Investment Memo · Semiconductor Memory / HBM Leader
Publication date 2026-06-30 | Data as of: Q1 2026 earnings (through 2026-03-31), share price snapshot as of 2026-06-26 close | Currency: Korean Won (₩), 1 trillion = 1조
I. Investment Thesis at a Glance (Tearsheet)
| Investment Rating | Hold (cycle high) |
|---|---|
| 12-month Target Price | ₩2,825,000 (probability-weighted) vs current ₩2,670,000 → +6% |
| Upside / Downside | Upside ₩3,400,000 (+27%) / Downside ₩1,900,000 (−29%) |
| Risk Level | MEDIUM-HIGH: 72% operating margin is an all-time cycle peak; low P/E is a hallmark of cycle tops, not cheapness |
| Time Horizon | HBM4 volume competition (Samsung/Micron entering) → memory pricing trends in 2H 2026 → AI CapEx cycle signals |
| Position Sizing | Max 3-5% of personal portfolio (cycle stock, don't chase ATH, budget by risk) |
| Liquidity / Concentration | #2 weight in KOSPI, ample liquidity; high foreign ownership, linked to Korea market |
| Market Cap | P/E (peak earnings) | Operating Margin | Net Cash |
|---|---|---|---|
| ≈ ₩1,900조 (≈ $1.4T) | ≈ 12x | 72% (unprecedented) | ₩35조 |
| Q1 Revenue Growth | Share | NVIDIA HBM4 Supply | Price vs ATH |
| +198% YoY | ~55–62% | ~54–70% | Near ATH |
Note: Shares ~728m; market cap/P/E from public quotes (USD at ~₩1,335/$). HBM share, target prices are estimated/modeled. P/E based on current peak annualized earnings is misleading for a cyclical; should be viewed against mid-cycle earnings.
1.1 Core Thesis (Probability-Weighted Framework)
| Scenario | Weight | Key Assumptions | Target Price / Change |
|---|---|---|---|
| ▲ Bull | 35% | HBM structural demand makes "this time different," FY27 earnings stay high, slight multiple re-rating | ₩3,400,000 +27% |
| = Base | 40% | Super cycle peaks 2026-27, earnings flatten then decline, P/E stays low (~10x) | ₩2,900,000 +9% |
| ▼ Bear | 25% | Memory cycle downturn (classic), HBM4 competition pressures pricing, earnings halve in downcycle | ₩1,900,000 −29% |
| Weighted Expectation | 100% | 0.35×3.40M + 0.40×2.90M + 0.25×1.90M | ₩2,825,000 +6% |
Target prices are model-derived using scenario-based earnings and cycle P/E. Cyclical stocks should be anchored to mid-cycle earnings; low peak P/E is normal.
1.2 Key Monitoring Indicators (KPIs to Track)
| Indicator | Current Value | Trigger to Reassess | Frequency |
|---|---|---|---|
| DRAM/NAND spot & contract pricing | Uptrend continues | Inflection to decline → cycle peak signal (bearish) | Monthly |
| HBM4 volume and share | HBM4E volume in 2027 | Maintain >50% NVIDIA share → bullish | Event-driven |
| Samsung/Micron HBM4 progress | Samsung targeting >30% share | Competitor ramp and price pressure → bearish | Event-driven |
| Operating margin | 72% (peak) | Stays >60% → bullish; rapid decline → bearish | Quarterly |
| CapEx intensity | Rampant (M15X/Yongin/) | Disciplined spend → healthy; over-investment → cycle risk | Quarterly |
| NVIDIA CapEx / HBM pull | Strong | Upward revision → bullish; slowdown → bearish | Event-driven |
| Inventory & utilization | Low inventory, full utilization | Inventory build → early cycle turn signal | Quarterly |
1.3 Trading Approach (Position Sizing & Hedging · Reference Only, Not Advice)
Not yet positioned: Price near ATH, low P/E on peak earnings — don't chase. Wait for a cyclical pullback or price correction to ₩2,000,000–2,300,000 for staggered entry; portfolio cap 3-5%.
Already positioned (low cost basis): Enjoy the super cycle but set discipline to trim. For cyclicals, "sell when you're making the most money" often beats "wait for a higher price." If prices turn or inventory builds, take profits in tranches.
Hedging: Use KOSPI / Philadelphia Semiconductor correlated tools or single-stock options to hedge cycle downside; non-Korean investors note KRW FX exposure.
Absolutely avoid: Full-position chasing at cycle peak under the illusion of low P/E — the classic "low P/E trap" in memory stocks.
II. Key Facts Snapshot
| Item | Value | Notes |
|---|---|---|
| Ticker / Exchange | 000660.KS · Korea Exchange | SK Group subsidiary; GDR listed in Luxembourg |
| Current Price / ATH | ₩2,670,000 / ₩2,987,000 | As of 2026-06-26; ATH set on 6-25 |
| Market Cap / Shares | ≈ ₩1,900조 (≈ $1.4T) / ~728m shares | Among the largest globally |
| Q1'26 Revenue | ₩52.58조 (+198% YoY, +60% QoQ) | First quarter above ₩50조 |
| Q1'26 Operating Profit | ₩37.61조 (margin 72%) | All-time peak; +405% YoY |
| Q1'26 Net Income | ₩40.35조 (margin 77%) | Net > Operating, includes non-op financial gains |
| QoQ comparison | Q4'25 rev ₩32.83조, OP ₩19.17조 | Q1'25 OP was ₩7.44조 (42% margin) |
| Cash / Debt | Cash ₩54.3조 / Interest-bearing debt ₩19.3조 | Net cash ₩35조 |
| Growth Drivers | HBM, high-capacity server DRAM, enterprise SSDs | HBM demand exceeds capacity plan for 3 years |
| HBM Roadmap | HBM3E mainstream; HBM4E sampling 2H 2026, volume 2027 | Leader in HBM market |
| CapEx | Significantly rising | M15X expansion, Yongin cluster, EUV (ASML) |
| NAND | Solidigm synergy | 321-layer QLC, enterprise SSD |
| Valuation | P/E ~12x (peak annualized) | Cyclical stock; low peak P/E is normal |
III. Investment Summary (TL;DR)
Core call: SK Hynix is the purest winner of the AI memory super cycle — the global HBM leader, NVIDIA's key supplier, with a Q1 operating margin of an unprecedented 72%. But it is still a cyclical stock: this is the cycle peak. A 72% margin is not sustainable, and the ~12x "low P/E" is precisely the typical signature of memory stocks at peak earnings — not cheapness. Rating: Hold. Current price near ATH, do not chase. Wait for cycle signals.
First, its AI position. AI model training and inference are ravenous for memory bandwidth, and High Bandwidth Memory (HBM) is the key to feeding GPUs. SK Hynix is the global HBM leader, with about 55-62% share and capturing ~54-70% of NVIDIA's next-generation HBM4 orders — in other words, behind every AI compute sold by NVIDIA, there is likely Hynix memory. In Q1 2026, revenue hit ₩52.58조, +198% YoY, first quarter above ₩50조; operating profit of ₩37.61조 at a 72% margin is an all-time peak for memory (the previous 2018 peak was ~50%).
专注投资分析、市场洞察与资产配置。不追短期波动,只理解真正驱动长期回报的东西。