SanDisk Looks 'Expensive' at $1,588 (But Isn't?)

📅 Published: 2026-05-11 · 🏢 Samsung / SK Hynix / Micron / SanDisk · 🎯 "Who do you buy in the AI memory supercycle?" 6 to 12 month view

🎯 One-Line Summary

"AI can't get enough memory. All four are flying. Who'll fly highest? Answer: SK Hynix > Samsung > SanDisk > Micron!" 🚀

As of May 2026, all four are up huge in a year. Samsung +420%, SK Hynix +904%, Micron +813%, SanDisk +4,109% (!!). But what matters isn't whether one share looks "expensive" or "cheap"; it's whether market cap divided by revenue (P/S) is fair. By that test, Samsung is the most undervalued and Micron the most expensive.

🏢 What Do They All Do?

One line: all four make "memory chips" for computers, phones, and AI servers. AI demand has driven the biggest boom in their history 🔥

Four-Way Side-by-Side

Most Confusing Point: "Per-share price doesn't matter!"

🤔 Don't get confused: SanDisk's ₩2.32M-per-share equivalent might look like "the most expensive stock," but by market cap it's just one-fifth of Samsung. "Cheap or expensive" should be measured by market cap divided by revenue (P/S).

🚀 What Happened Over the Past Year?

All four exploded, but SanDisk is in another league. 42x in a year is effectively crypto-grade volatility.

🌍 Fundamentals (Record Quarter Across the Board)

Latest Quarter (2026 1Q)

🤯 What 72% Operating Margin Actually Means

Analogy: imagine selling $100 of tteokbokki and keeping $72 as profit. Typical manufacturing runs 5 to 15%, even luxury handbags only hit 30 to 40%. SK Hynix at 72% is unprecedented in manufacturing history.

Why is this possible? SK Hynix has a near-monopoly on HBM (AI memory), so NVIDIA and friends pay whatever the asking price is. It's basically mining gold from the AI infrastructure rush.

🆚 The Most Important Concept: "P/S (Market Cap divided by Revenue)"

What is P/S?

Formula: market cap ÷ revenue = "the value the market puts on every $1 of revenue"

Example:

• SK Hynix: ₩1,389T market cap ÷ ₩210T revenue = 6.62x
• Reading: "To buy SK Hynix outright, you'd pay 6.62 years of revenue"

Benchmarks:

Four-Way P/S, the Key Table

🤯 Shock point: Micron is 2.5x more expensive than SK Hynix on P/S. Same revenue, but the market pays 2.5x more for Micron's stock.

🍰 SOTP, Samsung Is a Special Case

Samsung's P/S of 3.16x "looks the cheapest" because it blends in mobile, consumer electronics, and display alongside memory. Comparing it as a pure memory company would be wrong.

Analogy: you can't apply the same multiple to a single-menu tteokbokki shop (SK) and a tteokbokki-plus-grocery-store-plus-electronics conglomerate (Samsung). You have to value each business separately and add them up. That method is called SOTP (Sum-of-the-Parts).

Samsung SOTP:

• Memory (DS-Memory): ₩194T × 6.62x = ₩1,287T
• System LSI / Foundry: ₩65T × 7x = ₩454T
• Mobile (MX): ₩151T × 1x = ₩151T
• Consumer (CE / SDC): ₩130T × 0.7x = ₩91T
• Total fair market cap: ₩1,982T → fair price ₩332,080 (+16%)

So simple P/S says +200%, but the real answer is +16 to 30%.

🔑 Key: Samsung's memory unit alone trades at an implied P/S of 5.19x, a -22% discount to SK Hynix at 6.62x. The truly undervalued slice is memory.

🚨 PEG Is a Trap in Memory!

PEG = PER ÷ growth rate. Normally below 1.0 means attractive. All four memory names are at 0.03 to 0.15, which on the surface screams "buy."

🚨 It's a trap! Memory is cyclical, so today's +200% growth is one-off. A downcycle that takes EPS down 50% would push PEG from 0.03 to 5.0 overnight. PEG is useful for steady-growth industries like SaaS or healthcare, but basically useless for memory. Don't buy off this number alone.

💰 Where Do Prices Go in 6 to 12 Months?

Averaging revenue growth, P/S normalization, and SOTP models.

Blended Price Targets

🤔 Why is SanDisk only 3rd at +54%?

It's up +4,109% in a year, which makes it dangerously expensive. Fundamentals are strong ($42B backlog, about ₩61T!), but a single quarterly miss can mean -30 to -50%. Volatility bomb, keep the position small.

🤔 Why is Micron the lowest at +6%?

P/S 16.7x is bubble-territory. The AI premium is fully reflected, and mean reversion could mean -20 to -40%. The NVIDIA Vera Rubin exclusive is genuinely positive, but the price has already absorbed it.

💸 ₩1M Invested, What Happens in 1 Year?

Single-name scenarios with no diversification.

Risk-Adjusted Return

Risk-adjusted, SK Hynix and Samsung are safest. SanDisk's expected value is also ₩1.42M, but the range from -20% to +130% is far too wide.

🚦 Traffic Light Summary

🧠 Verdict: How Do You Actually Buy?

🎯 Suggested Weights (100% Memory Allocation Assumed)

✅ Time-Horizon Guide (Current Cycle Temperature 72°C, Peak Approaching)

★ The right holding period depends on where you are in the cycle. Undervalued zone (0 to 50°C) is the long-term entry window; overvalued zone (70°C+) is short-term trading only; past-peak (85°C+) is sell.

• Short-term trading (1 week to 3 months): 🟢 OK, momentum is strong but apply a strict -10% stop-loss. SK Hynix has the strongest momentum (1Y +904%)
• Medium-term hold (3 to 12 months): 🟡 cautious. Substantive models point to +23 to +42% upside, but cycle peak traversal risk is the offset
• Long-term (1 year+): 🔴 not advised right now. The cleaner long-term entry is the next downcycle (30 to 50°C zone, 1 to 2 years away). Buying here means stomaching -30 to 50% on the downcycle
• Dollar-cost averagers: 🟢 buy only 30% now, save 70% for the downcycle. Cut weight immediately if the cycle score hits 8

Match by Stock and Horizon

• SK Hynix: #1 across all horizons (momentum + fundamentals both strong). For long-term, re-enter after the downcycle
• Samsung: medium-term (waiting for HBM4 NVIDIA qualification trigger). Long-term buyers should stagger in on the downcycle
• SanDisk: short-term only (extreme volatility). Long-term buy zone is a pullback to the $1,300s
• Micron: not recommended short-term (16.7x P/S bubble) or long-term (cycle peak)

❌ Not Suitable For

• Investing with borrowed money: hard no
• Can't stomach -30% drawdowns: memory is cyclical, -50% is always on the table
• Looking for dividends: minimal payouts, high volatility
• Long-term value investors trying to enter at the peak: the Buffett approach only works in the 0 to 50°C zone. At 72°C, long-term buyers must be ready to hold through -30 to 50%

🤓 Cycle Industry: The Most Important Warning

Memory Is a Cyclical Industry

Like real estate, it cycles boom and bust every 2 to 3 years. We're currently near the boom peak.

🚦 Cycle Thermometer, Where Are We Now?

Diagnose by counting how many of 10 peak signals are on (each 1 point, score × 10 = temperature).

Current 7/10 = 72°C, "peak approaching, no traversal signal yet" 🌡️

Temperature-Based Action Guide and Expected Price Range

💡 Temperature and price relationship: Temperature is set by how many peak signals are lit, but cycle progression and stock prices typically move together. Below is the expected price range based on SK Hynix (currently ₩1.908M) plus average percent change across all four memory names. Expect ±15% error depending on ticker and timing.

Reading tips:

• Change rates are "additional move from current (72°C basis) to that temperature", not absolute returns
• Other three: apply same ratios to Samsung ₩285K, Micron $783, SanDisk $1,588

What Should You Do?

• 🟡 Be careful with new buys, chasing risk
• 🟡 If you already own, hold for now, but cut weight if signals 8, 9, or 10 flip
• 🟡 Prep to scale out, stagger selling, not all at once
• 🟢 Re-evaluate after the next earnings cycle (July to August 2026), guidance tone is the key

If the score moves to 8, cut exposure 30 to 50% immediately.

👀 What to Watch Next

🤔 Frequently Asked Questions

Q1. If only one, which? SK Hynix. The most stable and the only name where all three models converge in the +40s. AI-era memory #1 status won't waver for the next 2 to 3 years.

Q2. Samsung vs SK Hynix, both Korean, which?

• Short-term, conviction: SK Hynix (HBM #1, already proven)
• Long-term, undervalued: Samsung (rerating if HBM4 NVIDIA qualification clears)
• Suggested split: SK 60 / Samsung 40

Q3. Why is Micron the least attractive despite +813%? Already priced in. Revenue grew, but P/S sits at a bubble-grade 16.7x. For the same revenue growth, you buy SK at 6.62x or Micron at 16.7x. 2.5x worse value than SK.

Q4. SanDisk at +4,109%, too late? Fundamentals are real. NAND supercycle + $42B backlog = strong. But after a +4,109% run, short-term pullback risk is very high. Hold or watch, stagger in if the price retraces to $1,300s.

Q5. How to deploy ₩1M?

SK Hynix:  ₩400K → ~0.21 share (need fractional)
Samsung:   ₩300K → ~1.05 shares
SanDisk:   ₩200K → ~0.09 share
Micron:    ₩100K → ~0.09 share

SK and SanDisk are expensive per share, so you need fractional share buying through Toss Securities or similar. With regular orders, ₩1M won't even cover 1 share of SK.

Q6. What if the cycle ends? Historically -30 to -50%. SK Hynix dropped 60% after 2018. Keep memory below 30% of your portfolio. Always hold non-cyclical industries too (healthcare, consumer, SaaS).

Q7. Does buying all four memory names give you Korea/US diversification? The suggested weights (SK 40 / Samsung 30 / SanDisk 20 / Micron 10) naturally land at about 70% Korea / 30% US, which adds a bit of FX diversification. But they all live in the same industry, so country diversification alone does not neutralize cycle risk. If cycle-end signals flip, you still need to cut all four together.

🚨 Risk Warnings

⚠️ Cycle risk, memory swings boom/bust every 2 to 3 years. Next downcycle could mean -30 to -50%. Keep below 30% of portfolio.

⚠️ SanDisk extreme volatility, beta estimated 4 to 5 (4 to 5x the market). A single quarterly miss can mean -30%. ₩1M could lose ₩100K to ₩200K in a single day.

⚠️ Micron bubble risk, P/S 16.7x is the highest grade memory has ever seen. Mean reversion could mean -20 to -40%. No chasing.

⚠️ This report is not investment advice. Decide based on your own judgment and risk tolerance. Never invest with borrowed money.

📚 Glossary

📌 One-Sentence Conclusion

"The AI memory supercycle blew up all four names, but per-share price is misleading; market cap divided by revenue (P/S) is the real test of cheap or expensive. SK Hynix 40 + Samsung 30 + SanDisk 20 + Micron 10 is the answer. The cycle thermometer is at 72°C (peak approaching); if the score moves to 8, cut exposure immediately!" 🚀

For data, logic, SOTP, and scenario derivations, see the memory four-way analyst report. For comparison with the standout solo name of the season, see the SanDisk +3,685% beginner guide and SanDisk analyst report, the gap between a NAND pure-play bet and a four-name diversified basket becomes obvious. New to stocks entirely? Start with the stock investing beginner guide.

⚠️ This post is educational reference only; investment decisions are your own responsibility. Talk to a parent or guardian and never invest with borrowed money. Memory is cyclical, so the next downcycle could mean -30 to -50%.