On April 23, 2026, SK Hynix released its Q1 FY2026 earnings (consolidated K-IFRS) for the period ending March 31. All key metrics hit fresh all-time highs for the fourth straight quarter.
| Metric | Q1 2026 | QoQ | YoY |
|---|---|---|---|
| Revenue | KRW 52.58 trillion (~USD 37.9B) | +60% | +198% |
| Gross Profit | KRW 41.68 trillion | +85% | +313% |
| Operating Profit | KRW 37.61 trillion (~USD 27.1B) | +96% | +405% |
| EBITDA | KRW 41.34 trillion | +82% | +284% |
| Net Profit | KRW 40.35 trillion | +165% | +398% |
| Gross Margin | 79% | +10pp | +22pp |
| Operating Margin | 72% | +14pp | +30pp |
| EBITDA Margin | 79% | +10pp | — |
| Net Margin | 77% | +31pp | — |
For context, full-year 2025 operating profit was KRW 47.2 trillion — Q1 alone delivered ~80% of that.
A 72% operating margin is extremely rare among large-scale manufacturers, far exceeding TSMC's Q1 2026 guidance of ~54-56%. Even more striking is the 79% gross margin — closer to a software company than a hardware maker, and the clearest signal of this memory cycle's intensity.
But beneath the headline numbers, one detail must be flagged (see Section 4): net profit includes ~KRW 11.5 trillion in non-recurring gains. Stripping those out, core net profit is around KRW 28-29 trillion — still extremely strong, but the "77% net margin" needs proper context.
The Three Engines: Price Takes Center Stage
From SK Hynix's disclosed Q1 volume and price data:
DRAM: Bit shipments (B/G) flat QoQ, ASP up mid-60% QoQ. DRAM accounted for 78% of total revenue (vs. 76% in Q4 2025), or ~KRW 41 trillion.
NAND: Bit shipments down ~10% QoQ, ASP up mid-70% QoQ. NAND (including Solidigm) contributed 21% of revenue, or ~KRW 11 trillion.
The message is clear: growth was almost entirely price-driven, with volumes actually shrinking. In a tight market, the company chose to allocate capacity to high-value products and walk away from low-margin orders.
Engine 1: HBM — The Real Money Behind AI Compute
High Bandwidth Memory () is the absolute core of this earnings explosion. SK Hynix previously disclosed that HBM revenue more than doubled YoY in 2025, and in 2026 it is expected to account for ~28.8% of total revenue.
SK Hynix's lead in HBM is decisive:
- HBM3E 12-layer is the mainstream configuration for current AI server GPUs
- HBM4 completed mass production readiness globally first in September 2025, with volume shipments already underway in Q1 as per customer schedules; the company noted "close collaboration with customers from early product development" and continues to ramp
- Market share: 34% in HBM (Samsung 33%, Micron 26%); in HBM4, SK Hynix is expected to hold 54% share in 2026
More critically, SK Hynix's entire 2026 HBM capacity is already fully booked by customers, with some large clients prepaying to secure supply. A five-year DRAM long-term agreement with Google is also under negotiation.
Engine 2: Structural Shortage in Legacy DRAM
According to TrendForce, commodity DRAM contract prices rose 90-95% QoQ in Q1, but SK Hynix's blended DRAM ASP rose mid-60% — the gap reflects product mix (HBM prices are relatively stable, while generic DRAM surged).
The price surge stems from a deep supply-demand mismatch: AI servers' voracious demand for high-capacity DDR5, combined with major manufacturers prioritizing wafer capacity for HBM (which consumes ~3x the silicon of equivalent DRAM), has created a severe shortage of general-purpose DRAM. Analysts note that generic DRAM gross margins now even exceed those of premium HBM.
Engine 3: NAND — From Drag to Driver, But Volumes Shrink
NAND flash was the biggest "surprise," but also the part that needs the most careful unpacking:
- ASP up mid-70% QoQ, one of the steepest increases in NAND history
- But bit shipments fell ~10% QoQ — the company actively shifted capacity from low-margin products to high-value items like eSSD
- The 321-layer QLC-based PQC21 series has begun shipping to the client SSD market
- On the eSSD side, a balanced portfolio of high-performance TLC and high-capacity QLC targets AI demand
Through its integration of Solidigm, SK Hynix has built a differentiated position in enterprise SSDs, perfectly timed to capture structural demand for high-capacity, energy-efficient storage from AI inference workloads.
From Cyclical Stock to Infrastructure: A Fundamental Shift in Profit Model
What makes this cycle different is that SK Hynix may no longer be a traditional memory cyclical.
Historically, the memory industry followed a clear "build capacity → oversupply → price crash → cut production → recovery" cycle, with profit peaks rarely lasting more than 2-3 quarters. But this cycle shows several fundamental differences:
1. Structural demand: AI infrastructure investment is enterprise capex driven by hyperscalers' multi-year plans, decoupled from consumer electronics cycles. The company stated: "AI technology is rapidly evolving from training to inference and agentic AI. As data generated by AI agents grows, the demand base for various memory products is broadening."
2. Supply constraints: Manufacturing HBM requires advanced packaging and wafer resources that cannot be easily expanded. SK Hynix's 2026 capex will "significantly" increase, mainly for the Yongin cluster infrastructure and M15X ramp — but management frames it as "strategically securing mid-to-long-term capacity," not aggressive expansion.
3. Long-term contracts lock in pricing: Unlike spot market pricing in the past, HBM and some DRAM are now moving to multi-year contracts. The company has signed long-term agreements with Apple, Nvidia, Google, and others, insulating it from short-term price swings.
4. Competitive moat: HBM requires extremely high yields, packaging expertise, and customer collaboration. Samsung's HBM4 mass production has been delayed by yield issues, and Micron is still catching up. SK Hynix's first-mover advantage in this critical window could last for years.
The "Earnings Quality" Issue That Demands Attention
This is the most overlooked yet most important aspect of the report.
Net profit breakdown:
| Item | Amount |
|---|---|
| Operating Profit | KRW 37.61 trillion |
| Net Financial Income | +KRW 0.02 trillion |
| FX-related Net Gain | +KRW 1.57 trillion |
| Other Non-Operating Income | +KRW 12.44 trillion (includes KRW 9.94 trillion investment asset valuation gain) |
| Income Tax | -KRW 11.27 trillion |
| Net Profit | KRW 40.35 trillion |
In other words, ~KRW 11.5 trillion (29% of net profit) came from FX gains and asset revaluations — non-recurring items, especially the KRW 9.94 trillion unrealized gain.
Using operating profit as a base and applying a normal tax rate, "core net profit" is around KRW 28-29 trillion, with a core net margin of ~55%. That's still a stunning number, but much closer to real operating performance than the reported 77% net margin. It also explains why income tax surged +846% YoY and +368% QoQ — the tax base (pre-tax profit) includes large non-operating items.
Another red flag: accounts receivable.
AR jumped from KRW 18.2 trillion in Q4 2025 to KRW 33.8 trillion in Q1 2026, up 85% QoQ, outpacing revenue growth of 60%. This means revenue has been recognized but cash hasn't been collected yet. Operating cash flow of KRW 26.43 trillion was weighed down by a KRW 17.65 trillion working capital drain — a sign that a significant portion of this quarter's "paper prosperity" hasn't turned into actual cash inflows.
This isn't necessarily a problem (in a seller's market, extended payment terms are normal), but investors should closely monitor the AR/revenue ratio in coming quarters.
Financial Health and Shareholder Returns: The "KRW 100 Trillion" Target Emerges
Alongside the earnings boom, SK Hynix's balance sheet has undergone a fundamental transformation:
| Metric | Q1 2025 | Q4 2025 | Q1 2026 |
|---|---|---|---|
| Cash (incl. short-term financial assets) | KRW 14.31 trillion | KRW 34.94 trillion | KRW 54.33 trillion |
| Interest-bearing Debt | KRW 23.33 trillion | KRW 22.25 trillion | KRW 19.32 trillion |
| Debt/Equity Ratio | 29% | 18% | 12% |
| Net Debt/Equity Ratio | 11% | -11% | -21% |
Net cash in Q1 2026 reached ~KRW 35 trillion, and the trend is accelerating. Management introduced a previously unstated strategic target:
"Securing a net cash level exceeding KRW 100 trillion as a long-term goal, pursued in parallel with expanding shareholder returns."
This implies SK Hynix aims to build a "strategic reserve" to weather any market downturn. At the current cash generation pace, this target is achievable within 6-8 quarters.
On shareholder returns, management made new explicit commitments:
- FY2025 dividend of KRW 3,000 per share, total ~KRW 2.1 trillion
- Plan to cancel 50 million treasury shares (2.1% of outstanding shares)
- "We will actively evaluate additional shareholder return measures such as dividends, share buybacks, and cancellations within the year, and prepare implementation plans" — essentially telling the market that larger buybacks/cancellations are coming.
Q2 Guidance and Product Roadmap: Management Tells You Where They're Going
Q2 2026 Shipment Guidance (from company official)
- DRAM: Bit shipments up high single digits QoQ, focusing on high-density server modules and mobile
- NAND: Bit shipments up mid-teens QoQ, including Solidigm
The signal: Q2 sees both volume and price growth — DRAM modestly recovering, NAND ramping significantly, suggesting the company believes prices have reached a relative steady state and can release capacity.
Product/Technology Roadmap (first-time milestones in this report)
HBM:
- Continue strengthening comprehensive execution in "performance, yield, quality, supply stability"
- HBM4 ramp-up according to customer schedules
DRAM:
- World's first 1cnm LPDDR6, debuting in a major customer's next-gen flagship smartphone in H2 (likely Samsung Galaxy S series)
- 1cnm 192GB SOCAMM2 already in mass production, optimized for Nvidia's Vera Rubin platform
NAND:
- Client SSD: PQC21 shipping (321-layer QLC based on CTF technology)
- Enterprise SSD: Balanced portfolio of high-performance TLC and high-capacity QLC for diverse AI needs
SOCAMM2 (Small Outline CAMM2) deserves special attention — it's a custom memory module for Nvidia's next-gen AI server platform, and SK Hynix is the first and currently only mass-production supplier. This means, beyond HBM, SK Hynix has established another "entry ticket" into the Nvidia ecosystem.
Risk Factors: The Other Side of the Coin
Even with such strong data, several variables warrant attention:
1. Geopolitics and trade policy: The US government has mentioned the possibility of 100% tariffs on semiconductors not manufactured in the US. The Indiana packaging plant can mitigate some pressure, but US-China-Korea dynamics remain a systemic risk.
2. Competitive catch-up: If Samsung resolves its HBM4 yield issues, it could divert orders; Micron has been certified by Nvidia and is rapidly expanding capacity. SK Hynix's current share advantage requires continuous iteration to maintain.
3. Earnings quality: This quarter's net profit includes ~KRW 11.5 trillion in non-recurring gains. Without similar items in future quarters, net profit growth could appear to "decline." Markets need to distinguish core operating profit from one-time gains.
4. Accounts receivable growth: AR up 85% QoQ, far outpacing revenue growth. Cash collection in coming quarters needs monitoring.
5. Weak consumer electronics: The company acknowledged "cost pressure from rising memory prices has led to some demand softening in PC/smartphone markets, including adjustments in device shipments and product mix." While server demand is sufficient to offset, consumer signals bear watching.
6. Cyclical risk: The stock has risen over 5x in the past year, pricing in a lot of optimism. Any sign of AI capex slowdown could trigger a sharp correction.
Conclusion: A New Order for Memory
SK Hynix's Q1 2026 earnings report is, on the numbers, nearly flawless — 79% gross margin, 72% operating margin, fourth straight record, net cash doubling to KRW 35 trillion. But its deeper significance is validating a new industry narrative:
Memory is no longer a supporting act in the AI era. Historically, semiconductor profit pools were dominated by logic chips (CPU/) and foundry (TSMC), while memory makers earned cyclical, hard-fought margins. But as AI workloads demand extreme bandwidth, capacity, and energy efficiency, memory — especially HBM — has become the bottleneck and value center of the entire AI compute chain.
When SK Hynix's quarterly operating margin surpasses TSMC's, when its HBM capacity becomes the "entry ticket" for Nvidia's next-gen GPUs, and when management sets "KRW 100 trillion net cash" as a long-term target — the global semiconductor power structure is quietly shifting.
For investors, the question is no longer "how many more quarters can this cycle last?" but "is AI infrastructure capex truly a multi-year trend?" Key watchpoints from this report include: whether SOCAMM2 opens a new growth curve beyond HBM, whether 1cnm LPDDR6 builds a new moat in premium smartphones, the quality of AR collection, and the scale of shareholder returns in H2.
Management has given its answer — bet on technology leadership, bet on long-term contracts, bet on AI's longevity, while maintaining a financial reserve to weather any downturn. This Q1 report is just one footnote in that story.
Risk disclaimer: This content is for informational purposes only and does not constitute investment advice. Market risks exist; invest with caution.
On investment analysis, market insight, and asset allocation — not chasing short-term noise, but understanding what truly drives long-term returns.
