whether it can sustain the position across HBM4E and the following generation. Every new stack creates another qualification contest; the industry does not award permanent leadership for one successful generation.

Foundry’s weakness is more structural. Samsung can improve utilisation and win selected customers, but TSMC’s moat consists of thousands of accumulated customer interactions, design tools, packaging options and predictable execution. A durable Samsung foundry rerating requires years of repeat external production. The current improvement is a turn in earnings, not yet a turn in competitive structure.

The market at KRW 230,500 is rewarding some future success but not the full optimistic case. It no longer capitalises the KRW 89.5 trillion quarterly operating-profit print literally; otherwise the stock would appear extraordinarily cheap at approximately five times annualised peak earnings. It does capitalise a lasting HBM franchise and a healthier post-cycle earnings floor than Samsung had before AI memory.

The market is most likely misjudging the distinction between HBM volume visibility and HBM price durability. Long-term agreements and qualification make shipments more visible, but they do not freeze margins. Customer allocation can shift, new generations can change yields and conventional-memory prices can normalise around the HBM business. Investors who treat contracted capacity as recurring software revenue are overstating quality. Investors who treat HBM as ordinary DRAM are understating it.

The most critical one-year variable is HBM4 execution: shipment share, mix, yield and customer breadth. The most critical three-year variable is the supply response, especially whether 2027 fab and packaging additions lower conventional-memory prices faster than HBM profits grow. The most critical five-year variable is capital productivity: whether Samsung turns foundry, HBM and device integration into returns above its cost of capital rather than merely larger assets.

Samsung becomes a better investment under one of two conditions. The first is price: the ordinary shares fall into the KRW 140,000–148,000 zone while HBM qualification and the balance sheet remain intact. The second is proof: HBM share rises above 30%, foundry reaches sustained external-customer profitability, and shareholder returns improve enough to justify a higher conservative valuation.

The research judgment should be overturned positively if Samsung establishes HBM co-leadership across two consecutive generations, reports foundry breakeven for several quarters with a diversified external customer base, and adopts a recurring capital-return policy that reduces excess cash and the preferred-share discount. It should be overturned negatively if HBM share falls below 20–25%, Device Solutions margin drops below 30% before new capacity fully arrives, or foundry losses widen despite advanced-node volume.

The core bull reasons are:

• Samsung’s HBM4 qualification and commercial supply approval convert a previously hypothetical catch-up into a real customer position.
• Device Solutions produced KRW 89.2 trillion of quarterly operating profit and KRW 120.8 trillion of memory revenue, proving that the current upcycle is translating into cash rather than only bookings.
• Approximately KRW 167.6 trillion of net cash allows Samsung to finance HBM, packaging and foundry investment without balance-sheet stress.
• HBM consumes disproportionate wafer and packaging capacity, raising Samsung’s potential mid-cycle memory earnings above pre-AI cycles if it sustains a high-20s or better share.
• Foundry’s HBM base-die and U.S. customer demand give it a credible path toward better utilisation and a smaller earnings drag.

The core bear reasons are:

• A 70% Device Solutions margin is a scarcity outcome and cannot be used as a normalised earnings base.
• SK hynix still holds roughly half of estimated HBM bit output, leaving Samsung dependent on continued qualification and allocation gains every generation.
• Memory capacity additions from Samsung, SK hynix, Micron and CXMT make the company’s shortage-through-2028 claim much less reliable than its 2027 outlook.
• TSMC’s foundry share above 70% and Samsung’s uncertain 2-nanometre yields indicate that the foundry recovery may remain narrow and customer-concentrated.
• The current price exceeds conservative value, while the dividend yield is far below the Korean ten-year government-bond yield, leaving no downside protection under flat earnings.

A first pre-mortem begins in 2027. SK hynix’s M15X expansion, Micron’s Idaho output, Samsung’s own additions and Chinese DRAM capacity ramp close the conventional-memory gap earlier than suppliers expect. Contract DRAM prices fall 35%, NAND loosens, and Samsung HBM share stabilises near 25% rather than moving above 30%. Device Solutions margin falls from 70% to approximately 25%. Normalised EPS drops toward KRW 10,000 and the market applies a 12-times cycle multiple plus cash, producing a share price around KRW 130,000–150,000, a loss of roughly 35–45% from the current close.

A second pre-mortem is technology-specific. During 2027–28, SK hynix and Micron win most HBM4E and next-generation allocations because Samsung’s stack yield or power performance misses customer targets. Samsung’s HBM share falls below 20%. Its 2-nanometre foundry yield remains around the middle of the wide publicly reported range, and external customers keep their highest-value designs at TSMC. Foundry depreciation rises while revenue mix disappoints. Group normalised EPS falls toward KRW 9,000, the ordinary-share P/B multiple compresses toward 1.5 times and the stock approaches KRW 110,000–130,000. The combined earnings and multiple decline could cut the share price roughly in half.

Samsung Electronics remains one of the world’s most capable manufacturing organisations, but its stock is presently an investment in the quality of a normalisation path rather than in the record quarter itself. HBM4 has improved the long-term earnings floor, and the net-cash balance prevents financial distress. The current ordinary-share price already assumes that the HBM recovery persists and that conventional-memory conditions remain healthier than historical mid-cycle averages.

The strongest reason to own the shares is that Samsung may be transitioning from a lagging HBM supplier into a durable co-source while foundry losses contract. The strongest reason to wait is that the shares offer no discount to conservative value and only modest upside to base value. The next evidence should come from HBM allocation, realised mix, foundry customer quality and capital return, not another four-digit comparison with a depressed quarter.

【Company-profile scores】

• Fundamental quality: high
• Growth: medium
• Moat: strong
• Financial soundness: strong
• Management credibility: medium
• Valuation attractiveness: medium
• Risk level: high
• Suitable investor type: cyclical

【Investment rating】

• Rating: Hold
• One-line thesis: HBM4 recovery raises mid-cycle earnings, but KRW 230,500 already prices a durable share gain without conservative-case protection.
• Ideal buy price:

【Ideal Buy Price】140,000–148,000 KRW

Basis: at least a 20% margin of safety below the KRW 185,000 conservative scenario value, provided HBM qualification remains intact and net cash stays above KRW 150 trillion.

• Acceptable hold price: KRW 205,000–275,000, approximately ±15% around the KRW 240,000 base value.
• Clearly overvalued price: KRW 385,000 or higher, at least 10% above the KRW 350,000 optimistic value.
• Current-price classification: acceptable hold.
• Whether to wait for a better price: yes. A new purchase should wait for KRW 148,000 or below with HBM share still at least 25%, or for stronger operating proof that raises conservative value. The opportunity cost is missing a further HBM-led rerating if supply remains tight through 2028.
• Target holding horizon: 3–5 years.
• Expected annualized return: conservative approximately -6%; base approximately 2%; optimistic approximately 16%, assuming a three-year convergence period and dividends near the present rate.
• Max-loss risk: approximately 44–52% if 2027–28 memory supply normalises rapidly, Samsung loses next-generation HBM allocation and foundry remains structurally loss-making.
• Reassessment-trigger signals:
  • Samsung HBM share falls below 25% for two consecutive quarterly industry estimates.
  • HBM4 contributes less than 50% of HBM revenue in the second half of 2026.
  • Device Solutions operating margin falls below 30% before major new industry capacity begins production.
  • Foundry/System LSI losses exceed KRW 2 trillion for two consecutive quarters despite advanced-node revenue growth.
  • Net cash falls below KRW 100 trillion without a corresponding increase in cancelled buybacks, dividends or high-return operating assets.

【Valuation Range】

• current: 230,500 KRW (close as of 2026-08-06)
• bear (conservative · ideal buy zone): [140,000, 148,000]
• base (fair · acceptable hold zone): [205,000, 275,000]
• bull (optimistic · above the clearly-overvalued line): [385,000, 425,000]

Key data synthesis:

Research uncertainties remain material. First, Samsung does not disclose Memory, Foundry and System LSI operating profit separately, so the estimated memory contribution depends on external assessments of logic losses. Second, HBM yields, customer-specific allocation and realised contract pricing are confidential; qualification confirms technical acceptance but not long-run share. Third, foundry yield reports conflict widely and cannot be reconciled from public data. Fourth, the maintenance-versus-growth-capex split is an analytical estimate because the company does not publish it. Fifth, contemporary documentation of the 1975 IPO price and proceeds is not sufficiently reliable for inclusion.

The primary source base consists of Samsung Electronics’ second-quarter 2026 earnings release, divisional presentation and cash-flow materials; the 2025 audited business report and financial statements; shareholder-return and governance disclosures; the company’s IR event archive; competitor filings; and independent market-share, capacity and qualification reporting.

Independent competitive and cycle checks rely principally on TrendForce for HBM and foundry estimates, Reuters for customer qualification, capacity, labour and capital-market reporting, Micron’s official results, and public capacity announcements. Where sources conflict, the report uses ranges and assigns greater weight to company filings for realised financial figures and to independent sources for market share and customer qualification.

Other tickers mentioned

• 000660.KO — principal HBM and DRAM competitor with the current volume-leadership position.
• MU.US — focused U.S. memory peer and third major HBM supplier.
• 2330.TW — benchmark for advanced-foundry yields, ecosystem strength and capital productivity.
• NVDA.US — leading AI-accelerator platform whose qualifications shape HBM allocation.
• AAPL.US — premium-device ecosystem competitor and major display and component customer reference.
• 0981.HK — Chinese foundry competitor affecting mature-node capacity and geopolitical risk.
• WDC.US — NAND and storage-cycle reference.
• ASML.US — critical lithography-equipment supplier governing advanced-capacity expansion.
• AMAT.US — semiconductor-equipment reference for memory and foundry capital spending.
• LRCX.US — memory-process-equipment reference and indicator of wafer-capacity additions.

This report is based on public information and does not constitute investment advice. Markets carry risk; invest with caution.