Executive Summary
- Samsung Electronics is set to announce approximately ₩40.5 trillion ($26.9 billion) in Q1 2026 operating profit next week — the largest single-quarter profit ever recorded by a Korean company, driven by an "unprecedented supercycle" in memory chips.
- DRAM prices have surged nearly tenfold year-over-year (DDR4 8Gb: $1.35 → $13), HBM revenue has tripled, and conventional DRAM is sold out through 2027 — creating the most extreme supply-demand imbalance in semiconductor history.
- The earnings trajectory is accelerating: analysts project ₩50 trillion in Q2 and ₩60 trillion+ in Q3 as HBM4 shipments ramp, raising a critical question — is this a structural repricing of memory's value, or the peak of a cycle that will inevitably correct?
Chapter 1: The ₩40 Trillion Milestone — Anatomy of a Record
When Samsung Electronics releases its preliminary Q1 2026 earnings next Tuesday, it will mark a watershed in Korean corporate history. At approximately ₩40.5 trillion in operating profit on revenue exceeding ₩120 trillion, Samsung will shatter not just its own records but redefine the scale of what a Korean company can earn in a single quarter.
To put this in perspective: Samsung's Q2 2024 operating profit was ₩4.68 trillion. In the span of nine months, quarterly profits have increased nearly ninefold — a velocity of earnings growth that has no precedent in the global semiconductor industry. The trajectory tells the story: Q2 2024 (₩4.68T) → Q3 (₩12.2T) → Q4 (₩20.1T) → Q1 2026 (₩40T+). Each quarter has roughly doubled the previous one.
The driver is deceptively simple: memory chips have become the world's most valuable commodity. But the mechanics behind this transformation reveal a structural shift that goes far beyond cyclical supply-demand dynamics.
The DRAM Price Explosion
The headline number is staggering: DDR4 8Gb DRAM prices have risen from $1.35 per unit in Q1 2025 to $13 by end of February 2026 — a 9.6x increase. This is not a gradual appreciation. It is a price explosion that has caught even industry veterans off guard.
The cause is a perfect storm of converging forces:
AI infrastructure demand: Every major hyperscaler — Microsoft, Google, Meta, Amazon, Oracle — is building data centers at unprecedented scale. The ₩690 billion ($650 billion+) in combined AI capital expenditure announced by Big Tech requires enormous quantities of memory. A single Nvidia Vera Rubin GPU server rack requires more HBM than an entire data center did five years ago.
Supply-side constraints: Years of cautious capacity expansion during the 2022-2023 downcycle left the industry structurally short. Samsung, SK Hynix, and Micron — the three companies that control over 95% of global DRAM production — all prioritized profitability over volume. Now, with demand exploding, there are simply not enough wafers to go around.
The HBM cannibalization effect: High Bandwidth Memory (HBM) chips command 5-10x the revenue per wafer compared to conventional DRAM. As chipmakers divert wafer capacity to HBM production for AI accelerators, conventional DRAM supply shrinks further, driving prices higher in a self-reinforcing spiral. Samsung has committed to tripling its total HBM volume this month, which means even less capacity for standard DRAM.
War-induced energy costs: The Iran war and Hormuz blockade have added a second-order effect — rising energy costs for semiconductor fabrication, helium shortages threatening EUV lithography, and naphtha prices doubling to $1,100/ton, increasing petrochemical input costs for packaging materials. While Samsung's Korean fabs are partially insulated by nuclear power and strategic reserves, the global supply chain stress has tightened inventories further.
HBM: The Crown Jewel
Samsung's HBM revenue surpassed ₩3 trillion in Q1 — more than tripling from approximately ₩1 trillion in Q1 2025. This surge was driven by dramatically increased supply volumes to Nvidia, Samsung's most important HBM customer.
The breakthrough came in February 2026, when Samsung became the world's first company to mass-produce sixth-generation HBM4. This is not merely an incremental improvement. HBM4, built on Samsung's cutting-edge 1c (sixth-generation) DRAM process, offers 50% more bandwidth and 30% better power efficiency than HBM3E. Critically, Samsung's HBM4 uses 1c DRAM dies while competitors SK Hynix and Micron are still on 1b (fifth-generation) for their HBM products — giving Samsung a technology lead for the first time in the HBM race.
Samsung plans to fill more than half of its expanded HBM capacity with HBM4 starting Q2, positioning itself to dominate both the next-generation AI memory market and the premium pricing tier simultaneously.
With profit margins on both HBM and DRAM approaching 70%, Samsung is projected to generate approximately ₩33 trillion in operating profit from DRAM alone in Q1 — meaning memory chips account for over 80% of the company's total profits.
Chapter 2: The Supercycle — Structural or Cyclical?
Why This Cycle Is Different
The semiconductor memory industry has historically been brutally cyclical. Boom-bust patterns driven by overinvestment, inventory gluts, and demand corrections have defined DRAM economics since the 1980s. Samsung, SK Hynix, and Micron have all endured periods of devastating losses — as recently as 2023, when Samsung's semiconductor division posted its first quarterly loss in 14 years.
But the current supercycle has characteristics that distinguish it from previous booms:
Demand driver durability: Unlike previous cycles driven by PC or smartphone upgrades (discrete, one-time purchase events), AI infrastructure spending is continuous and compounding. Every new model generation requires more compute and more memory. The shift from training to inference — where AI models serve billions of queries daily — creates persistent, growing demand rather than front-loaded purchasing.
Supply discipline: After the brutal 2022-2023 downcycle, all three major DRAM makers adopted capital discipline. Capacity expansion plans are measured and deliberate. Samsung's ₩73 trillion investment plan, announced in March 2026, is heavily weighted toward HBM4 and advanced packaging rather than commodity DRAM expansion. This supply restraint, combined with the wafer cannibalization from HBM, creates structural scarcity.
Oligopoly pricing power: The memory industry has consolidated to just three players controlling 95%+ of the market. This concentration, combined with sold-out capacity through 2027, gives producers unprecedented pricing power. Customers cannot switch suppliers — there are no alternatives.
Geopolitical barriers to entry: China's attempts to build domestic DRAM capacity (CXMT/ChangXin Memory) have been severely constrained by export controls on EUV lithography equipment and other advanced tools. The prospect of a new entrant disrupting the oligopoly has diminished significantly.
The Bull Case: ₩60 Trillion by Q3
The financial investment industry projects Samsung's operating profit will reach ₩50 trillion in Q2 and exceed ₩60 trillion in Q3. The logic is straightforward:
- HBM4 shipments begin in earnest from Q2, with revenue expected to approach ₩10 trillion per quarter by H2 2026
- NAND flash prices have also surged (128Gb: $4.35 → $12.67), with Q1 NAND operating profit expected to triple or quadruple from Q4 2025's ₩2.6 trillion
- Conventional DRAM prices show no signs of peaking, with next year's orders already sold out
- Samsung's market capitalization has held above ₩1,000 trillion ($670 billion) despite the broader market downturn caused by the Iran war
If these projections hold, Samsung would generate approximately ₩150-160 trillion in full-year 2026 operating profit — roughly equivalent to the GDP of a mid-sized country like Hungary or Morocco.
The Bear Case: Echoes of 2018
Skeptics point to historical precedent. The 2017-2018 DRAM supercycle saw Samsung post record profits of ₩58.9 trillion for full-year 2018 — followed by a brutal correction as inventory overshoot and smartphone demand weakness crashed prices. DRAM prices fell 50%+ in 2019.
Could the current cycle follow the same pattern? Several risks exist:
AI ROI reckoning: If hyperscalers conclude that their massive AI investments are not generating sufficient returns — the "Sollow Paradox 2.0" that has been debated extensively — capital expenditure could contract sharply. A NBER working paper published in March found that 90%+ of executives reported AI had produced no measurable impact on productivity.
Energy-driven demand destruction: The Iran war has pushed Brent crude above $100 and global energy prices to levels that are forcing some data center projects to pause. The physical constraint of powering AI infrastructure — estimated at 12-25% of US electricity demand by 2030 — could become a binding ceiling on memory demand growth.
Google's TurboQuant shock: Google's announcement of TurboQuant — a KV cache compression technology achieving 6x memory reduction — briefly wiped $30 billion from memory stocks in late March. While the "Jevons Paradox" argument (efficiency gains increase total usage) ultimately prevailed, it demonstrated that technology breakthroughs can rapidly alter memory demand assumptions.
China's semiconductor self-sufficiency: Hua Hong's 7nm breakthrough and Huawei's Ascend 950PR — achieving 41% Chinese AI chip market share — suggest China is building a parallel semiconductor ecosystem that may eventually develop domestic HBM alternatives, potentially fragmenting the market.
Chapter 3: The Competitive Landscape — Samsung's Redemption Arc
From Laggard to Leader
Samsung's path to this moment has been anything but smooth. Throughout 2024 and into early 2025, the company was widely perceived as having fallen behind SK Hynix in the critical HBM market. SK Hynix was Nvidia's preferred supplier, Samsung's HBM3E faced yield issues, and its foundry business was hemorrhaging money.
The turnaround began in H2 2025 when Samsung resolved its HBM3E quality issues and began volume shipments to Nvidia. The decisive move came in February 2026 with the world-first mass production of HBM4 on 1c DRAM — leapfrogging SK Hynix's 1b-based approach.
Samsung's ₩73 trillion ($49 billion) investment plan, announced March 20, 2026, doubled down on this strategy: heavy investment in HBM4 production at its Pyeongtaek mega-fab, with more than 50% of the facility pivoting to base die production for HBM4 packaging. The company is betting that the convergence of memory and logic — where the base die of HBM stacks increasingly resembles a logic chip — plays to its strengths as the only company with both leading-edge memory and foundry capabilities.
SK Hynix: The Incumbent Under Pressure
SK Hynix remains the world's largest HBM supplier by volume, with a dominant position in HBM3E for Nvidia's current-generation Blackwell platform. Its stock has surged — at one point reaching ₩1.5 million per share, making it one of the most valuable semiconductor companies globally.
But Samsung's HBM4 first-mover advantage creates a genuine threat. SK Hynix's HBM4 product uses 1b DRAM, which Samsung argues is a generational disadvantage. While SK Hynix maintains that its mature 1b process offers superior yields and reliability, the market is watching closely for Nvidia's qualification decisions.
Micron: The American Third
Micron, the sole American DRAM maker, reported Q2 FY2026 revenue of $23.86 billion — a threefold increase — with guidance of $33.5 billion for Q3. Its HBM4 development is on track but trails both Korean competitors. Micron's position has been strengthened by the "reshoring" narrative and CHIPS Act subsidies, but its smaller scale limits its ability to match Samsung and SK Hynix on HBM production volumes.
Chapter 4: Investment Implications — The HALO Trade's Memory Dimension
The Great Rotation Continues
Samsung's record earnings are the latest manifestation of the "Great Rotation" — the structural shift from software to hardware, from bits to atoms, that has defined markets in 2026. While the Nasdaq has entered correction territory (-10%+) and SaaS stocks have been decimated, semiconductor and energy companies have surged.
The HALO trade — Heavy Assets, Low Obsolescence — continues to favor companies with physical infrastructure and pricing power. Samsung, with its ₩40 trillion quarterly profit and 70% margins, exemplifies this thesis.
Key Data Points for Investors
| Metric | Q1 2025 | Q4 2025 | Q1 2026E | Change YoY |
|---|---|---|---|---|
| Revenue | ~₩67T | ~₩95T | ~₩120T+ | +79%+ |
| Operating Profit | ~₩6.6T | ₩20.1T | ~₩40.5T | ~6x |
| DRAM ASP (DDR4 8Gb) | $1.35 | ~$8.50 | $13.00 | +863% |
| HBM Revenue | ~₩1T | ~₩2.2T | ~₩3T+ | +200%+ |
| NAND ASP (128Gb) | ~$2.80 | ~$10 | $12.67 | +352% |
Scenario Analysis
Scenario A: Sustained Supercycle (45% probability)
- DRAM prices remain elevated through 2027 as AI demand continues to outpace supply
- Samsung full-year 2026 operating profit exceeds ₩150 trillion
- HBM4 leadership translates into sustained market share gains
- Trigger: Continued hyperscaler capex growth, no major AI demand correction
- Historical parallel: 1999-2000 DRAM supercycle lasted 6+ quarters before correction
Scenario B: Managed Deceleration (35% probability)
- Prices plateau in H2 2026 as some demand destruction occurs from energy costs
- Samsung full-year profit ₩120-140 trillion — still historically exceptional
- Inventory build begins but remains controlled
- Trigger: AI ROI skepticism grows, some hyperscaler projects delayed
- Historical parallel: 2018 peak followed by gradual 12-month correction
Scenario C: Sharp Correction (20% probability)
- Technology breakthrough (quantum computing, advanced compression) or macroeconomic shock triggers rapid demand collapse
- Prices revert 40-50% within 6 months
- Samsung profit falls to ₩60-80 trillion annualized
- Trigger: AI "winter," severe global recession, or breakthrough in memory alternatives
- Historical parallel: 2019 DRAM crash, 2008 financial crisis
Chapter 5: The Broader Significance — Memory as Strategic Asset
From Commodity to Crown Jewel
The transformation of DRAM from a commodity product to a strategic national asset represents one of the most significant shifts in the global technology landscape. Memory chips are no longer interchangeable components — they are the bottleneck that determines the pace of AI advancement.
This explains why Samsung has maintained a ₩1,000 trillion+ market capitalization even as Korean markets experienced multiple circuit-breaker events due to the Iran war and energy crisis. Investors have recognized that memory — unlike most other industries — actually benefits from the current macro environment: AI demand is accelerating, supply is constrained, and the oligopoly structure prevents price competition.
The Korean Strategic Advantage
Korea's position as the world's memory chip hegemon — Samsung and SK Hynix together control approximately 70% of global DRAM and over 50% of NAND production — has become a core pillar of national economic resilience.
Korea's March 2026 exports reached $80.4 billion, a record high with 48.3% year-over-year growth — the fastest in 40 years — driven almost entirely by semiconductor exports. This occurred even as the Korean economy faced severe headwinds from the Hormuz blockade (Korea imports 60% of its oil through the strait) and broader global uncertainty.
The ₩25 trillion wartime supplementary budget passed by the National Assembly in March was, in a sense, funded by Samsung and SK Hynix's extraordinary profits — the tax revenue from semiconductor earnings providing fiscal ammunition during the energy crisis.
Risks: Concentration and Fragility
Korea's semiconductor-dependent growth model carries inherent risks. The KOSPI's 40%+ weighting in Samsung and SK Hynix creates dangerous concentration. When memory stocks fall — as they did during the Hormuz panic (KOSPI -8.8% circuit breaker) — the entire market crashes.
More fundamentally, the emergence of China's parallel semiconductor ecosystem (Hua Hong 7nm, SMIC 5nm, Huawei Ascend 950PR achieving 41% AI chip market share domestically) suggests that the current oligopoly may face structural challenges within 3-5 years. China's 15th Five-Year Plan explicitly targets 80% semiconductor self-sufficiency — a goal that, if even partially achieved, would fundamentally alter the supply-demand dynamics that underpin Samsung's current earnings.
Conclusion
Samsung's ₩40 trillion quarterly profit is not merely a corporate milestone — it is a marker of a structural transformation in which physical infrastructure, scarce materials, and strategic supply chains have become the defining assets of the AI era.
The memory supercycle reflects the broader "Great Rotation" from bits to atoms: while software companies are being disrupted by the AI agents they helped create (the SaaSpocalypse), hardware companies with physical moats — fabs, clean rooms, proprietary processes — are accumulating unprecedented profits.
For investors, the key question is not whether Samsung will hit ₩40 trillion — that appears certain. The question is whether this represents the new baseline or the peak. History suggests caution: semiconductor supercycles always end. But the structural case — AI demand durability, supply discipline, oligopoly pricing power, and geopolitical barriers to entry — suggests this cycle may have more runway than skeptics expect.
Samsung's redemption from HBM laggard to HBM4 first-mover adds a narrative dimension that the market has yet to fully price. If Q2 and Q3 deliver on the ₩50-60 trillion projections, Samsung may establish itself not just as Korea's champion but as the defining company of the physical AI era.
Sources: Seoul Economic Daily, Reuters, LSEG SmartEstimate, DRAMeXchange, TrendForce, Samsung Electronics disclosures, industry sources
Leave a Reply