Research cut 2026-08-04T23:59:59+09:00. Report figures are model outputs unless explicitly labelled reported or preliminary.
AI memory has reset earnings; the valuation depends on how long scarcity lasts
Samsung's 2026 earnings step-up is being driven by memory pricing and a richer AI product mix rather than a uniform recovery across the group. HBM, server DRAM and enterprise SSDs now carry the near-term estimate, while mobile, displays, appliances and Harman provide scale but much less incremental profit. Our case assumes the memory upcycle remains strong through 2027 and then normalizes; it does not capitalize current scarcity margins as a permanent condition.
Device Solutions generated KRW81.7tn of Q1 2026 Revenue and KRW53.7tn of operating profit. The result was driven by high-value AI products, higher memory prices and constrained supply; it should not be read as a broad 43% group margin across every business.
Samsung began mass-product sales of HBM4 and SOCAMM2 for NVIDIA's Vera Rubin platform and planned HBM4E samples in Q2. Qualification, yield, packaging capacity and customer mix now have an unusually direct path into group earnings.
DX generated KRW52.7tn of Q1 Revenue but only KRW3.0tn of operating profit. Displays and Harman contributed another KRW10.5tn of segment Revenue and KRW0.6tn of operating profit. These businesses broaden cash flow but do not replace the semiconductor thesis.
The model allows peak AI-memory economics through 2027, then reduces EBITDA and net margins as industry supply, customer bargaining and capital intensity catch up. A flat peak-margin extrapolation would produce a higher target but a weaker research case.
Derived model output
| Period | Revenue | Growth | EBITDA margin | EPS |
|---|---|---|---|---|
| FY2026 | KRW 700,000,000.0m | 109.8% | 50.0% | 41,543.03 |
| FY2027 | KRW 820,000,000.0m | 17.1% | 50.0% | 52,731.93 |
| FY2028 | KRW 900,000,000.0m | 9.8% | 48.0% | 55,045.87 |
The rating is a model indication pending analyst approval. It is not personalized investment advice.
The variant view is about conversion, not the existence of demand.
Device Solutions generated KRW81.7tn of Q1 2026 Revenue and KRW53.7tn of operating profit. The result was driven by high-value AI products, higher memory prices and constrained supply; it should not be read as a broad 43% group margin across every business.
Samsung began mass-product sales of HBM4 and SOCAMM2 for NVIDIA's Vera Rubin platform and planned HBM4E samples in Q2. Qualification, yield, packaging capacity and customer mix now have an unusually direct path into group earnings.
DX generated KRW52.7tn of Q1 Revenue but only KRW3.0tn of operating profit. Displays and Harman contributed another KRW10.5tn of segment Revenue and KRW0.6tn of operating profit. These businesses broaden cash flow but do not replace the semiconductor thesis.
The model allows peak AI-memory economics through 2027, then reduces EBITDA and net margins as industry supply, customer bargaining and capital intensity catch up. A flat peak-margin extrapolation would produce a higher target but a weaker research case.
Samsung Electronics operates two broad divisions. Device Solutions houses memory, System LSI and Foundry. Device eXperience includes mobile devices, networks, visual display and digital appliances. Samsung Display and Harman add OLED, automotive and connected-audio exposure. The group combines leading manufacturing scale, large research spending and a net-cash balance sheet with substantial semiconductor-cycle and capital-intensity risk.
| Business line | What it does | Primary earnings driver |
|---|---|---|
| Memory | DRAM, NAND, HBM and enterprise storage products serving AI, server, mobile and other computing demand. | ASP, bit growth, product mix, yield, packaging capacity and customer qualification |
| System LSI and Foundry | Application processors, image sensors and contract semiconductor manufacturing across advanced and mature nodes. | Design wins, advanced-node utilization, yield, wafer pricing and capex |
| Mobile eXperience and Networks | Galaxy smartphones, tablets, wearables and telecom-network equipment. | Shipments, premium mix, component cost, channel inventory and marketing |
| Display, Visual Display and Appliances | OLED panels, televisions, monitors and home appliances, including premium and AI-enabled product lines. | Panel utilization, premium mix, replacement demand, tariffs and input cost |
| Harman | Automotive electronics, connected systems and premium consumer audio. | Automotive production, design-win conversion, order backlog and component availability |
A compact view of how products, capacity, counterparties and risks connect to the financial model.
Revised in v5AI infrastructure is changing the memory mix more quickly than total semiconductor unit demand. HBM, high-capacity server DRAM and enterprise SSDs command different pricing, yield and qualification economics from commodity memory.
A memory shortage can create exceptional operating leverage because fabrication and development costs are high while incremental pricing flows rapidly into gross profit. The same mechanism reverses when capacity and inventory outrun demand.
HBM competition is not decided by nominal capacity alone. Yield, thermal performance, advanced packaging, customer qualification and on-time delivery determine whether announced output converts into premium Revenue.
Samsung's device businesses remain strategically important because they provide scale, internal component demand and ecosystem control. Their lower margins also mean a smartphone or appliance recovery has less valuation impact than an equivalent percentage change in Memory.
| Product or market | Role | Why it matters |
|---|---|---|
| Advanced semiconductor packaging | Advanced chip integration capability | Packaging expands the value captured per AI chip program and can differentiate Samsung's foundry offering, but depends on yield and customer qualification. |
| DRAM | Core volatile-memory portfolio | Industry supply discipline, bit demand, pricing, process migration and product mix make DRAM one of the largest swing factors in semiconductor revenue and operating profit. |
| Enterprise SSD | Datacenter storage portfolio | Hyperscaler capital spending, NAND pricing, qualification cycles, storage density and controller performance determine enterprise SSD volume, mix and margins. |
| Exynos processors and image sensors | System LSI processor and imaging portfolio | External design wins and adoption in Samsung devices affect logic revenue, utilization and the economics of advanced-node development. |
| Foundry services | Contract semiconductor manufacturing | Customer wins, advanced-node yield, wafer utilization and capital intensity determine whether foundry scale contributes earnings or absorbs substantial depreciation and development cost. |
| Galaxy smartphones and devices | Flagship consumer-device franchise | Launch timing, unit shipments, average selling prices, component costs and ecosystem retention directly affect the Device eXperience division's revenue and profit. |
| HBM4 | Fourth-generation high-bandwidth memory | Qualification, yield, capacity ramp and customer supply agreements can materially shift Samsung's memory mix, pricing and semiconductor earnings. |
| High Bandwidth Memory | AI memory product family | HBM mix and qualification can influence memory pricing, utilization and profit per wafer, making it a central link between AI infrastructure spending and Samsung's semiconductor earnings. |
| Image sensors | Mobile and automotive imaging portfolio | Design wins, pixel-performance leadership, handset demand and manufacturing yield affect logic-semiconductor revenue and external customer penetration. |
| Memory semiconductors | DRAM and NAND product portfolio | Industry supply growth, inventories, bit demand, product mix and contract prices drive revenue and create substantial operating leverage in the Device Solutions division. |
| Device eXperience Division | Consumer devices and networks division | Smartphone shipments and mix, component costs, product launches, channel inventory, television demand and appliance replacement cycles are its principal earnings drivers. |
| Device Solutions Division | Semiconductor division | DRAM and NAND pricing, HBM qualification, process yields, foundry utilization and capital expenditure can move Samsung's consolidated profit much more sharply than changes in semiconductor revenue alone. |
| Digital Appliances Business | Home-appliance business | Housing and replacement demand, raw-material costs, logistics and premium-product mix drive appliance sales and margins. |
| Memory Business | Memory semiconductor business | Memory pricing, bit shipments, product mix, yields and industry inventory cycles are major drivers of Device Solutions revenue and operating profit. |
| Mobile eXperience Business | Smartphone and mobile-device business | Premium-device mix, unit shipments, component costs and ecosystem monetization directly influence Samsung's consumer-electronics revenue and margins. |
| System LSI Business | Logic semiconductor design business | Design wins, flagship-device demand and external customer adoption determine scale, while internal Galaxy usage can support volumes. |
| Visual Display Business | Television and display-products business | Panel costs, premium-screen mix, global television demand, replacement cycles and competitive pricing shape revenue, inventory requirements and profitability. |
Public disclosures support customer types, channels and selected counterparties; they do not support a complete customer roster or private contract economics.
| Relationship | Role | Company relevance |
|---|---|---|
| Advanced Micro Devices, Inc. | AI memory and computing partner | Successful qualification and shipment into AMD accelerator and CPU platforms can support high-value memory volumes and create foundry opportunities. |
| Broadcom Inc. | AI memory and foundry collaboration partner | The relationship can create multi-technology revenue across memory, foundry and packaging, subject to product execution and the non-binding nature of an MOU. |
| Asset, input or provider | Role | Operational significance |
|---|---|---|
| Pyeongtaek semiconductor campus | Major Korean semiconductor production site | Capacity additions, equipment installation, yields, power and water availability, and utilization at this site influence semiconductor supply and the return on invested capital. |
| Taylor semiconductor site | United States semiconductor investment | Construction timing, incentives, equipment readiness, customer commitments and process ramp execution determine when the site begins contributing revenue and absorbing depreciation. |
| Period | Revenue | EBITDA | Net income | EPS |
|---|---|---|---|---|
| FY2023 | KRW 258,935,494.0m | KRW 49,715,527.0m | KRW 14,473,401.0m | 2,130.74 |
| FY2024 | KRW 300,870,903.0m | KRW 80,313,430.0m | KRW 33,621,363.0m | 4,950.00 |
| FY2025 | KRW 333,605,938.0m | KRW 97,013,846.0m | KRW 44,260,956.0m | 6,605.00 |
A statistical control and a separate analyst driver model are kept visible.
Revenue = demand x addressable exposure x order conversion x capacity realizationEBITDA = Revenue x EBITDA marginEPS = Net income / diluted sharesFCFF = EBIT x (1 - tax) + D&A - capex - change in working capital| Metric | Selected model | Periods | WAPE | Bias | Interval coverage |
|---|---|---|---|---|---|
| GAAP diluted EPS | damped_median_delta | 15 | 47.1% | -23.3% | 60.0% |
| Net income | damped_median_growth | 15 | 40.8% | -16.3% | 66.7% |
| Revenue | robust_ensemble | 15 | 11.2% | -6.9% | 66.7% |
The control is selected through expanding-window backtesting. Small samples make these diagnostics descriptive rather than conclusive. The annual analyst case uses operating evidence and does not overwrite the frozen control.
| Period | Revenue | GAAP EPS | Model | Basis |
|---|---|---|---|---|
| FY2026 Q3 | KRW 180,000,000.0m | 11,500.00 | Memory-cycle driver bridge | Q2 guidance, server-memory demand, HBM mix, normalized diluted shares and a lower peak-margin assumption than the statistical net-income control |
| Year | Revenue growth | EBITDA margin | Net margin | Capex / revenue | Basis |
|---|---|---|---|---|---|
| FY2026 | 109.8% | 50.0% | 40.0% | 12.0% | Official Q1 results and Q2 guidance, followed by strong but not accelerating H2 memory pricing and AI-product mix |
| FY2027 | 17.1% | 50.0% | 42.7% | 13.0% | HBM, server DRAM and enterprise SSD demand remain strong while 2nm and advanced packaging scale; repurchases lower shares |
| FY2028 | 9.8% | 48.0% | 40.0% | 13.0% | Memory supply response begins to moderate price and margin, partly offset by AI mix and foundry utilization |
| FY2029 | 7.8% | 46.0% | 37.0% | 12.0% | Mid-cycle normalization across memory, with mobile, display, automotive and services supporting group breadth |
| FY2030 | 6.2% | 44.0% | 35.0% | 11.5% | Steady-state case with lower semiconductor scarcity rents and continued capital intensity |
| Year | Revenue | EBITDA | EBIT | Net income | EPS | FCFF |
|---|---|---|---|---|---|---|
| FY2026 | KRW 700,000,000.0m | KRW 350,000,000.0m | KRW 301,000,000.0m | KRW 280,000,000.0m | 41,543.03 | KRW 188,300,000.0m |
| FY2027 | KRW 820,000,000.0m | KRW 410,000,000.0m | KRW 352,600,000.0m | KRW 350,140,000.0m | 52,731.93 | KRW 216,480,000.0m |
| FY2028 | KRW 900,000,000.0m | KRW 432,000,000.0m | KRW 364,500,000.0m | KRW 360,000,000.0m | 55,045.87 | KRW 228,600,000.0m |
| FY2029 | KRW 970,000,000.0m | KRW 446,200,000.0m | KRW 373,450,000.0m | KRW 358,900,000.0m | 55,557.28 | KRW 243,470,000.0m |
| FY2030 | KRW 1,030,000,000.0m | KRW 453,200,000.0m | KRW 375,950,000.0m | KRW 360,500,000.0m | 56,328.12 | KRW 249,260,000.0m |
EPS follows modeled net income and diluted shares without a manual override; the full series is shown in the table above.
The target blends a forward earnings multiple with a lower-weight DCF cross-check.
| Method | Key assumption | Value per share | Weight |
|---|---|---|---|
| Forward P/E | FY2027 EPS 52,731.93 at 6.0x | KRW 316,392 | 95.0% |
| DCF | WACC 9.5%; terminal growth 2.5% | KRW 488,653 | 5.0% |
| Blended target | Rounded under house policy | KRW 325,000 | 100.0% |
| WACC / g | 1.5% | 2.0% | 2.5% | 3.0% | 3.5% |
|---|---|---|---|---|---|
| 7.5% | KRW 587,624 | KRW 629,600 | KRW 679,973 | KRW 741,538 | KRW 818,496 |
| 8.5% | KRW 504,788 | KRW 534,126 | KRW 568,353 | KRW 608,803 | KRW 657,344 |
| 9.5% | KRW 442,697 | KRW 464,143 | KRW 488,653 | KRW 516,934 | KRW 549,928 |
| 10.5% | KRW 394,435 | KRW 410,655 | KRW 428,903 | KRW 449,583 | KRW 473,217 |
| 11.5% | KRW 355,853 | KRW 368,452 | KRW 382,451 | KRW 398,096 | KRW 415,698 |
| Scenario | Probability | EPS factor | P/E | Target | Return | Case |
|---|---|---|---|---|---|---|
| Bear | 25.0% | 0.68x | 5.0x | KRW 179,000 | -25.3% | Memory supply responds faster, HBM qualification or mix disappoints and mobile input costs remain elevated. |
| Base | 55.0% | 1.00x | 6.0x | KRW 316,000 | 31.8% | AI memory demand supports 2026-27 earnings before margins normalize as industry supply expands. |
| Bull | 20.0% | 1.20x | 7.5x | KRW 475,000 | 97.7% | HBM leadership, advanced packaging and foundry utilization improve together while the memory shortage persists longer. |
DCF terminal value and the chosen P/E multiple are assumptions, not observed facts. The rating remains pending analyst approval.
Medium probability / Very High impact. New industry supply, customer inventory or slower AI capex can reduce ASP and utilization while depreciation remains fixed.
Valuation impact: A 500 bp FY2027 net-margin shortfall lowers EPS by roughly KRW6,175 and the 6x P/E indication by about KRW37,000 per share.
Mitigation: A richer HBM/server mix, product leadership and net cash provide resilience, but do not remove cycle risk.
Medium probability / High impact. Delayed customer qualification or poor yield can shift premium demand to competitors and strand expensive capacity.
Valuation impact: A 10% reduction in modeled FY2027 net income lowers EPS and the P/E indication by roughly the same proportion before any multiple change.
Mitigation: Samsung reported HBM4 mass-product sales and continues advanced packaging and process investment; customer concentration remains relevant.
Medium probability / High impact. Advanced-node investment requires high utilization and yield; delays or weak external demand reduce returns and free cash flow.
Valuation impact: A 200 bp increase in capex to Revenue reduces FY2027 FCFF by KRW16.4tn and weakens the DCF cross-check.
Mitigation: HPC design wins, 2nm development and internal semiconductor demand support the roadmap, but economic returns remain execution-dependent.
Medium probability / High impact. Technology restrictions, customer rules and regional incentives can change accessible markets, equipment supply and manufacturing economics.
Valuation impact: The effect can reach Revenue, capex, working capital and the valuation multiple simultaneously; a single deterministic percentage is not defensible without a specified rule and exposure.
Mitigation: Korean and US manufacturing footprints diversify capacity, while advanced semiconductor tools and end demand remain globally interdependent.
| Catalyst | Timing | Evidence to watch | Financial link |
|---|---|---|---|
| Third-quarter 2026 results | October 2026 | Memory Revenue, DS operating margin, HBM mix, inventory and capex | Revenue, margin, EPS and FCFF |
| HBM4E samples and customer qualification | Second half 2026 | Customer approval, shipment timing, yield and packaging capacity | Memory ASP, mix and utilization |
| 2nm and advanced-node ramp | 2026-2027 | Utilization, design wins, yield and external foundry Revenue | Foundry Revenue, margin and capex returns |
| Share cancellation and repurchases | Board disclosures | Shares acquired or cancelled and remaining authorization | Diluted shares and per-share value |
No separate event-intelligence layer was frozen for this initiation. Reported results and disclosed operating developments are reflected only through the dated source ledger and analyst assumptions.
The model gives priority to balance-sheet capacity, reinvestment needs and cash conversion. Net cash is added in DCF, but only once; it is not also capitalized through the earnings multiple.
Expansion capex is explicitly deducted from FCFF in the year it is expected to be spent. A project can be strategically attractive and still reduce near-term free cash flow.
Reported actuals, research-cut dates, evidence hashes and approved ratings are outside the chatbot edit surface. Narrative and assumptions may change only through a validated, versioned patch.
Private contracts, complete customer rosters and undisclosed order economics remain unknown rather than estimated as facts.
| Company | Why included | How used |
|---|---|---|
| SK hynix | Memory and HBM peer | HBM execution, memory pricing and cycle comparison |
| Micron Technology | Global memory peer | DRAM/NAND pricing, capex and supply response |
| TSMC | Foundry benchmark | Advanced-node utilization and manufacturing economics; business mix differs |
| Apple | Premium device ecosystem benchmark | Premium mobile demand and ecosystem economics; semiconductor model differs |
Reported: official filing or exchange disclosure. Derived: deterministic calculation from reported figures. Forecast: model output using disclosed assumptions. Judgment: analyst interpretation. Scenario: bounded alternative, not a prediction.
| Source | Date | Use | Status |
|---|---|---|---|
| Samsung Electronics Q1 2026 earnings presentation | 2026-04-30 | Latest complete official financial, segment and outlook presentation | available |
| Samsung Electronics Q2 2026 earnings guidance | 2026-07-07 | Official Q2 Revenue and operating-profit guidance | available |
| Samsung Electronics Q4 and FY2025 earnings presentation | 2026-01-29 | FY2025 results and comparative segment data | available |
| Samsung Electronics 57th AGM results | 2026-03-18 | Audited FY2025 totals, EPS and dividends | available |
| FMP structured Samsung market and financial snapshot | 2026-08-04 | Price, normalized history, provider estimates and provisional Q2 components | available |
| Argus Samsung point-in-time statistical forecast control | 2026-08-04 | Expanding-window Revenue/EPS backtest and Q3 control | available |
| Argus verified Samsung Electronics company-twin release | 2026-07-26 | Evidence-backed products, facilities, partners, risks and relationships | available |
| Samsung begins commercial HBM4 shipments | 2026-02-12 | First-party HBM4 product imagery and product-status evidence | available |
| Samsung and AMD expand AI-memory collaboration | 2026-03-18 | First-party partner and HBM4 deployment evidence | available |
| Samsung unveils the Galaxy S26 series | 2026-02-26 | First-party mobile product imagery | available |
| Samsung semiconductor operations photo essay | 2017-11-10 | First-party semiconductor manufacturing and campus imagery | available |
| HARMAN agreement to acquire ZF's ADAS business | 2025-12-23 | First-party automotive technology and portfolio evidence | available |
| Customer allocation, HBM pricing, yields and foundry contracts | Known data gap | known_gap |
The point-in-time statistical baseline is frozen before event scenarios. Annual estimates use an explicit revenue-to-margin-to-EPS bridge. Events do not change the control forecast automatically; they create reviewable low/high deltas and require promotion into a new immutable version.