SK Hynix Deep Dive: HBM Hegemon, Cycle Peak, and the Korea Stock vs. US ADR Choice

1. Target Basics: Korean Stock vs. US ADR

SK Hynix trades in two main channels. Get these clear first, so the trade-offs below have a baseline.

Korean KOSPI Common Stock 000660.KS

• Listing: Korea Exchange KOSPI main board (listed since 1996).
• Quote currency: Korean Won (KRW), quoted in "KRW per share."
• Trading hours: 09:00–15:30 Korea time, no after-hours electronic session; for Hong Kong investors that means afternoon trading.
• Liquidity: As a KOSPI heavyweight, daily trading in the Korean shares is active, with deep retail and institutional participation. Before the ADR listing, it was already one of the two largest market caps on KOSPI.
• Access: Requires an overseas brokerage account, or channels such as a "Korean Stock Direct Access" program, a local bank custody/trust account, or an overseas custodian.

US Nasdaq ADR SKHY

• Listing: Nasdaq Global Select Market, listed July 10, 2026, ticker SKHY.
• Structure: Level-3 Sponsored ADR. The ratio is not "1 KOSPI share = 2 ADRs" — the correct ratio is 10 ADRs = 1 KOSPI common share, i.e. each ADR represents 0.1 of a Korean share. This ratio has been unchanged since the July 2026 listing.
• Quote currency: U.S. dollars (USD); no need to convert currency to trade the underlying shares.
• Trading hours: U.S. Eastern time 09:30–16:00, plus pre-market and after-hours; for Asian investors this means "overnight / pre-market" participation.
• Liquidity: At launch, cornerstones such as Baillie Gifford, Coatue, and Situational Awareness anchored the deal. The IPO was priced at $149, raising roughly $2.65 billion; the shares have consistently traded at a notable premium post-listing, reflecting an inverse re-rate of the "Korea Discount."
• Access: Any broker that supports U.S. stocks can buy and sell SKHY directly.

Premium / Discount and FX Risk

In theory, ADR price = Korean share × 0.1 ÷ KRW/USD. In practice, SKHY has frequently traded at a 5–15% premium during Asian-session pre-market windows (i.e. U.S. pre-market) in the early days, and higher at some points. Three structural factors drive this:

1. Korean Convenience Premium: U.S. investors cannot buy the Korean stock during U.S. cash hours, so they are willing to pay up.
2. Settlement / custody friction: T+1 settlement, ADR fees, and FX-hedging costs.
3. Retail sentiment: U.S. forums treat SKHY as a "pure AI Beta" name.

FX risk affects only ADR holders: if you hold the Korean share, your won exposure is embedded directly in the share price; if you hold only the ADR, you carry an extra layer of KRW/USD exposure — won depreciation eats into USD returns, and the reverse also holds.

2. Company Fundamentals

Business Mix

SK Hynix posted 2025 full-year consolidated revenue of roughly ₩97.15 trillion (about US$63.8 billion) and operating profit of ₩47.2 trillion, for an operating margin of about 48.6%. Product mix:

• DRAM: 2025 Q4 revenue about ₩24.9 trillion, ~76% of mix, the largest cash cow. DRAM further splits into conventional DRAM (DDR4 / DDR5) and HBM, and HBM's share within DRAM has already climbed to a record high.
• NAND Flash: 2025 Q4 revenue about ₩7.6 trillion, +59% YoY, reflecting eSSD demand pulled along by AI servers. Management plans to gradually reduce NAND mix and concentrate capex on HBM and advanced DRAM nodes.
• HBM (included under DRAM): 2025 full-year HBM revenue more than doubled, the core engine driving margin expansion.

Industry Position and Competitive Landscape

• HBM market share: Depending on the source, ~56–60% in 2025, still about 58% in 2026 Q1, leadership is solid; Samsung ~20–22%, Micron ~17–20%.
• DRAM market share: 2026 Q1 — SK Hynix ~29% (#2), Samsung 38% (#1), Micron ~24%, China-based CXMT ~8%.
• NAND market share: 2024 — SK Hynix ~21% (#2), behind Samsung.
• Key customers: Nvidia (the largest AI GPU buyer), AMD, Google, Microsoft, Amazon, and other cloud / AI chip vendors. Non-HBM customers include PC OEMs, server ODMs, Apple (mobile NAND), etc.

Most Recent Quarter (2026 Q1)

• Revenue: ₩52.58 trillion (about US$35.5 billion), QoQ +60%, YoY +~140%.
• Operating profit: ₩37.61 trillion; operating margin 72% (a record high), a sharp QoQ jump.
• Net income: ₩40.35 trillion; net margin 77%.
• Capex guidance: On the call management stressed that 2026 capex will be "meaningfully higher than 2025," with the bulk going to the Yongin semiconductor cluster phase-1 fab, the Cheongju (P&T7) advanced packaging plant, and EUV scanner purchases for HBM4 and subsequent HBM4E expansion.

3. Industry Cycle Analysis

Inventory Cycle Position

Memory is a classic cyclical industry. DRAM has followed a "down 2–3 years, up 1–2 years" rhythm over the past two decades. This up-cycle started in the second half of 2024, and across multiple research houses the consensus is that DRAM and NAND spot / contract prices peak in mid-2026, with a possible "high-level chop / initial rollover" after Q3, returning to the mid-stage of the down-cycle from 2027 onward — assuming AI capex stays elevated.

Memory Pricing Trends

• DRAM contract price: 2026 Q1 +90–95% QoQ; Q2 forecast another +58–63%.
• NAND contract price: 2026 Q1 +55–60%; Q2 forecast another +70–75%.
• HBM: 12-Hi HBM4 carries a 20–40% premium versus HBM3e; SK Hynix is now quoting Samsung-equivalent prices on Nvidia bids, marking a new chapter in Korean vendors' pricing power with Nvidia.

AI Server Demand and Profit Impact

AI server pull-through of HBM is the "non-linear amplifier" for the entire memory industry today. A single Nvidia Rubin / Vera Rubin GPU needs 6–8 HBM stacks, so the memory content per GPU is far higher than in a generic server. That makes HBM an extension lever on the "AI capex slope." The moment hyperscaler capex slows (for example, Nvidia data-center revenue YoY growth dropping below 30%), HBM demand weakens fast and gross-margin pressure shows up almost immediately.

4. Stock Price Perspective

Why the Recent Whipsaw

From early 2026 to mid-year, SK Hynix rode a roller coaster:

1. Late January — the Q4 print plus record full-year earnings led the market to reprice HBM's pricing power, and the stock jumped.
2. Late February to early March — KOSPI dropped roughly 10% in a single session, dragging SK Hynix with it; Korean retail showed clear profit-taking pressure along the way.
3. April 23 — the Q1 release, with a record 72% gross margin, sent the stock vertical again; at the same time management signaled "meaningfully higher 2026 capex," which some funds read as near-term depreciation pressure, triggering short-term profit-taking.
4. July 9–10 — the ADR listing at $149; opening-day volatility was wild. The Korean common stock saw pressure as arbitrage desks sold ADRs / bought KOSPI, but the ADR then rallied and re-widened the premium.

Put together, 2026 volatility comes down to a tug-of-war between three forces: (a) HBM fundamentals are extremely strong; (b) "end-of-cycle" psychology at a cycle peak; (c) cross-market arbitrage flows introduced by the new ADR listing.

5. Key Risks

1. Chip price downside risk: If DRAM / NAND contract prices turn down in H2 2026, gross margin compresses within two quarters; if Samsung grabs HBM4 share with aggressive pricing, ASP comes under pressure too.
2. Capex overrun risk: Yongin and EUV tool costs are steep. If AI capex slows in 2027, overcapacity risk emerges, and depreciation / interest expense erodes EPS.
3. Korea-U.S. regulatory and geopolitical risk: HBM sits on the U.S. export-control sensitive list for China. If controls widen or Korea-U.S. friction flares, China business and Nvidia shipment cadence could be affected.
4. FX risk: KRW/USD swings directly affect the ADR's USD quote; won strength is negative for the ADR (compresses USD returns) but neutral for the KOSPI share.
5. ADR float and arbitrage risk: SKHY's float is still thin in the early days, and the arbitrage mechanism is immature. Premiums could widen to unreasonable levels, then snap back sharply once arbitrage channels open or cornerstones unlock, causing violent moves.
6. HBM spec transition risk: Yield or schedule slippage on the HBM4 → HBM4E → HBM5 node jumps would cost lead time on Nvidia's next-gen GPUs.
7. Customer concentration: Nvidia dominates HBM demand; pricing and shipment cadence are heavily tied to a single customer.

6. Conclusion: Korea Stock vs. US ADR Trade-offs

If you want exposure to this HBM hegemon, the decision is really three-dimensional:

Korean Common Stock 000660.KS

• Pros: Deepest liquidity, no ADR fee or premium, tight bid-ask, direct exposure to KOSPI opening prints; most "natural" for Hong Kong-based investors using Futu, IB, or any broker with Korean direct access.
• Cons: Trading concentrates in the Asian afternoon; FX conversion cost on foreign dividend remittance exists; Korean shares get thinner U.S.-side research coverage.

US ADR SKHY

• Pros: Available through any standard U.S. broker; trades on U.S. cash hours; can be managed on the same screen as NVDA / MU / AMD / GOOG and other AI holdings; easy to slot into ETFs.
• Cons: Premium swings wildly early on; float still small; hidden ADR custody and FX-hedging costs; price is sensitive to cornerstone selling and arbitrage flows; in theory the ADR should converge toward the KOSPI share over time, but the timing is uncertain.

Practical Decision Rules

• Short-to-medium term (under six months), want flexible entries and exits, use a U.S. account, AI-beta trading orientation: pick SKHY, but accept the premium swings.
• Medium-to-long term (one year+), care about fundamentals and dividends, want cost and FX exposure under control: pick the KOSPI common stock.
• You can hedge on either side with FX or hedges, but for retail the cost is high.

Before making any choice, investors should verify the latest FX ratio, ADR custody fee schedule, and the prevailing premium, and make sure their own time window and risk tolerance match the high-volatility nature of the name.

The content above is information compilation and objective analysis only. It is not investment advice or a solicitation to buy or sell. Semiconductors and AI memory are high-volatility industries, exposed to cycle, geopolitics, and customer-concentration risk. Any investment decision should be made independently by the reader, who bears full responsibility for the outcome.