4.SK hynix Stock Analysis: Record Earnings, HBM4 Competition, and the Post-Selloff Valuation
SK hynix remains the revenue leader in high-bandwidth memory, but its July 2026 decline exposed a harder question: can HBM4 execution, cash generation, and shareholder returns support the valuation through 2027?
Key Takeaway
The operating thesis remains intact as long as SK hynix converts HBM4 demand into shipments and free cash flow without allowing depreciation, competition, or long-term contract pricing to erode normalized returns. The bottom is therefore more likely to be confirmed by earnings revisions and execution than by one dramatic trading session.
July 29 Snapshot
Price and market capitalization use the July 29, 2026 KRX regular-session close; NXT after-market prices are excluded because they reflect a different session and timestamp. The earnings figures are preliminary, consolidated K-IFRS results and remain subject to review.
Why the Stock Fell Despite Record Earnings
SK hynix reported the strongest quarterly operating result in its history, yet the numbers were below unusually high market expectations. Reported operating profit of KRW 60.5 trillion compared with an LSEG SmartEstimate near KRW 64 trillion, while revenue of KRW 79.3 trillion was also below an estimate near KRW 84 trillion. Slower-than-expected recognition of some HBM4 shipments was one reason the headline result failed to clear the market's hurdle.
What the market disliked
- An earnings miss relative to elevated consensus estimates.
- No detailed, immediate shareholder-return framework.
- 2026 capital expenditure guided to the high-KRW 40 trillion range.
- New common shares underlying the Nasdaq ADS offering.
- Growing evidence that Samsung Electronics and Micron are shipping HBM4.
What did not collapse
- SK hynix still described AI-memory demand as structurally strong.
- The company had completed long-term agreement discussions with about 10 customers.
- HBM4 mass shipments began in the second quarter and were scheduled to ramp in the second half.
- The balance sheet moved to a substantial net-cash position.
- External AI infrastructure indicators remained strong.
The broader Korean market amplified the move
The stock did not fall in isolation. The KOSPI dropped 6% on July 29 after a severe technology-led decline the previous session. Leveraged positioning, forced selling, and a rapid reassessment of AI-related valuations magnified the reaction to the earnings miss. That context matters: part of the decline was company-specific, but part was a market-wide liquidity event.
The Nasdaq offering created a real dilution and flow overhang
SK hynix sold 177.9 million American depositary shares at US$149 each. Every 10 ADSs represent one common share, so the offering required 17.79 million new common shares. Relative to the 712.7 million common shares previously listed, that increased the share count by roughly 2.5%. The capital strengthens funding capacity, but it also changes per-share economics and creates an arbitrage channel between the KRX ordinary shares and Nasdaq-listed ADSs.
Business Model, Earnings, and Financial Position
SK hynix (KRX: 000660) is a KOSPI-listed memory semiconductor producer. Its core businesses are DRAM, including high-bandwidth memory used with AI accelerators, and NAND flash, including enterprise solid-state drives through Solidigm. The company is both an AI infrastructure growth company and a cyclical memory manufacturer; investors must evaluate both identities at the same time.
Q2 2026 preliminary results
| KRW trillion | Q2 2026 | Q1 2026 | Q2 2025 | Change |
|---|---|---|---|---|
| Revenue | 79.3 | 52.6 | 22.2 | +51% QoQ; +257% YoY |
| Operating profit | 60.5 | 37.6 | 9.2 | +61% QoQ; +557% YoY |
| Operating margin | 76% | 72% | 41% | +4 percentage points QoQ |
| Net income | 93.9 | 40.3 | 7.0 | Includes material non-operating gains |
The operating result is the cleaner measure of current business performance. Q2 net income exceeded operating profit because investment-asset gains added KRW 63.3 trillion, with analysts linking much of the gain to the completed sale of the company's Kioxia investment. That makes the quarter's net income unsuitable for simple annualization or a mechanical trailing P/E calculation.
Balance-sheet strength is rising, but so is the investment burden
SK hynix ended June with KRW 88 trillion of cash and cash equivalents and KRW 18.6 trillion of debt, equivalent to approximately KRW 69.4 trillion of net cash. That provides strategic flexibility. However, management also indicated that 2026 capital expenditure would rise into the high-KRW 40 trillion range from KRW 30.2 trillion in 2025.
The relevant long-term test is not whether the company can fund the projects. It is whether HBM, server DRAM, and enterprise SSD demand can keep utilization and pricing high enough to absorb the resulting depreciation. Projects including M15X, the first Yongin cluster cleanroom, advanced packaging capacity, P&T7, and M17 expand the opportunity set, but they also increase the cost base that 2027 earnings must support.
HBM Demand Is Strong, but HBM4 Competition Is Real
The evidence does not support a broad conclusion that AI-memory demand has ended. NVIDIA reported fiscal first-quarter 2027 revenue of US$81.6 billion, with data-center revenue up 92% year over year. Microsoft said Azure revenue grew 40% in its fiscal third quarter and that customer demand continued to exceed available capacity. Those indicators support continued infrastructure spending, although they do not guarantee SK hynix's future market share or pricing.
SK hynix entered the transition from HBM3E with the strongest revenue position. Counterpoint Research estimated that the company held 58% of global HBM revenue in Q1 2026, compared with 21% each for Samsung Electronics and Micron. The leadership is meaningful, but the share was below 69% a year earlier.
| Company | Q1 2026 HBM revenue share | HBM4 status | Investment implication |
|---|---|---|---|
| SK hynix | 58% | Mass shipments began in Q2; second-half ramp planned | Leader, but execution and revenue timing must validate the premium |
| Samsung Electronics | 21% | Commercial shipments and mass production announced in February 2026 | Stronger qualification and yields could pressure share and pricing |
| Micron Technology | 21% | Volume shipment of 36GB 12-high HBM4 announced in March 2026 | A credible third supplier reduces scarcity value over time |
Why HBM4 is harder than the previous cycle
HBM4 competition is not only about stacking DRAM dies. It also depends on logic-base-die design, foundry coordination, packaging yield, thermal performance, power efficiency, customer qualification, and delivery timing. SK hynix said its HBM4 met customer-required speed and power targets and that HBM4E samples were completed in the first half. The investment case improves if those claims translate into accelerating second-half revenue without margin dilution.
CXMT is a different but important risk
China's ChangXin Memory Technologies is not yet the central HBM4 competitor, but it matters in conventional DRAM. Counterpoint estimated CXMT at 8% of global DRAM revenue in Q1 2026, versus 3% a year earlier. If Chinese capacity pressures commodity DRAM pricing, SK hynix would need HBM, server DRAM, and enterprise SSD profitability to offset weaker economics in legacy products.
NAND and enterprise SSDs can reduce dependence on one product
SK hynix highlighted expanding enterprise SSD sales and planned to raise 321-layer NAND to roughly half of domestic NAND production capacity by year-end. Solidigm's high-capacity data-center SSD portfolio is strategically important because AI infrastructure requires both accelerated memory and large, efficient storage pools. A stronger NAND contribution would broaden earnings quality, although NAND remains cyclical and capital intensive.
Why It Matters for Global Investors
SK hynix is more than a Korean semiconductor stock. Its HBM shipment pace is a read-through for AI accelerator supply, advanced packaging demand, and the pricing power of the global memory industry. Strong execution can support customers across the AI hardware ecosystem; delays or faster competitor qualification can change system costs and supplier economics well beyond Korea.
Valuation: What Would Confirm a Bottom?
At the July 29 KRX close, SK hynix had a market capitalization of approximately KRW 1,023.4 trillion. The share-price decline makes headline multiples look lower, but the inputs are unusually unstable. Current earnings include exceptional investment gains, while 2026 consensus estimates assume extremely high memory profitability.
| Metric | Input displayed July 29, 2026 | Mechanical result at KRW 1,401,000 | Why caution is required |
|---|---|---|---|
| Forward P/E | 2026 consensus EPS: KRW 315,006 | About 4.4x | Consensus may move sharply with HBM4 timing, memory prices, and one-time gains |
| Spot P/B | March 2026 BPS: KRW 231,952 | About 6.0x | Backward-looking book value does not yet include the full Q2 earnings and capital raise |
| Forward P/B | 2026 consensus BPS: KRW 482,445 | About 2.9x | Depends on retained earnings, asset values, and the sustainability of extraordinary profitability |
The calculations above divide the July 29 close by the EPS or BPS figures displayed by Naver Finance/FnGuide. They are not company guidance, and provider estimates may update after the preliminary earnings release.
Five conditions that would provide stronger bottom evidence
- HBM4 revenue catches up with shipments: second-half growth should be visible in revenue, not only production announcements.
- 2027 normalized estimates stabilize: operating-profit and free-cash-flow forecasts should stop falling after excluding one-time investment gains.
- Capital spending produces acceptable returns: depreciation growth should remain manageable relative to gross profit and operating cash flow.
- ADS and ordinary-share flows normalize: the conversion relationship should stop creating abnormal selling or valuation gaps across markets.
- Management defines capital allocation: investors need a credible balance among growth investment, balance-sheet reserves, dividends, and potential buybacks.
Catalysts, Risks, and Scenarios
| Potential catalyst | What would confirm it | Downside risk | What would weaken the thesis |
|---|---|---|---|
| HBM4 ramp | Clear second-half volume and revenue acceleration | Qualification or shipment delays | Repeated revenue deferrals or lower pricing |
| Long-term agreements | More customers, deposits, and attractive contract economics | Upside capped by fixed terms | Contract pricing trails spot-market gains |
| Enterprise SSD growth | Solidigm volume, margins, and cash contribution improve | NAND oversupply | Price declines offset unit growth |
| Capital returns | Specific dividend or buyback framework later in 2026 | Cash retained indefinitely | No measurable per-share return despite net cash |
| AI infrastructure demand | Hyperscaler capacity additions and accelerator shipments remain strong | AI spending digestion | Customer capex cuts or inventory accumulation |
HBM4 execution restores confidence
Second-half HBM4 revenue accelerates, long-term agreements protect utilization, enterprise SSD profitability improves, and management announces a disciplined shareholder-return plan. Earnings estimates stabilize despite higher depreciation.
Strong demand, lower valuation premium
AI-memory demand remains healthy, but Samsung and Micron gain share and contract pricing limits some upside. SK hynix stays highly profitable, while the stock requires time to absorb dilution, capex, and reduced expectations.
Peak earnings arrive before peak capacity
HBM4 shipments remain delayed, hyperscalers slow spending, conventional DRAM faces Chinese supply pressure, and depreciation rises into weaker pricing. Consensus earnings fall faster than the share price, making the apparently low P/E misleading.
Evidence before conviction
Monitor HBM4 revenue recognition, Q3 margin guidance, 2027 customer commitments, capex phasing, the shareholder-return announcement, SKHY conversion flows, and updated consensus estimates after the preliminary results are reviewed.
Access for International Investors and Company Information
International investors now have two prominent listed routes. The ordinary shares trade on the KOSPI market under code 000660 in Korean won. American depositary shares trade on Nasdaq under SKHY, with 10 ADSs representing one common share. Prices can diverge temporarily because of KRW/USD moves, depositary mechanics, fees, settlement, and non-overlapping trading hours.
| Item | Verified information |
|---|---|
| Official name | SK hynix Inc. |
| Primary listing | KOSPI, KRX code 000660 |
| U.S. security | Nasdaq ADS ticker SKHY; 10 ADSs = 1 common share |
| Industry | Memory semiconductors: DRAM, HBM, NAND flash, and enterprise SSDs |
| Q2 2026 snapshot | Revenue KRW 79.3T; operating profit KRW 60.5T; net cash about KRW 69.4T |
| Largest shareholder | SK Square; roughly 20.0% after the new-share issuance, based on disclosed ownership and the expanded share count |
Related companies
FAQ
Why did SK hynix fall after record quarterly earnings?
The result was below extremely high consensus expectations, HBM4 revenue recognition was slower than investors expected, and management did not provide an immediate detailed shareholder-return plan. New-share dilution, higher capex, and a broad KOSPI selloff amplified the decline.
Does the decline mean the HBM cycle is over?
No single trading session can establish that conclusion. AI infrastructure demand indicators remain strong, and SK hynix reported continued customer requests and long-term agreements. The more important risk is whether competitors gain share and whether supply growth eventually weakens pricing.
Can U.S. investors buy SK hynix through an ADR?
Yes. SK hynix ADSs trade on Nasdaq under SKHY. Ten ADSs represent one KRX-listed common share. Investors should still consider currency exposure, depositary fees, liquidity, settlement, and price differences between trading sessions.
Why can the forward P/E be misleading?
Memory earnings are cyclical, consensus estimates can change quickly, and Q2 net income included substantial investment gains. A low multiple may reflect peak or non-recurring earnings rather than a durable valuation floor.
What would most clearly weaken the investment thesis?
Repeated HBM4 delays, falling 2027 operating-profit estimates, weaker hyperscaler spending, rapid share gains by competitors, deteriorating free cash flow after capex, or rising depreciation without corresponding revenue would materially weaken the case.
Official and Primary Sources
- SK hynix Q2 2026 preliminary financial results
- SK hynix ADS offering disclosure
- Nasdaq listing information for SKHY
- SEC final prospectus for the ADS offering
- Samsung Electronics HBM4 production announcement
- Micron HBM4 volume-production announcement
- Counterpoint Research quarterly DRAM and HBM market shares
- NVIDIA fiscal Q1 2027 results
- Microsoft fiscal Q3 2026 earnings call
- Reuters report on the earnings reaction and market expectations
This article is an independent analysis based on publicly available information. It is provided for informational purposes only and does not constitute a recommendation to buy or sell any security. Investors are responsible for their own decisions and should consider market, currency, liquidity, tax, and regulatory risks.
댓글
댓글 쓰기