A Quarter That Dwarfed the Last Three Years Combined
Samsung's Q2 FY26 preliminary results landed well above the elevated bar that analysts had set. The world's largest memory maker turned a ₩4.7T operating profit a year ago into ₩89.4T — a number that eclipses its combined earnings for the prior three years.
Q2 FY26 Operating Profit vs Consensus
The beat came despite bonus-related provisions under the May wage deal — without those charges, analysts estimated operating profit would have exceeded ₩100T. Revenue more than doubled to ₩171T, also ahead of expectations. But these are preliminary figures; Samsung will not release net income, divisional breakdowns, or cash-flow data until the full report around July 30. That leaves investors filling in the blanks — and the blanks, particularly around foundry losses and bonus provisioning, may explain some of the market's hesitation.
The Micron Template, Repeating in Seoul
Samsung beat across the board, but the stock fell anyway. That sentence captures the central puzzle — and it is a puzzle the market already solved once this quarter, with Micron.
The selloff is less about Samsung's numbers and more about where the stock was going into the print. Samsung had rallied 165% YTD, and the five trading days before the announcement had already seen a sharp drawdown — from ₩362,500 on June 18 to ₩309,500 by July 3 — as investors de-risked into the event. The actual −7% drop on July 7 was the final leg of a pre-positioning unwind, not a sudden repricing of fundamentals.
"Buy the Rumor, Sell the Fact" Mechanics
The AI memory super-cycle has been the dominant narrative for months. DRAM prices rising 44% QoQ and NAND 53% QoQ were tracked in real time by every sell-side desk. By the time Samsung made it official, the trade was crowded.
The Micron Template
On June 25, 2026, Micron stock hit a record $1,254 after reporting Q3 FY26 results that were staggering — revenue $41.5B (4× YoY), adjusted EPS $25.11, gross margin ~85%. Two weeks later, it was trading at $975, a 22% decline from the peak. The pattern — extraordinary delivery, already priced in, then rotation out — is now repeating at Samsung with near-identical mechanics.
The Sustainability Question
Counterpoint Research flagged that the three major memory makers are averaging 75–80% operating margins — levels its report called "excessive profiteering." The implicit warning: if margins stay here, regulatory scrutiny or a capacity arms race that eventually crushes pricing becomes a real tail risk.
Post-Earnings Stock Performance: Samsung vs Micron
The selloff is not a verdict on Samsung's execution — it is a verdict on how much good news the stock price had already absorbed. For investors, the lesson from Micron is that post-earnings drawdowns in this cycle can run deep and last weeks, even when the numbers are historic.
In the five sessions before July 7, Samsung fell from ₩362,500 (June 18 close) to ₩309,500 (July 3 close), a 14.6% pre-event decline. The pattern — sell into strength ahead of a known catalyst — mirrors Micron's pre-earnings volatility. SK Hynix followed a similar trajectory, falling from ₩2,521,000 to below ₩2,000,000 over the same window.
The Memory Engine: Structural, Not Cyclical
Samsung's Q2 revenue doubled year-on-year to ₩171T, and the driver is unambiguous: memory. The AI buildout has created a demand shock that the industry's supply chain cannot absorb, pushing prices higher across every memory category — not just the high-bandwidth memory (HBM) that goes into AI accelerators, but conventional DRAM and NAND as well.
DRAM & NAND Price Increases (QoQ %)
The price surge has a structural underpinning that distinguishes it from historical memory cycles. Nvidia CEO Jensen Huang and OpenAI COO Brad Lightcap have both publicly warned that memory supply is the key bottleneck for AI development. When the two most important companies in AI both say they cannot get enough of what you make, your pricing power is not cyclical — it is structural.
Manufacturers are prioritizing HBM production to meet data-center demand, which in turn constricts supply of conventional memory used in smartphones, PCs, and enterprise servers. The result is a rising tide that lifts all memory boats: even products that were commodity-priced two years ago now command premiums. Analysts expect shortages to persist through at least 2027.
YTD Stock Performance: Samsung vs SK Hynix
Samsung's 165% YTD gain would be extraordinary in any other context — but it trails SK Hynix's 260% surge. The gap reflects portfolio composition: SK Hynix is a pure-play memory bet, while Samsung carries a consumer-electronics division and a foundry business that is losing money. The foundry and logic-chip (LSI) segments are expected to report widening losses in Q2, partly because bonus expenses under the May wage deal are allocated across the entire semiconductor division.
Samsung will release divisional results around July 30. Based on analyst estimates cited in the preliminary coverage: Memory is expected to account for the vast majority of operating profit (₩80T+ range), Mobile (MX) likely contributed a solid quarter, while Foundry/Logic (LSI) losses may have widened due to bonus provisioning and competitive pressure. These are analyst estimates pending official disclosure.
The $70 Billion Bet
Samsung has announced plans to invest over $70 billion in production-capacity expansion and R&D in 2026 alone. Together with SK Group, the two Korean chipmakers plan to build four new fabrication plants in the country's southwest — a combined 800 trillion won investment aimed at doubling South Korea's memory production capacity within five years. Samsung separately committed to 2,100 trillion won in domestic investment through 2040.
Samsung Annual Capex Trajectory
Bull Case: Confidence in Structural Demand
- $70B in a single year signals conviction that AI-driven memory demand is structural, not cyclical
- Customers are increasingly seeking longer-term supply agreements, reinforcing elevated pricing
- Building new fabs takes years — supply cannot respond quickly even if demand continues surging
Bear Case: Seeds of Overcapacity
- Memory historically defined by boom-and-bust cycles — the larger the boom, the harder the correction
- Concurrent Samsung + SK expansion risks flooding the market when new capacity comes online
- Meta is developing its own AI cloud infrastructure — a potential signal of demand fragmentation
The Samsung-SK 800 trillion won joint expansion plan is a national project. South Korea aims to double its memory production capacity within five years, positioning itself as the indispensable manufacturing hub for the AI era. The four new plants — two by Samsung Group, two by SK Group — are concentrated in the country's southwest. While the strategic logic is clear, the scale of concurrent investment raises questions about execution risk, labor availability, and power infrastructure.
The Message From the C-Suite
While Samsung's formal earnings call is expected alongside the full report on July 30, the preliminary release and accompanying media coverage already sketch the management narrative. The message from the C-suite is confidence: demand is structural, the capex plan is necessary, and the pricing environment remains favorable. But analysts are already sharpening their questions around margin sustainability and whether 75–80% operating margins in memory are politically and competitively durable.
"Demand is so strong that they are trying to ship more products to their server customers, which typically tends to carry higher margins." — Sanjeev Rana, CLSA Securities Korea. The suggestion — that Samsung may be asking for significant price increases from customers like Nvidia — speaks to extraordinary bargaining power. The counterpoint, from Counterpoint Research, is that such margins "constitute excessive profiteering" and could draw regulatory attention.
The Memory Makers: Who Wins, Who Signals
Samsung does not operate the AI memory super-cycle alone. Three players dominate the market, and each tells a different part of the story.
| Company | Ticker | Key Metric | YTD Return | Key Dynamic |
|---|---|---|---|---|
| Samsung Electronics | 005930.KS | ₩89.4T OP (+19×) | +165% | Broad portfolio; memory + foundry drag |
| SK Hynix | 000660.KS | Pending (est. strong) | +260% | Pure-play HBM leader; highest AI leverage |
| Micron Technology | MU | $41.5B Rev (+4×); EPS $25.11; GM 85% | −22% from peak | US pure-play; post-earnings selloff template |
Micron: The Template
Micron's fiscal Q3 2026 (ended May 28) was enormous — revenue quadrupled YoY to $41.5 billion, EPS hit $25.11, and gross margin reached a company-record 85%. The stock surged to an all-time high of $1,254 on June 25, the day results were released. Two weeks later, it traded at $975. The 22% decline from peak is the cautionary template now hanging over Samsung: extraordinary numbers do not guarantee sustained stock appreciation when expectations have run far ahead of even historic results.
SK Hynix: The Pure Play
SK Hynix's 260% YTD surge reflects its status as the highest-beta play on AI memory. Unlike Samsung, SK Hynix lacks a consumer-electronics division or a struggling foundry business to dilute its earnings. Its premium over Samsung captures the market's preference for pure-play exposure in a sector where focus matters.
Nvidia: The Customer and the Signal
Nvidia's Jensen Huang has been explicit: memory is the bottleneck. When the most important company in the AI supply chain says your product is the constraint on its growth, your demand is structural. Nvidia's HBM orders from Samsung and SK Hynix are a direct transmission mechanism from AI model training to memory pricing.
Where We Stand After the Selloff
Samsung's −7% post-preliminary drop lands in familiar territory: the Micron pattern suggests post-earnings drawdowns in this memory cycle can run 15–22% from peak and last two to three weeks. If Samsung follows the same trajectory from its June 18 high of ₩362,500, the stock could find support between ₩285,000–₩310,000. The July 30 full report is the next major catalyst — divisional breakdowns, net income, cash flow, and formal guidance will either validate the structural-demand thesis or feed the sustainability doubt.
Bull Case
- Memory ASPs continue rising through H2 2026, driven by HBM demand and conventional supply tightness
- July 30 final report shows foundry losses stabilizing and memory margins holding above 70%
- Micron finds a floor and the "buy-the-rumor, sell-the-fact" cycle resets
- AI hyperscaler capex guidance confirms sustained investment through 2027
Bear Case
- Memory margins compress as capacity expansion starts to outpace demand growth
- Regulatory scrutiny on 75–80% memory margins materializes (Korea, US, or EU)
- Micron continues to slide, dragging sector sentiment
- Meta's self-built AI cloud signals a fragmentation of demand
- July 30 full report reveals foundry/LSI losses larger than anticipated
What to Watch
- July 30, 2026: Samsung full Q2 report — divisional breakdown, net income, FCF, formal guidance
- DRAM/NAND monthly contract prices: Track whether the 40–53% QoQ pace holds or decelerates
- Micron (MU) price action: A floor in MU would signal the sector selloff is exhausting
- Nvidia earnings (Aug 2026): HBM demand read-through from the ultimate AI bellwether