SK Hynix (KRX: 000660) is the world's second-largest memory semiconductor company by revenue and the global leader in High Bandwidth Memory (HBM). Founded in 1983 as Hyundai Electronics, the company was acquired by SK Group in 2012 and has since undergone a fundamental strategic transformation — from a commodity DRAM/NAND supplier toward a specialized provider of AI-optimized memory architectures. Its FY2025 results cemented that transition: the company generated KRW 97.1 trillion in revenue and KRW 47.2 trillion in operating profit, the highest absolute and margin figures in its 42-year history.
SK Hynix's product portfolio spans two primary segments: DRAM (~70-75% of revenue), which includes HBM, server DDR5/DDR4, mobile LPDDR5X, and graphics GDDR7; and NAND Flash (~25-30% of revenue), dominated by enterprise SSDs under the Solidigm brand (acquired from Intel in 2021). The company operates major manufacturing facilities in Icheon (M14, M16) and Cheongju (M15, M15X) in South Korea, plus two China fabs in Wuxi (DRAM) and Dalian (NAND).
| Metric | FY2022 | FY2023 | FY2024 | FY2025A | FY2026E | FY2027E | FY2028E |
|---|---|---|---|---|---|---|---|
| Revenue | 49.4 | 32.8 | 66.2 | 97.1 | ~190 | ~250 | ~295 |
| Revenue YoY | — | -33.6% | +101.8% | +46.7% | ~+95% | ~+32% | ~+18% |
| Operating Profit | 5.5 | -7.7 | 23.5 | 47.2 | ~125 | ~165 | ~195 |
| OPM (%) | 11.1% | Deficit | 35.5% | 49.0% | ~65% | ~66% | ~66% |
| EBITDA | ~11.0 | ~2.0 | ~37.0 | ~67.0 | ~150 | ~195 | ~230 |
| EBITDA Margin | 22% | 6% | 56% | 69% | ~79% | ~78% | ~78% |
| Net Profit | 5.0 | -9.2 | 13.3 | 42.9 | ~98 | ~130 | ~155 |
| Net Profit Margin | 10.1% | Deficit | 20.1% | 44.2% | ~52% | ~52% | ~53% |
* FY2025A = Audited actuals. FY2026E–FY2028E = consensus-based estimates; wide analyst dispersion exists (bullish outliers at KRW 231T OPM for FY2026). Source: SK Hynix IR, DART, broker consensus.
The KRW 30.9T year-over-year revenue increase in FY2025 (from KRW 66.2T to KRW 97.1T) was not uniform across product lines. Understanding the contribution by segment is critical for assessing both the sustainability of current margins and the risk profile going forward.
| Segment | FY2024 Rev. | FY2025 Rev. | YoY Change | % of Total (FY2025) | Primary Driver |
|---|---|---|---|---|---|
| HBM (within DRAM) | ~KRW 15T | ~KRW 30T+ | +KRW 15T+ (2x+) | ~31% | HBM3E 12Hi volume ramp; NVIDIA H200/GB200 demand; ASP premium sustained |
| Conventional DRAM | ~KRW 31T | ~KRW 38T | +KRW 7T (+23%) | ~39% | Server DDR5 demand surge; supply tightness from HBM capacity diversion; 4Q25 ASP +50% |
| NAND / Enterprise SSD | ~KRW 18T | ~KRW 25T | +KRW 7T (+39%) | ~26% | AI datacenter eSSD demand (H2 recovery); 321-layer QLC product launch; Solidigm contribution |
| Mobile / Consumer DRAM | ~KRW 2T | ~KRW 4T | +KRW 2T | ~4% | LPDDR5X volume growth in flagship smartphones; modest contribution overall |
* Segment-level revenue is not separately disclosed; figures are estimated from SK Hynix earnings calls, IR materials, and broker channel checks. HBM revenue "more than doubled YoY" confirmed by management guidance (Jan 2026).
The key structural insight: HBM now contributes roughly 31% of revenue but is estimated to generate over 50% of operating profit, given ASPs that are 5–10x conventional DRAM and margins well above the corporate average. This concentration cuts both ways — it is the engine of the current supercycle, but it also means any HBM-specific headwind (pricing, share loss, demand slowdown) has an outsized impact on profitability.
Looking ahead to FY2026E, the incremental revenue jump (from ~KRW 97T to ~KRW 190T) is projected to be driven primarily by three forces: (1) HBM volume and ASP expansion as HBM4 ramps for the Rubin platform (~+KRW 50–60T contribution), (2) continued conventional DRAM ASP strength from persistent supply tightness (~+KRW 20–25T), and (3) NAND/eSSD market expansion tied to AI inference infrastructure build-out (~+KRW 10–15T). The operating leverage on this incremental revenue is exceptionally high, explaining why consensus operating profit forecasts more than double YoY.
| Metric | FY2023 | FY2024 | FY2025A | FY2026E | Notes |
|---|---|---|---|---|---|
| Profitability | |||||
| Return on Equity (ROE) | -13.1% | 20.1% | 44.1% | ~52%E | Sharp structural improvement; watch sustainability |
| Operating Margin (OPM) | Deficit | 35.5% | 49.0% | ~65%E | HBM mix-driven; could compress if HBM pricing softens |
| EBITDA Margin | ~6% | ~56% | ~69% | ~79%E | 4Q25 EBITDA: KRW 22.7T (69% margin) |
| Balance Sheet & Leverage | |||||
| Debt Ratio (Debt/Assets) | ~55% | ~31% | 18% | ~22%E | Significant deleveraging in FY2025; CapEx will re-lever |
| Net Cash / (Net Debt) | Net Debt | Net Debt | Net Cash KRW 12.7T | Net Cash (narrowing) | FCF being consumed by aggressive CapEx plans |
| Current Ratio | 1.1x | 1.4x | 1.86x | ~1.6xE | Adequate liquidity; Q3/Q4 cash build notable |
| Cash Flow | |||||
| Free Cash Flow (FCF) | KRW -14.2T | KRW 13.9T | KRW ~20T | KRW ~15T E | FCF yield ~2.4% at current price; compressed by CapEx |
| CapEx | KRW 17.5T | KRW 15.6T | KRW 29T | KRW 35T+ E | Aggressive; M15X + Yongin + Indiana investments ongoing |
| Operating Cash Flow | KRW 3.3T | KRW 29.5T | KRW ~49T | KRW ~50T E | Strong OCF generation; key funding source for CapEx |
| Per-Share Data (KRW) | |||||
| EPS (Basic) | -12,700 | 18,310 | 58,897 | ~134,500 E | Trailing EPS (TTM) as of 2025 = KRW 62,161 |
| DPS (Dividend Per Share) | KRW 1,200 | KRW 1,305 | KRW 3,000 | KRW 1,500 fixed + variable | FY2025 total: KRW 2.1T payout; 25% fixed DPS hike |
| Dividend Yield | — | — | ~0.26% | ~0.13% fixed min. | Low yield; SK Hynix is not an income stock — growth play |
| Book Value Per Share (BPS) | — | ~28,500 | ~43,900 | ~60,000 E | Supports current PBR of ~2.6x at spot |
* FCF = Operating Cash Flow – CapEx. Dividend policy (2025–27): Fixed KRW 1,500/share annually + 50% of cumulative FCF allocated to shareholder returns. Source: SK Hynix IR, GuruFocus, DART.
| Multiple | SK Hynix (spot) | Samsung Elec. DS | Micron (FY26E) | SK Hynix FY26E | Comment |
|---|---|---|---|---|---|
| P/E (Forward) | 8.4x | ~10x | ~9x | ~8.4x | Cheap vs. history; market pricing in cycle peak |
| P/B (Price-to-Book) | ~2.6x | ~1.5x | ~3.2x | ~1.9x FY26E | Justifiable given ROE profile |
| EV/EBITDA | 11.1x | ~8x | ~7x | ~5.5x FY26E | Significant compression ahead if earnings materialize |
| EV/FCF | 25.6x | ~20x | ~22x | ~55x FY26E | FCF inflated by heavy CapEx; less attractive on FCF basis |
| Dividend Yield | 0.26% | ~1.2% | 0% | ~0.13% | Not an income investment |
SK Hynix's ownership structure is anchored by SK Square (formerly part of SK Telecom, spun off in 2021) as the strategic parent, with the remaining ~80% held by a diverse base of institutional and retail investors. Foreign institutional ownership stands at approximately 54%, reflecting high global confidence in the AI memory thesis — but also creating vulnerability to sentiment shifts in global tech equities.
Total shares outstanding: 728,002,365 (all common stock). Source: SK Hynix IR, December 31, 2025.
Governance & Ownership Notes
SK Square's 20.1% stake gives it de facto strategic control, but not majority control. On contentious resolutions, the NPS (~7.7%) and large foreign holders effectively hold the balance of power. This governance dynamic is generally investor-friendly, but it also means SK Hynix's capital allocation decisions — including future M&A, major CapEx commitments, and shareholder return policies — require broad consensus, which can slow strategic pivots compared to founder-controlled structures.
| # | Investment Pillar | Core Claim | Key Evidence |
|---|---|---|---|
| ① | HBM4 First-Mover Advantage | SK Hynix is the exclusive early supplier to NVIDIA's Rubin platform | HBM4 mass production began Feb 2026; Micron ~1 quarter behind; Samsung in validation stage |
| ② | Dual-Engine Revenue Cycle | Both HBM premiums AND conventional DRAM prices are rising simultaneously | Commodity DRAM prices surged ~50% in 4Q25; DDR5 margins approaching HBM3E levels per KB Securities |
| ③ | Balance Sheet Transformation | From leveraged cyclical to net-cash industrial with meaningful shareholder returns | Net cash position; debt ratio 18%; KRW 2.1T FY2025 dividend; KRW 12.2T treasury share cancellation |
The bull case above is widely understood and largely priced in. The more important analytical work is stress-testing the risks. Below are five structural challenges that could materially impair the investment thesis over a 12–24 month horizon.
| Dimension | SK Hynix | Samsung Electronics (DS) | Micron Technology |
|---|---|---|---|
| Financial Profile (FY2025 / FY2025 Equiv.) | |||
| Revenue | KRW 97.1T | ~KRW 180T (DS only) | ~USD 31B (~KRW 43T) |
| Operating Margin | 49.0% | ~30–35% | ~30% |
| Debt Ratio | 18% | ~25% | ~45% |
| HBM Competitive Position | |||
| HBM Market Share (2Q25) | 62% | 17% (recovering) | 21% |
| HBM4 Status | Mass production Feb 2026 | PRA completed; validation stage | Expected 2Q2026 |
| NVIDIA Relationship | Strategic partner; Rubin co-development | Re-certified in 3Q25; rebuilding trust | Approved supplier |
| 2026 HBM4 Share Forecast | ~55% (Daishin) | ~28% (Daishin) | ~17% |
| Strategic Context & Key Risk | |||
| Key Advantage | HBM technology leadership; NVIDIA trust | Scale, vertical integration, NAND leadership | US domestic content; CHIPS Act support |
| Key Risk | Single-customer concentration; export controls | HBM quality credibility rebuild; DS profitability | Technology gap vs. Korean peers |
| FY2026 CapEx | KRW 35T+ | KRW 40T+ (DS) | ~USD 8B (~KRW 11T) |
The competitive dynamic in HBM is shifting. Samsung's HBM3E debacle — which handed SK Hynix a near-monopoly window from 2023–2025 — is unlikely to repeat for HBM4. Samsung's engineering teams have learned from their mistakes, and NVIDIA has strong incentives to qualify at least two suppliers to preserve bargaining power on pricing. Investors should model a gradual SK Hynix share erosion in HBM rather than sustained dominance.
SK Hynix operates a structured shareholder return framework under its 2025–2027 policy, which was revised in November 2024 to reflect the improved balance sheet and profitability. The policy has three components:
| Component | Details | FY2025 Actual | Notes |
|---|---|---|---|
| Fixed Annual Dividend | KRW 1,500/share minimum (raised 25% from KRW 1,200 in Nov 2024); paid quarterly | KRW 1,500/share | Reliable floor; very low yield (~0.13%) at current price |
| Variable / Additional Dividend | 50% of cumulative FCF (after fixed dividend) allocated to shareholder returns | KRW 1,500/share additional (KRW 1T total) | Highly variable; depends on FCF after KRW 35T+ CapEx in FY2026 |
| Treasury Share Buyback & Cancellation | Cancel remaining 50M treasury shares (~2.1% of total, ~KRW 12.2T at current price) | Announced Jan 2026 | Significant EPS accretion; more impactful than dividend in near term |
Important caveat for foreign investors: With FY2026E CapEx of KRW 35T+ and operating cash flow of approximately KRW 50T, the theoretical FCF is only KRW 15T or so — substantially below the headline operating profit. The 50% FCF allocation policy means the variable dividend could be quite modest in FY2026–27 if CapEx runs as planned. SK Hynix is a capital-intensive growth story, not a dividend income story. The treasury share cancellation provides a more meaningful economic benefit than the cash dividend.
Current price: ₩1,170,000 (April 15, 2026). Consensus average: ~₩1,405,000. 65 analysts cover the stock; significant dispersion between bulls and bears reflects genuine uncertainty about HBM cycle duration.
SK Hynix is a genuinely exceptional business operating in a cyclical industry at or near its peak earnings power. The HBM4 technology lead is real, the NVIDIA partnership is deep, and the balance sheet transformation has been remarkable. At 8.4x forward P/E and 5.5x FY2026E EV/EBITDA, the stock is not obviously expensive in absolute terms. The consensus rating remains BUY with a target of ₩1,405,000.
However, investors must honestly acknowledge what they are buying: a high-margin commodity cyclical at or near peak margins, with concentrated customer exposure, significant US-China geopolitical risk, a capital-return program constrained by a KRW 35T+ CapEx commitment, and an increasingly capable competitor in Samsung. The AI memory supercycle is real — but cycles always end. Position sizing should reflect both the exceptional upside potential and the genuine bear case scenarios outlined in this report.
Semiconductors are the heartbeat of the AI era — and I think that era is only just beginning.
Much of what passes for AI debate today is really a debate about semiconductors. The models, the datacenters, the inference chips — none of them exists without memory. And in the memory stack, nothing matters more right now than HBM. That's why SK Hynix isn't just a stock story; it's a front-row seat to one of the most consequential technological shifts in modern history.
AI is not simply a technological revolution; it is increasingly a question of national power. The US, China, and the EU are not investing hundreds of billions into AI infrastructure out of commercial interest alone. They are doing it because whoever leads in AI will have a structural advantage in economic productivity, military capability, and geopolitical influence for decades to come. In that context, the demand for advanced memory semiconductors is not a business cycle. It is a strategic imperative backed by sovereign capital.
That said, I'm not naive about the risks outlined in this report. The share price will be volatile; global liquidity conditions, interest rate cycles and macro risk-off episodes will cause meaningful drawdowns, some of them sharp. The US-China export control situation is a genuine wildcard, and Samsung's competitive response in HBM4 is something to watch closely. These are not small risks.
But here is the core of my thinking: technology leadership compounds. SK Hynix has spent over a decade building the manufacturing expertise, process know-how, and customer trust that gives it the HBM position it holds today. That kind of structural advantage doesn't disappear in a quarter. The companies at the frontier of HBM (SK Hynix being the clearest example) are sitting at the intersection of the most powerful demand wave in semiconductor history and a genuinely high barrier to entry. The volume of global AI investment will continue to sustain revenue and operating income at levels that would have seemed implausible three years ago.
For investors with a long time horizon who can tolerate near-term volatility, I believe SK Hynix represents one of the more compelling ways to gain direct exposure to the AI infrastructure buildout. The price will fluctuate. The demand for memory that powers intelligence will not.
This report is prepared for informational purposes only based on publicly available information, including SK Hynix IR releases, DART filings, broker research summaries, and market data. It does not constitute an offer to buy or sell any security, nor investment advice. All forward-looking estimates are based on publicly available consensus data and carry material uncertainty. Past performance does not guarantee future results. Currency risk (KRW/USD) applies to all non-KRW investors. US-based investors should note SK Hynix trades on KRX (000660); GDR access is via HXSCL (OTC) with lower liquidity. Investors should consult a qualified financial adviser before making any investment decisions.
Sources: SK Hynix Newsroom & IR (news.skhynix.com), DART, Counterpoint Research, Bank of America, Goldman Sachs, Daishin Securities, Hana Securities, KB Securities, Citi, GuruFocus, PitchBook, WSTS. As of April 20, 2026.