SK Hynix (000660) Deep Dive: The Incremental Function of an Earnings Supercycle — Q2 Momentum, HBM Dominance & the ₩1,500,000–₩1,700,000 Debate

By Analyst J | Capitalsight.net

Executive Summary: SK Hynix stands at the most extraordinary profitability inflection in the history of the global DRAM industry. The company is on track to post Q1 2026 operating profit of approximately ₩35.6–₩37.5 trillion — nearly doubling the ₩19.2 trillion recorded in Q4 2025 — and two domestic strategy estimates project full-year 2026 operating profit in the range of ₩184.7–₩249.3 trillion, followed by ₩238.2–₩381.9 trillion in 2027. With the stock trading at ₩1,128,000 against a current-year PER of 3.9x–5.4x and forward PBR of 2.5x–3.1x, the market is applying a historically compressed valuation multiple to what may be the most structurally durable memory upcycle ever recorded — driven not by conventional PC refresh cycles but by the insatiable bandwidth demands of Agentic AI infrastructure. The primary catalyst cluster for Q2–Q3 2026 includes DRAM/NAND ASP surges of approximately +40%/+38% quarter-over-quarter, the formalization of an American Depositary Receipt (ADR) issuance, shareholder return policy reconfirmation, and the imminent execution of binding long-term supply agreements (LTAs) with hyperscale customers.

Analyst J's Key Takeaways

• Investment Moat: SK Hynix possesses the only commercially mature, high-volume HBM3E production capability in the world, with TSV utilization running at 100% since Q4 2025 and HBM ASP commanding a 127%–173% premium over conventional DRAM on a $/GB basis. This technological lead — hardened by years of co-engineering partnerships with Nvidia and the inherent complexity of TSV stacking yield management — constitutes a structural supply-side moat that cannot be replicated by Samsung or Micron within the next 12–18 months at equivalent scale and yield.
• Primary Catalyst: The Q2 2026 earnings inflection is the most important near-term catalyst. With DRAM ASP projected to increase approximately +40% QoQ (on top of the +69% increase in Q1) and NAND ASP rising approximately +38% QoQ, the company's operating profit is forecast to surge from ₩35.6–₩37.5 trillion in Q1 to ₩49.5–₩61.4 trillion in Q2, establishing a new all-time record for a single quarter. This ASP acceleration is being driven by tightening supply across all memory categories as AI-driven server buildouts absorb available capacity, while customers have not yet begun speculative inventory accumulation — suggesting the demand is genuine and the supply constraint is structural rather than demand-pull.
• Consensus Target: Domestic strategy estimates diverge materially: the more bullish view targets ₩1,700,000 (50.7% upside from ₩1,128,000) anchored to a 2026 PBR of 3.80x against a projected ROE of 94.1%; the more conservative view targets ₩1,500,000 (33.0% upside) using a SOTP framework that values the HBM business at 18.6x 2026 after-tax profit (PEG 0.5x) and the Non-HBM business at 2.1x adjusted book value. Analyst J views both targets as achievable within a 12-month horizon, with the ₩1,500,000–₩1,700,000 range representing the appropriate fair value band and the bullish target as plausible if the 2027 earnings upgrade cycle continues at its current velocity.

The Core Thesis: Why This Stock Now?

The investment case for SK Hynix in April 2026 is ultimately an argument about the relationship between incremental supply additions and incremental demand — what domestic strategy research has aptly characterized as "the incremental function." In conventional memory cycles, demand is the variable that surprises; this cycle is structurally different because the supply side is the constraint. Global DRAM bit supply growth for 2026 is tracking at roughly 17%–22% year-over-year, which is historically modest. But AI server DRAM demand — particularly for High Bandwidth Memory — is growing at a compound rate that makes any supply addition immediately insufficient. Nvidia's AI server platform is absorbing an estimated 1,689GB of HBM per server box in 2026, up 60% from 1,056GB in 2025, and AI server unit shipments are growing at 18% year-over-year to an estimated 1,754K units. The arithmetic is unambiguous: incremental HBM demand is consuming every bit SK Hynix can produce, and the company's TSV (Through-Silicon Via) utilization rate has been running at 100% since Q4 2025 with no near-term relief on the supply side.

The deeper structural point is that the driver of this cycle is not a transient demand event — it is the architectural evolution of AI from inference to Agentic AI. In a world where AI models are running continuous, multi-step reasoning loops rather than single-shot inference, the memory bandwidth requirements per GPU cluster scale non-linearly. Contemporary AI infrastructure, increasingly built around both GPU-centric and CPU-centric compute nodes, requires a magnitude of low-latency, high-bandwidth memory that makes HBM's addressable market materially larger than street consensus was estimating as recently as six months ago. Domestic strategy research notes that 2027 server DRAM demand estimates have been subject to continuous upward revision as Agentic AI deployment accelerates — a pattern that has historically preceded multi-year earnings revision cycles of the kind SK Hynix is now entering.

The Q2 2026 narrative is particularly important because it is not simply about beating a single quarter's consensus estimate. What makes Q2 analytically significant is the convergence of multiple independent value-creation events within a single quarter: the ASP inflection across both DRAM and NAND, the ADR issuance process reaching deal roadshow stage with reported positive feedback from U.S. institutional investors, the reconfirmation of shareholder return commitments at the general meeting, and the anticipated execution of binding LTA contracts that will provide multi-year revenue visibility. Each of these events, taken individually, would constitute a meaningful re-rating catalyst. The fact that they are converging simultaneously in Q2 creates a compound momentum effect that domestic strategy analysts describe as a "sequential, multiple stock price momentum" scenario. The ADR, in particular, is structurally important: it introduces SK Hynix to the U.S. institutional investor universe that currently underweights Korean semiconductor names due to market access frictions, and it will force a re-evaluation of the stock's global peer group from Korean KOSPI context to the Micron/Nvidia/TSMC competitive landscape.

Perhaps the most underappreciated element of the current thesis is the competitive landscape within the customer base. Buyers of SK Hynix's memory products — particularly in the AI server segment — have not yet initiated speculative hoarding or double-ordering. This is an unusual condition in a memory upcycle. Historically, rapid ASP appreciation is accompanied by panic buying that inflates demand readings and sets the stage for an eventual inventory correction. The current cycle, as characterized by domestic research, involves customers who are receiving precisely the memory they need for active deployments, without building excess safety stock. Emergency "5x" (penta) and "6x" (hexa) ordering patterns — where customers queue multiple orders for the same product to secure allocation — are expected to emerge in Q2–Q3 2026 as supply tightness becomes acute. When speculative demand layering begins, it will amplify ASP upside beyond current forecasts. The implication is that the current trajectory, which already implies staggering profitability, may still understate the peak of this upcycle.

Competitive Position & Business Segments

SK Hynix's business divides into two primary segments: DRAM and NAND. Within DRAM, the most strategically important sub-segment is High Bandwidth Memory (HBM), which has been the company's defining competitive differentiator since 2024. Understanding the HBM supply chain architecture is essential to appreciating why SK Hynix's competitive position is not merely a temporary first-mover advantage but a compounding structural lead.

HBM production requires a manufacturing process fundamentally different from conventional DRAM. Instead of selling a flat DRAM chip, HBM requires stacking multiple DRAM dies vertically and connecting them through thousands of TSV copper interconnects, then packaging the stack on a silicon interposer alongside the GPU die. The yield management challenges of this process — maintaining electrical integrity across 8–12 die stacks while achieving commercially viable defect rates — have proven significantly more difficult to solve at scale than Micron and Samsung initially anticipated. SK Hynix has been producing HBM3E at commercial volumes since early 2024, while Samsung's HBM3E quality issues delayed its acceptance into Nvidia's supply chain and Micron is scaling HBM3E production but from a lower base. Based on domestic research estimates, SK Hynix's blended HBM maximum yield has improved from 63.1% in Q1 2025 to 82.4% in Q1 2026, and its TSV wafer capacity is expanding from 450K wafers/quarter in Q4 2025 to 600K wafers/quarter by Q4 2026 — a 33% capacity increase on an already fully utilized base.

The HBM ASP premium is the financial expression of this competitive advantage. According to domestic research analysis, HBM ASP is projected at approximately $15.2–$16.4/GB in 2026, compared to conventional DRAM ASP of approximately $8.1–$9.5/GB in the same period — an HBM Advantage of 127%–133% premium. This premium has been remarkably stable despite Samsung's attempts to re-enter the Nvidia supply chain and Micron's increasing HBM3E production, which suggests that supply constraints for qualified HBM remain binding and customer switching costs (extensive co-engineering, yield qualification processes) are high. One domestic strategy estimate projects HBM revenue reaching approximately $39.9 billion USD in 2026, representing approximately 28.8% of total DRAM revenue — a counterintuitive decline in HBM revenue share as conventional DRAM ASPs surge so dramatically that they grow faster than HBM volumes.

The NAND business has undergone an equally dramatic transformation. Historically a drag on consolidated margins — NAND operating margins were essentially zero through much of 2025 — the segment is now generating 48%–57% operating margins as ASP recovery has been swift and production discipline among major suppliers has improved. The supply side for NAND has been constrained by the industry's reluctance to invest in new manufacturing capacity during the 2022–2023 downturn, and SK Hynix's NAND revenue is projected to grow from ₩21.2 trillion in 2025 to ₩80.4 trillion in 2026 under the more bullish domestic estimate — a nearly 4x increase that reflects both volume growth and explosive ASP appreciation. NAND is no longer a margin dilutant; it is becoming a significant earnings contributor in its own right.

Against global peers, SK Hynix's operational profile is exceptional. The company achieved commodity DRAM operating margins of approximately 84% in Q1 2026 — described by domestic research as "industry-highest levels" — a figure that reflects not just pricing power but the efficiency of a mature manufacturing base amortized against dramatically higher ASPs. By comparison, Micron (MU), the closest comparable U.S. pure-play memory company, achieved HBM revenue of approximately $1 billion in its most recent quarter at significantly lower volumes. Samsung's memory division, despite its larger overall scale, has been disadvantaged by HBM yield issues and a more cautious pricing strategy relative to SK Hynix's contractual approach.

Financial Breakdown & Forecasts

The two domestic strategy estimates analyzed in this report represent meaningfully different views of SK Hynix's financial trajectory, and the divergence is analytically important. The following consolidated table presents key financial metrics from both perspectives alongside the actual FY2024–FY2025 data to provide proper context for the scale of this earnings ramp.

Metric (KRW bn unless noted)	FY2024 (Actual)	FY2025 (Actual)	1Q26E	2Q26E	FY2026E (Bull)	FY2026E (Base)	FY2027E (Bull)	FY2027E (Base)
Revenue	66,193	97,147	55,800–56,095	72,300–83,236	338,015	275,608	487,058	344,171
Operating Income	23,467	47,206	35,600–37,473	49,500–61,361	249,322	184,669	381,860	238,220
Operating Margin %	35.5%	48.6%	63.7%–66.8%	68.4%–73.7%	73.8%	67.0%	78.4%	69.2%
Net Income (Controlling Interest)	19,789	42,919	34,076 (Bull)	49,760 (Bull)	212,292	142,965	329,997	183,411
EPS — KRW (Bull / Base)	27,182	58,955	—	—	291,608	207,050	453,292	265,627
Forward PER (at ₩1,128,000)	41.5x	19.1x	—	—	3.9x (Bull)	5.4x (Base)	2.5x (Bull)	4.2x (Base)
DRAM Operating Margin	~55–62%	~62%	74–76%	79–80%	80% (Bull)	78.6% (Base)	84% (Bull)	—
NAND Operating Margin	~0–9%	~9%	48–57%	56–65%	57% (Bull)	64.2% (Base)	60% (Bull)	—
ROE %	26.8%	35.6%	—	—	94.1% (Bull)	74.9% (Base)	67.1% (Bull)	52.1% (Base)
BPS (KRW)	101,515	165,544	—	—	454,307	359,045	897,628	608,138
HBM Revenue (USD bn, Base est.)	~$10.8bn	~$22.8bn	$8.7bn	$10.4bn	~$39.9bn	~$39.9bn	—	—

The magnitude of the earnings ramp deserves to be stated plainly: under the bullish domestic estimate, SK Hynix is on track to generate ₩249.3 trillion in operating profit in 2026 alone — more than ten times its 2024 figure of ₩23.5 trillion. Even the more conservative estimate of ₩184.7 trillion represents a nearly 8x increase over the same period. The BPS expansion from ₩165,544 at end of 2025 to ₩454,307–₩359,045 at end of 2026 is equally dramatic and explains why both analysts are anchoring their fair value estimates to book-value multiples rather than earnings multiples — at these profit levels, book value is accumulating faster than the stock price is appreciating, making PBR a natural valuation floor.

The quarterly earnings trajectory from the bullish domestic estimate is particularly informative: operating profit of ₩37.5 trillion in Q1, ₩61.4 trillion in Q2, ₩72.1 trillion in Q3, and ₩78.4 trillion in Q4 2026. This is not a step-function jump followed by plateau; it is a continuously accelerating trajectory that extends through fiscal year 2027, where quarterly operating profit is forecast to reach even higher levels. The driver of this relentless improvement is the combination of ASP increases landing on a fixed-cost manufacturing base — once the silicon and TSV infrastructure is installed and amortized, each additional dollar of ASP increase flows almost entirely to the operating income line.

Valuation Reality Check & Target Price Assessment

The two domestic estimates diverge most significantly on two key assumptions: the pace of DRAM/NAND ASP appreciation, and the treatment of employee performance bonuses. The bullish estimate projects 2026 full-year DRAM ASP (in 1Gb equivalent) reaching $1.61/Gb on average, with ASP jumping from $1.14/Gb in Q1 to $1.87/Gb by Q4 2026 — a trajectory that implies sustained pricing discipline and structural supply tightness through year-end. The more conservative estimate is built on per-GB blended DRAM ASP of approximately $10.1/GB (a different denomination metric but converging to a similar directional story), with HBM ASP rising from $15.2/GB in Q1 to $17.5/GB by Q4 2026. The Q1 2026 operating profit gap between the two estimates — ₩37.5 trillion (bullish) vs. ₩35.6 trillion (conservative, after performance bonus deduction of approximately ₩3.9 trillion) — is largely attributable to the accounting treatment of the approximately 10% employee performance bonus pool, which the more conservative estimate expenses fully in the operating line while the bullish estimate treats it differently. Neither approach is analytically wrong; the more conservative treatment is arguably more GAAP-consistent.

The valuation methodologies are the most analytically interesting aspect of the divergence between the two reports. The bullish domestic estimate uses a simple PBR framework: 2026 BPS of ₩454,307 × a target PBR of 3.80x (calibrated against the projected ROE of 94.1%) yields a fair value of ₩1,726,367, rounded to ₩1,700,000. This methodology is straightforward and conservative in one sense — a 3.80x PBR against a 94.1% ROE implies a price-to-earnings ratio of only about 4x, which is extraordinarily cheap for any company generating nearly 100% return on equity. The risk in this approach is that PBR multiples in memory semiconductors are historically volatile at cycle peaks, and the market may apply a forward-looking discount anticipating ROE normalization as capacity additions ramp in 2027–2028.

The more conservative domestic estimate uses a Sum-of-the-Parts (SOTP) framework that explicitly separates the HBM business from the conventional DRAM and NAND businesses. The HBM segment is valued at 18.6x projected 2026 after-tax profit of ₩33.0 trillion (applying a PEG ratio of 0.5x to the long-term HBM EPS growth rate of 37.2%, which itself is anchored to big tech capex CAGR of 28.6% from 2015 to 2026 plus degree of operating leverage of 1.3x). This yields an HBM business value of approximately ₩613.8 trillion. The Non-HBM business is valued at 2.1x adjusted book value, yielding approximately ₩429.2 trillion. Combined equity value of approximately ₩1,043 trillion divided by the net share count of approximately 691 million shares yields a target price of ₩1,509,711 — rounded to ₩1,500,000. This SOTP approach is methodologically more rigorous because it explicitly models the probability that HBM remains a high-growth, premium-margin business over the next several years while acknowledging that conventional DRAM and NAND are lower-multiple commodity businesses.

Analyst J views the SOTP approach as the more intellectually defensible framework, but notes that it may be undervaluing the HBM segment by using a PEG of 0.5x at a time when HBM is structurally supply-constrained and customer switching costs are extremely high. A PEG of 0.5x makes sense in a competitive market where pricing power is limited; it is arguably too conservative for a product where SK Hynix holds the only commercially proven high-volume HBM3E supply chain and competitors are 12–18 months behind on yield. Applying a PEG of 0.6x–0.7x to the HBM segment would yield a target price in the ₩1,600,000–₩1,700,000 range, consistent with the bullish estimate.

Analyst J's Fair Value Verdict

Based on a synthesis of the two domestic strategic frameworks — the PBR-anchored bullish estimate at ₩1,700,000 and the SOTP-based conservative estimate at ₩1,500,000 — the appropriate fair value range for SK Hynix over a 12-month horizon is ₩1,500,000–₩1,700,000, implying upside of 33%–51% from the current ₩1,128,000. The bullish ₩1,700,000 target, while aggressive in absolute terms, is arithmetically conservative when stress-tested against the projected earnings trajectory: at ₩1,700,000 and 2026E EPS of ₩291,608 (bullish estimate), the implied PER is only 5.8x — below the 5-year historical average forward PER for the stock even at cycle lows. The more conservative ₩1,500,000 target is well-supported by the SOTP analysis and represents the accumulation floor for risk-averse institutional investors. The domestic consensus target of approximately ₩37 trillion for Q1 2026 operating profit appears to slightly underestimate the performance bonus-adjusted trajectory, and full-year 2026 consensus operating profit estimates of approximately ₩200 trillion are running approximately 24.9% below the bullish domestic estimate of ₩249.3 trillion — a gap that, if correct, implies a significant upward revision cycle ahead. Analyst J's recommended accumulation zone is ₩1,050,000–₩1,200,000, with ₩1,500,000 as the near-term price target and ₩1,700,000 as the 12-month bull case target contingent on Q2–Q3 2026 operating results meeting or exceeding the domestic bullish estimates.

Key Risks & Downside Scenarios

The most significant risk to the current thesis is a sudden and unexpected normalization of AI infrastructure investment by hyperscale customers. SK Hynix's earnings model for 2026–2027 is predicated on AI server demand continuing at or above its current growth trajectory, with no pause or retrenchment. If major cloud providers (Amazon, Microsoft, Google, Meta) were to slow their capex commitments due to a combination of AI revenue disappointments, regulatory pressure, or macroeconomic deterioration, the impact on HBM demand would be immediate and severe. This is not currently the base case — Nvidia's forward order book and management commentary from hyperscalers suggest multi-year commitment to AI infrastructure spending — but it is the scenario that would most rapidly invalidate the earnings estimates underlying the current target prices. A 20% reduction in HBM demand assumptions would reduce the bullish 2026 operating profit estimate by approximately ₩30–₩40 trillion, or roughly 12%–16% of the full-year forecast.

The second major risk is Samsung's HBM competitive re-entry. Samsung has publicly committed to resolving its HBM3E quality issues and regaining a meaningful share of Nvidia's supply chain. If Samsung achieves commercial-scale HBM3E yields that satisfy Nvidia's qualification requirements before Q4 2026, the current HBM supply-demand balance would shift from acute shortage to moderate tightness — a transition that would put immediate downward pressure on HBM ASPs. The domestic conservative estimate already assumes that HBM's contribution to total DRAM revenue declines from 43% in 2025 to 28.8% in 2026, reflecting anticipated share dilution as Samsung and Micron scale. However, if Samsung's re-entry is faster or more volume-aggressive than assumed, even the conservative forecast could prove optimistic.

The HBM ASP pressure from Nvidia's Liquid Processing Unit (LPU) architecture deserves specific analytical attention. Recent reports indicate that Nvidia has made progress using LPU hardware to reduce the bandwidth burden on HBM during AI decoding operations. If LPU adoption at scale materially reduces per-server HBM content requirements, the 1,689GB/box HBM assumption embedded in 2026 AI server demand models would be overstated. The more conservative domestic estimate explicitly acknowledges this risk, noting that it could put a brake on HBM ASP appreciation even as stock price rerating from the ADR listing partially offsets this headwind. The net effect is uncertainty around the trajectory of HBM ASP in H2 2026 — which is why both domestic estimates converge on Q2 2026 as the peak sequential ASP growth quarter, with more modest QoQ improvements thereafter.

The US ADR issuance, while primarily a catalyst, also carries execution risk. The April–May 2026 deal roadshow timeline implies an ADR listing in the coming weeks, but market conditions, exchange rate volatility (the KRW/USD trajectory directly affects ADR economics), or unexpected regulatory hurdles could delay the process. More importantly, the ADR pricing will reflect the market's global peer group valuation of SK Hynix — if U.S. institutional investors apply a Micron-like valuation framework (Micron trades at a meaningfully lower multiple than SK Hynix's implied domestic fair value), the ADR could create modest short-term selling pressure from arbitrage. The long-term impact of successful ADR listing is unambiguously positive for the shareholder base and liquidity.

Geopolitical and export control risks cannot be dismissed. SK Hynix's primary manufacturing facilities for cutting-edge DRAM are located in Korea, but the company's NAND manufacturing has significant China exposure through its Wuxi facility. Any escalation of U.S.-China trade restrictions on semiconductor technology — including potential restrictions on HBM exports to China — would create supply chain disruption risk and could limit SK Hynix's addressable market. Additionally, memory is a capital-intensive industry where exchange rate movements (particularly KRW weakness against USD) can create a transient competitive cost advantage, but also complicate financial reporting for global investors.

Strategic Outlook

SK Hynix is the most asymmetric investment opportunity in the global semiconductor sector as of April 2026. The stock is trading at less than 4x–5x current-year earnings in a profitability cycle that shows no signs of peaking before mid-2027 by the most credible domestic research estimates. The competitive structure of the HBM market — a duopoly between SK Hynix and, at increasing scale, Micron and Samsung — provides a degree of pricing discipline that has never existed in conventional DRAM, where three-player competition typically ended cycles faster than the fundamentals warranted.

The shareholder return dimension is likely to become an increasingly important re-rating catalyst as the cash accumulation becomes too large to ignore. Based on the more conservative domestic estimate, SK Hynix's 2026 FCF is approximately ₩60.3 trillion. The company's stated policy of returning 50% of FCF to shareholders through dividends and buybacks — combined with the ADR issuance that is expected to be accompanied by share repurchase and cancellation — implies a shareholder return yield that will reach double digits by 2027. The domestic conservative estimate projects a 2027F DPS of ₩57,075 (a theoretical maximum based on FCF of ₩155.3 trillion × 50% allocation), implying a dividend yield of approximately 5.1% on the current share price. At these return levels, SK Hynix transitions from a pure growth story to a hybrid growth-and-income profile that widens its institutional investor base considerably.

For global institutional investors — particularly those currently underweight Korean equities — the ADR listing provides the most natural entry point into this thesis. The ADR will offer USD-denominated exposure to the world's most profitable memory company at a valuation multiple that would be considered deeply discounted by any U.S. technology sector standard. The comparable exercise is instructive: Nvidia trades at approximately 17.7x forward PER on FY2028 estimates showing +32.8% revenue growth. SK Hynix is forecast to deliver +44% revenue growth in 2027 (bullish estimate) while trading at approximately 2.5x forward PER on the same timeframe. The magnitude of this valuation gap — which reflects a combination of market access friction, Korean corporate governance discount, and cyclicality skepticism — is unlikely to persist once U.S. institutional investors have direct ADR access and can perform the comparative earnings analysis side by side.

The strategic recommendation is a Buy with a 12-month price target range of ₩1,500,000–₩1,700,000. The recommended accumulation zone is ₩1,050,000–₩1,200,000 on any weakness. Investors should be prepared for intra-quarter volatility driven by Korean market dynamics, exchange rate movements, and any quarterly results that fall short of the aggressive domestic estimates. The Q1 2026 actual results (expected in late April 2026) will be the first definitive data point that tests whether the operating profit trajectory is tracking closer to the ₩37.5 trillion bullish estimate or the ₩35.6 trillion conservative estimate. Either outcome leaves the stock materially undervalued relative to a 12-month fair value above ₩1,500,000.

---

Disclaimer: The analysis provided on Capitalsight.net is for informational and educational purposes only and does not constitute financial, investment, or trading advice. Investing in the stock market involves risk, including the loss of principal. All financial data and forecasts cited in this article are sourced from domestic strategy research reports on SK Hynix (000660), dated April 18–20, 2026. Specific brokerage names have been anonymized per editorial policy. All investment decisions are solely the responsibility of the individual investor. Please consult with a certified financial advisor and conduct your own due diligence before making any investment decisions.