A 72% Operating Margin Is a Blow-Off, Not a Baseline: SK hynix's Best Quarter Ever Meets a Flat Tape. Downgrading to Hold After Three Outperforms and a 4.5x

Results vs. Consensus

Q1 2026 Scorecard

Year-Over-Year and Quarter-Over-Quarter (K-IFRS)

Quality of Beat

Revenue: The +4.0% consensus beat and +60% sequential jump were pricing-led and broad. DRAM ASP rose mid-60% and NAND ASP mid-70% QoQ; crossing ₩50tn for the first time is a genuine milestone. The quality of the revenue is high, but the rate of ascent (+60% QoQ) is itself a peak-cycle signature: revenue does not compound at this pace except at the top of a pricing spike.

Margins: The 72% operating margin is the defining figure and the defining risk. It is a real, operating number driven by extreme pricing, not an accounting artifact. But high-50s margins were already late-cycle at Q4; low-70s margins are unprecedented, which by definition makes them unsustainable. The next move in margin is down, and that is the crux of the downgrade.

EPS / net income: The +45.7% EPS beat is the lowest-quality headline of the four quarters we have covered. Net income exceeds operating profit on a ₩14tn non-operating tailwind (valuation and FX gains), so the 77% net margin and the ₩55,493 EPS materially overstate the operating result. Pre-tax income was ₩51.6tn. This is the same non-operating-inflation pattern we flagged in prior peak quarters, and it argues against extrapolating the headline.

Segment Performance

Product-Line Detail — Q1 2026

DRAM — Pure Price

DRAM bit shipments were roughly flat sequentially, in line with guidance, while ASP rose by a mid-60% percentage as conventional DRAM pricing accelerated. The entire DRAM revenue increase was price, not volume, which is the cleanest possible read of a supply-constrained market at the top of a pricing cycle. SK hynix steered its fixed bit budget to the highest-value demand (HBM, 128GB-plus server modules).

Assessment: A mid-60% ASP increase on flat bits is extraordinary and, by definition, late-cycle. It is excellent for the current quarter and a flashing indicator for the next several: ASP gains of this magnitude do not recur, and they set up difficult sequential comparisons once pricing normalizes.

NAND — ASP +mid-70%, Volumes Down

NAND bit shipments fell roughly 10% QoQ on a high base and a mix shift toward high-value enterprise SSD, while ASP rose sharply by a mid-70% percentage on strength across all product lines. The eSSD/AI-storage structural narrative remains intact.

Assessment: Even more than DRAM, the NAND result is pure pricing. A mid-70% ASP increase is the kind of move that, historically, marks the final innings of a NAND up-cycle before supply responds. We continue to weight NAND as the higher-beta, faster-correcting leg.

HBM — Still the Anchor, Still Sold Out

HBM remains the structural core. Management reiterated that customer HBM demand for the next three years far exceeds current supply, that HBM4 is ramping on agreed customer schedules, and that it allocates between HBM and general DRAM for balanced AI-ecosystem growth rather than to maximize near-term revenue. HBM4E samples arrive in H2 2026 for 2027 mass production.

Assessment: HBM is the reason this is a downgrade to Hold rather than to Underperform. The contracted, multi-year HBM demand provides a genuine earnings floor that prior memory cycles never had, which limits the downside. But HBM cannot offset a normalization in the conventional DRAM and NAND pricing that drove the 72% margin, and it is conventional pricing, not HBM, that is at an unsustainable extreme.

Key Topics & Management Commentary

Overall management tone: Confident and structurally bullish, with one notable tell: the first question of the call was about spot prices weakening, and management spent its answer explaining why that is not a peak. When the buy-side's opening question is whether this is the top, and management is in the position of rebutting it, the market is already wrestling with cycle position, which is exactly the debate the rating now turns on.

1. The First ₩50 Trillion Quarter

"First quarter revenue increased by 60% QoQ and 198% YoY to KRW 52.6 trillion, surpassing KRW 50 trillion mark for the first time… operating margin also improved by 13 percentage points QoQ to 72%, marking another all-time high." — Kim Woo-hyun, CFO

Revenue crossed ₩50tn and the operating margin reached 72%, both records, on broad pricing strength across HBM, conventional DRAM, and NAND. It is, by any measure, the best quarter the company has ever produced.

Assessment: A genuinely historic result. The discipline point is not that the quarter was anything less than spectacular; it is that "another all-time high" at a 72% margin is, in a cyclical industry, the definition of peak, and peaks are where disciplined ratings step back.

2. The Non-Operating Inflation, Stated Plainly

"Net non-operating profit reached KRW 14 trillion, reflecting… foreign exchange-related net gain of KRW 1.6 trillion… and valuation gains on investment assets of KRW 9.9 trillion." — Kim Woo-hyun, CFO

Management disclosed that net income was lifted by ₩14tn of non-operating income, dominated by ₩9.9tn of non-cash investment valuation gains, which pushed net income above operating profit and the net margin to 77%.

Assessment: The disclosure is transparent and the figure is large enough to matter: it inflates the EPS beat to +45.7% from an operating reality nearer the +4% revenue surprise. Any screen or model anchored to the headline EPS or net margin will badly misjudge the run-rate.

3. Spot Prices Soften, and Management Rebuts the Peak

"The moderate trend in spot prices, rather than being a sign of market peak out, appears to be a temporary phenomenon resulting from some inventory entering the market from some distribution channels due to the recent price increase." — Kim Woo-hyun, CFO

The opening analyst question flagged softening memory spot prices as a possible peak signal. Management argued the spot market is a small, unrepresentative slice and attributed the softness to channel inventory rather than a turn in underlying demand.

Assessment: The answer is reasonable, and probably right for the immediate quarter. But the question being asked first, by the first analyst, is itself the signal: the market is now actively looking for the top. Spot pricing is a leading indicator, and softening spot prices at a 72% margin is precisely the configuration that precedes a normalization. We treat management's rebuttal as credible-but-interested and elevate spot pricing to the primary watch item.

4. Pricing Framed as Structural

"The current price strength is driven by structural changes in the market, not by the temporary supply demand imbalance… the favorable pricing environment is expected to continue for the time being." — Kim Woo-hyun, CFO

Management held its structural-cycle argument: AI has elevated memory's strategic importance, supply is constrained by HBM capacity consumption, and customers prioritize volume over price, so pricing stays favorable.

Assessment: We have given this argument increasing credence over four quarters, and the supply-side half (HBM cannibalizing conventional capacity) is genuinely structural. But "structural" and "at a sustainable level" are different claims. The demand may be structural; a 72% margin is not. "For the time being" is doing the work.

5. HBM: Three-Year Demand Still Exceeds Supply

"Customers' demand for the next three years far exceeds our current supply capacity… we are trying to achieve optimal allocation between HBM and general DRAM… rather than trying to maximize revenue." — Park Joon-deok, Head of DRAM Marketing

HBM remains sold out multiple years forward, with management allocating scarce capacity for balanced ecosystem growth and ramping HBM4 on customer schedules. The three-year demand-over-supply gap is unchanged from prior quarters.

Assessment: This is the earnings floor that makes the downside bounded and the rating Hold rather than Underperform. Contracted multi-year HBM demand genuinely de-risks the franchise. It does not, however, address the cyclical risk in the conventional DRAM and NAND pricing that drove the margin to 72%.

6. Investment Up, Oversupply "Unlikely"

"We expect investment in 2026 to increase significantly YoY… both customers and suppliers agree on the importance of securing long-term visibility… There should be no major concerns about oversupply as in the past." — Kim Woo-hyun, CFO

Management guided 2026 investment significantly higher (Yongin, M15X, EUV) and argued the demand-visibility regime makes a repeat of past oversupply unlikely. The Yongin Phase 1 clean room was pulled forward to February 2027.

Assessment: The perennial bear mechanism, stated and dismissed by management. A significant 2026 CapEx increase, across the whole industry, is exactly the capacity that historically arrives into a softer 2027. The demand-visibility argument may genuinely cushion the next downturn, but it does not eliminate it, and the spend is happening precisely as margins peak.

7. The Balance Sheet and the ₩100 Trillion Target

"Achieving financial soundness with net cash of more than KRW 100 trillion and expanding shareholder returns are goals that can be pursued in parallel." — Kim Woo-hyun, CFO

Net cash reached ₩35tn (cash ₩54.3tn, debt ₩19.3tn), and management set a ₩100tn net-cash target while committing to additional shareholder returns (buybacks and cancellations) within 2026, on top of FY25's ₩2.1tn dividend and ₩12.2tn cancellation.

Assessment: The balance-sheet strength and continued capital return are genuine supports that further bound the downside. A net-cash fortress and ongoing buybacks are why we are comfortable at Hold rather than more bearish. But cash generation at a 72% margin is itself peak cash generation, and the ₩100tn target assumes the earnings power persists.

8. The US ADR Filing

"On March 24, we confidentially submitted a registration statement to the U.S. Securities and Exchange Commission related to the proposed ADR offering, and we are proceeding with the goal of listing on the U.S. securities markets within the year." — Kim Woo-hyun, CFO

SK hynix confirmed a confidential SEC registration for a US ADR, targeting a listing within 2026, while declining to specify size, structure, or timing. This follows the "looking into various options" language from the Q4 call.

Assessment: The most credible non-cyclical re-rating catalyst in the story. A US listing would widen the investor base, improve dollar-capital access and index eligibility, and could compress the Korea discount. It is the single development most likely to keep the stock supported even as the earnings cycle peaks, and it is the main reason a return to Outperform on a pullback is plausible.

9. Product Roadmap: LPDDR6, SOCAMM2, HBM4E

"We completed development of the industry's first 1c nm-based LPDDR6… we began mass production this month of our 192 GB SOCAMM2 product… optimized for NVIDIA's Vera Rubin platform." — Kim Woo-hyun, CFO

SK hynix completed the industry-first 1c-nm LPDDR6 (+33% speed, +20% power efficiency) for H2 supply, began mass production of 192GB SOCAMM2 for NVIDIA's Vera Rubin platform, shipped the first CTF-based 321-layer QLC client SSD, and will sample HBM4E in H2 for 2027 mass production.

Assessment: The product cadence confirms the technology lead is intact and extending. None of this is in question. The roadmap supports the earnings floor and the long-term franchise value; it does not change the near-term cycle-position math.

10. Demand Broadens to Agentic AI

"AI technology is now evolving rapidly beyond the training phase of large-scale models into the inference and agentic AI stage… the demand base is broadening for both DRAM and NAND." — Kim Woo-hyun, CFO

Management's demand framework mirrors the broader industry's agentic-AI thesis, with inference and agents broadening memory demand beyond HBM, offset partly by some softening in price-sensitive PC and mobile.

Assessment: Directionally sound and consistent across the franchise's prior calls. The demand story is not the issue; the price level is. Agentic AI may sustain demand growth, but it will not hold conventional memory ASP at a level that supports a 72% margin indefinitely.

Guidance & Outlook

SK hynix guides bit shipments and frames pricing qualitatively. For Q2 2026:

The shift to positive bit-shipment guidance (DRAM +high-single-digit, NAND +mid-15%) is the operationally meaningful change: SK hynix is moving from pure price capture toward volume-plus-price. That is constructive for absolute profit but introduces the classic late-cycle dynamic, adding volume just as ASP gains of mid-60% (DRAM) and mid-70% (NAND) become impossible to repeat. The forward earnings question is no longer whether SK hynix can grow, but whether a 72% margin can hold, and the honest answer is that it cannot for long.

Analyst Q&A Highlights

Is the Softening Spot Price a Peak Signal?

The opening question, tellingly, asked whether weakening memory spot prices signal a peak. Management said no, framing the softness as a channel-inventory artifact.

Q: "The memory spot prices have been on a steep upward trend, but are now showing some signs of weakness. Some are concerned that this may be a signal of a peakout… Does the company see this as just a temporary adjustment?"
— Rok-ho Kim, Hana Securities

A: "The spot market itself takes up a very small part of the overall DRAM market… The moderate trend in spot prices, rather than being a sign of market peak out, appears to be a temporary phenomenon resulting from some inventory entering the market from some distribution channels."
— Kim Woo-hyun, CFO

Assessment: The most important exchange on the call for our rating. The answer is credible for the quarter, but the question itself, asked first, confirms the buy-side is now hunting for the top. Softening spot at a 72% margin is the leading edge of a normalization; we elevate it to the primary watch item.

Is the Price Cycle Structural?

A follow-on asked for the broader price-cycle outlook. Management reiterated its structural-change argument.

Q: "It seems as if the unprecedented super cycle is driving up the memory prices. What is the company's outlook on the memory price trends down the road?"
— Sun-woo Kim, Meritz Securities

A: "We see that the current price strength is driven by structural changes in the market, not by the temporary supply demand imbalance… the favorable pricing environment is expected to continue for the time being."
— Kim Woo-hyun, CFO

Assessment: The demand-side structural case is real; the supply-side constraint (HBM consuming capacity) is real. Neither makes a 72% margin durable. We accept structural demand and reject structural 72% margins.

Enhanced Multi-Year LTAs

An international analyst asked about the enhanced LTAs and how they reduce volatility. Management described firm multi-year structures still being negotiated under supply constraints.

Q: "Could you provide us an update on the current progress on your enhanced new long-term agreements… whether they apply to DDR, NAND Flash, and HBM?"
— Nicolas Gaudois, UBS

A: "Unlike past LTAs, we are comprehensively reviewing various approaches and structural options, but due to current supply constraints, we are limited from accommodating all customer requests… it can also reduce the volatility that has repeatedly plagued the memory industry."
— Kim Woo-hyun, CFO

Assessment: The LTA-driven volatility reduction is the structural bull case that, if proven through a down-cycle, would eventually justify a higher through-cycle multiple and a return to Outperform. It remains the most important multi-year watch item, still unproven against a price decline.

HBM4 Qualification and Share

An analyst asked for the HBM4 qualification status and year-end share. Management reaffirmed the three-year demand backlog and allocation discipline.

Q: "What is the update on HBM4 qualification and the expected timing for full scale shipment… what is the company's expectation about HBM4's share out of the total HBM?"
— Ricky Seo, HSBC

A: "Customers' demand for the next three years far exceeds our current supply capacity… we are trying to achieve optimal allocation between HBM and general DRAM… rather than trying to maximize revenue."
— Park Joon-deok, Head of DRAM Marketing

Assessment: HBM remains the floor under the stock. A supplier rationing scarce capacity rather than maximizing revenue still has pricing power. This is why the downside is bounded and the rating is Hold, not lower.

Investment Expansion and Oversupply

An analyst raised the bear case directly: aggressive industry investment and the risk of repeating past oversupply. Management argued the visibility regime makes that unlikely.

Q: "Does the company also plan to expand investment…? It is understandable if there are some concerns about past oversupply issues recurring."
— Young-min Ko, DAOL Investment & Securities

A: "We expect investment in 2026 to increase significantly YoY… both customers and suppliers agree on the importance of securing long-term visibility… There should be no major concerns about oversupply as in the past."
— Kim Woo-hyun, CFO

Assessment: Management is asking investors to believe this capacity super-cycle is different. History argues for skepticism: a significant industry-wide 2026 build, poured precisely as margins peak, is the classic precursor to the next normalization. We respect the discipline and decline to underwrite a permanently tame cycle at a 72% margin.

NAND and Enterprise SSD

An analyst asked the NAND plan for AI-driven storage demand. Management pointed to the 321-layer migration.

Q: "With demand driven by AI expected to fuel rapid growth in NAND demand… what is SK hynix's plan to meet future market demand?"
— Jay Kwon, JPMorgan

A: "We plan to migrate more than 50% of our domestic production to 321-layer technology by the end of this year… this two-generation jump… is expected to yield significant productivity gain."
— Song Chang-seok, Head of NAND Marketing

Assessment: The 321-layer migration is a cost win and a supply addition. In NAND, the more commoditized leg now riding a mid-70% ASP increase, bit-supply growth is the variable that historically re-balances the market fastest. Constructive technology, cyclically double-edged at this price.

Shareholder Return and the ADR

The closing question addressed the ₩100tn cash goal and an ADR update. Management committed to additional returns and confirmed the SEC review is underway.

Q: "Regarding the goal of KRW 100 trillion in cash… what is the company's direction for shareholder return policies? And could you give us an update about ADRs?"
— Dong-Hee Han, SK Securities

A: "We plan to develop additional shareholder return measures within this year, including not only dividends but also share buyback and cancellation… About ADRs, the U.S. SEC review is currently underway."
— Kim Woo-hyun, CFO

Assessment: Continued capital return plus a progressing ADR are the two supports that keep this a Hold rather than a more bearish call, and the two developments most likely to drive a return to Outperform if the stock pulls back.

Market Reaction

All price levels are FMP-verified KRX closing prices; the benchmark is the KOSPI (^KS11). SK hynix discloses before the Korean market open, so the reaction is the print-day session.

• Pre-print close (2026-04-21): ₩1,224,000, an all-time high.
• Reaction (2026-04-22) close: ₩1,223,000, flat (−0.1%) on a record print, while the KOSPI rose +0.5%.
• Volume: ~2.9M shares versus a ~4.4M 30-day average (0.7x), below average, a notably muted, low-conviction session.
• Setup: +88% YTD and +604% over the trailing twelve months (from ₩173,800), one of the most extreme one-year runs in the franchise's history.

A flat reaction on below-average volume to the best quarter in company history is the single most important signal in this report. Two quarters ago the stock fell on a record (Q3); one quarter ago it rallied 5.1% (Q4); now it does nothing. That oscillation, and specifically the failure of a 72%-margin blow-off to move the price after a +604% year, is the market telling us the earnings are fully priced and then some. When the best quarter ever cannot lift the stock, the marginal buyer has been satisfied.

For our rating, this is the behavioral confirmation of the cycle-position call. We do not read it as the start of a decline, the franchise is too strong and the HBM floor too firm for that, but as the exhaustion of the upside that justified three Outperform calls. The risk/reward at ₩1.22M, after a 4.5x, on a peak margin that the market will not pay up for, is balanced at best.

Street Perspective

Debate: Is 72% a Peak or a New Plateau?

Bull view: Structural AI demand, HBM capacity constraints, and firm multi-year LTAs hold pricing higher for longer; with 2026 sold out, the high margin persists well beyond what skeptics expect, and the through-cycle margin has structurally re-rated.

Bear view: 72% is the highest operating margin memory has ever produced and is definitionally a peak. Spot prices are already softening, the industry is pouring CapEx, and the operating leverage that drove margins from 42% to 72% in a year reverses just as violently. The next print is the high-water mark.

Our take: Peak, with a longer plateau than prior cycles but a peak nonetheless. The structural demand and LTA arguments genuinely extend the high-margin period and cushion the eventual descent, which is why we downgrade only to Hold. But we will not underwrite 72%, or anything close, as a sustainable through-cycle margin.

Debate: Does the 4.5x Run Leave Any Asymmetry?

Bull view: Earnings are still exploding and the ADR catalyst, capital return, and ₩100tn cash target keep the stock supported; on contracted 2026 earnings the forward multiple is not demanding, and a US listing could re-rate it regardless of the cycle.

Bear view: A stock up ~4.5x in three quarters and +604% in a year, going flat on a record, has exhausted its buyers. The non-operating-inflated headline makes the trailing multiple look cheaper than the operating reality, the classic peak-earnings trap.

Our take: The asymmetry that made the July initiation compelling at ₩268,500 is gone at ₩1,224,000. The downside is bounded by HBM and the balance sheet; the upside is capped by a peak margin the market will not pay up for. Balanced risk/reward is a Hold.

Debate: Can the ADR Re-Rate the Stock Through the Cycle?

Bull view: A US listing compresses the long-standing Korea discount, widens the investor base, and improves index eligibility, a non-cyclical catalyst that could lift the multiple even as the earnings cycle peaks.

Bear view: The ADR is unconfirmed in size and timing, and listing into a cyclical peak risks the new US investor base buying the top. A re-rating is speculative until terms are set.

Our take: The ADR is the most credible reason to return to Outperform, and the main upside option in the Hold. We will watch its progress closely; concrete terms on attractive pricing would change the calculus.

Thesis Scorecard: 2025_Q4 Signposts Revisited

Our Q4 2025 update set signposts for Q1. They graded bullish operationally, which is precisely the point: the downgrade is not about the franchise stumbling. It is about cycle position and price.

Scorecard summary: Operationally, the signposts are bullish, HBM intact, LTAs holding, ADR filed exactly as flagged. But two grades carry the rating: the margin expanding to 72% (a peak signal, not a baseline) and the conventional ASP blow-off (mid-60s/mid-70s, unsustainable), alongside a rising out-year supply risk and the first PC/mobile/spot cracks. The franchise passed; the cycle is at its top. That combination, a flawless company at a peak price, is the textbook setup for taking a gain.

Bottom Line: Downgrading to Hold

Rating decision: We downgrade SK hynix to Hold from Outperform. We initiated at Outperform in July 2025 at ₩268,500, maintained it through the Q3 and Q4 records, and the thesis delivered a ~4.5x to ₩1,224,000. We are taking the win. This is not a downgrade on the business, which printed the best quarter in its history and remains the supply-constrained HBM leader with 2026 sold out, a net-cash balance sheet, and a US ADR in registration. It is a downgrade on cycle position and price.

The case is straightforward. A 72% operating margin is a blow-off, the highest memory has ever produced, and from here the margin compresses rather than expands. The stock is up ~4.5x in three quarters and +604% in a year, and it went flat on a record print, the market is no longer paying for the prints. The headline EPS is non-operating-inflated. The asymmetry that made the July initiation a 3:1 opportunity is gone; at ₩1.22M, with the upside capped by a peak margin and the downside bounded by HBM and the balance sheet, the risk/reward is balanced. That is a Hold.

What would move us back to Outperform: a meaningful pullback that restores asymmetry (toward the ₩850,000–950,000 area); a US ADR on attractive terms that re-rates the multiple independent of the cycle; or hard evidence that multi-year LTAs have structurally raised the through-cycle margin floor, proven by pricing holding through any spot-led softness.

What would move us to Underperform: spot-price weakness broadening into contract pricing and a clear margin roll from the 72% peak; signs the 2026 CapEx wave is building 2027 oversupply faster than demand absorbs it; or a Samsung/Micron HBM4 breakthrough that contests the scarcity premium and removes the earnings floor.

Signposts for Q2 2026 earnings (late July 2026):