After the "strongest in history," Hynix has found it hard to exceed expectations.

```

SK Hynix delivered its "strongest ever" financial report today, yet it also became the most nuanced miss amid the AI memory bull market.

In fiscal year 2026 Q2, Hynix recorded revenue of 79.32 trillion KRW, up 257% year-on-year and 51% quarter-on-quarter—ostensibly impeccable, but about 5.5% below market expectations (consensus: 83.9 trillion KRW). Operating profit was 60.54 trillion KRW, soaring 557% YoY and 61% QoQ, but also about 5.7% below consensus (64.2 trillion KRW).

Operating profit margin reached 76.3%, setting a new record high for the memory industry; gross margin hit 83%—already surpassing Nvidia (about 75%) and approaching software platform giants like Meta (about 82%).

Even more intriguing is the market reaction: before the report, Hynix's share price had fallen about 47% from its June 22 peak. After the report, its US ADR dropped more than 5% before rebounding; in Korea, shares opened up 3-4% the next day before plummeting.

In short: The absolute numbers are historic, but flexibility fell short of expectations; the market is in a tug-of-war between "strongest ever" and "disappointing performance."

Revenue and Operating Profit Margin Trend

Why did profits jump 557% but still miss? LTA is the answer

The root of Hynix's miss lies in the ultra-high proportion of HBM revenue.

LTA (Long-Term Agreement, 3-5 year long-term supply contracts) are long-term procurement deals AI chip customers sign with memory vendors like Hynix, with prices locked at the point of signing and not fluctuating monthly like the spot market. For Nvidia, the biggest buyer of AI training chips, this lock-in guarantees supply stability; but for Hynix, it's a double-edged sword—as the average DRAM sales price (ASP) QoQ increase dropped sharply from about 60% in Q1 to 30% in Q2, and NAND ASP growth fell from about 70% to 50-55%.

The arithmetic is simple: about 50% of Hynix revenue is locked in by LTA, which happens to come from its highest HBM share. Meanwhile, spot DRAM prices kept soaring within the same quarter—according to TrendForce, generic DRAM contract prices rose 58-63% QoQ in Q2, NAND prices rose 70-75%, yet Hynix's actual ASP growth was only about half that, meaning Hynix "ate" the volume increase but didn't fully "eat" the price hike elasticity. KIS Securities already warned of this in its July 13th report, lowering Q2 operating profit projections to 60.4 trillion KRW—the actual 60.54 trillion KRW, almost spot on.

It should be emphasized, the miss isn't due to business deterioration. On the contrary, operating profit margin improved from 71.5% in Q1 to 76.3%, with record highs in absolute profit, revenue, gross and operating margins. The issue is: the market based expectations on spot price hikes, while Hynix chose a steadier, less exciting LTA path.

ASP Comparison of the Big Three Memory Giants

The Watershed for the Big Three: Samsung, Micron beat, Hynix miss

In this super cycle for memory, the three giants unexpectedly diverged in performance.

In Q2, Samsung Electronics posted revenue of 171 trillion KRW and operating profit of 89.4 trillion KRW, beating consensus on both revenue and profit. Its DRAM ASP rose over 40% QoQ, NAND ASP reached mid-60%, and HBM4 has been in mass production since February, with sales exceeding $1.2 billion. Samsung once lagged in HBM but is now catching up with its group-wide resources.

Micron's Q3 financials were a decisive win: revenue of $41.46 billion, 16% above estimates; non-GAAP EPS of $25.11, beating by 22.5%; gross margin 84.9%; Q4 guidance for $50 billion in revenue (±$1 billion) and about 86% gross margin. DRAM ASP surged over 60% QoQ, NAND ASP up about 85%, both far outperforming Hynix.

At the company-wide level, Hynix had the highest operating margin (76.3% vs. Samsung Group 52% vs. Micron ~68%), but was the only one to miss consensus this quarter. Notably, Samsung Semiconductor (DS) division margin is estimated around 80%, higher than Hynix—but the group's mobile, display, etc., dilute the overall rate. This contrast shows: HBM dominance delivers the highest company margin and most stable customer ties, but strong LTA lock-in most compresses spot price upside. In short, Hynix traded ASP elasticity for certainty.

HBM4 Mass Production & 2H Ramp-up: Catalyst for ASP Elasticity Recovery

If LTA caused the Q2 miss, then the HBM4 ramp is the key variable that could reverse expectations in 2H24.

Hynix confirmed in its report that HBM4 started mass production and shipping in Q2, with shipments to increase through the rest of the year; advanced HBM4E samples have been delivered to clients in 2026 H1. KIS Securities projects that as HBM4 ramps from Q3, Hynix’s ASP will realign closer to market averages, meaning the QoQ DRAM ASP growth could recover from 30%. The "discount" imposed by LTAs might be partially offset.

On the demand side, Nvidia’s next-gen AI accelerator (Vera Rubin) ramp-up will directly stimulate HBM4 demand; also, Agentic AI applications are shifting AI memory demand from training to inference. Hynix management stressed that current customer demand still outstrips supply, with their key advantage being the ability to meet massive shipment requirements. In other words, the ASP discount in Q2 was not due to a lack of orders, but because contracted prices are simply below spot—the better the HBM4 shipment mix, the more upside there is for unit revenue.

On supply, Hynix’s capex continues to ramp up: 2026 full-year capex is seen at the high end of 40-50 trillion KRW, mainly for speeding up M15X DRAM fab mass production, P&T7 advanced packaging, M17 NAND base, and the Yongin semiconductor cluster starting production in early 2027. By the end of Q2, Hynix had cash reserves swelling to 88 trillion KRW, with net cash of 69.4 trillion KRW (about $45.6 billion); its balance sheet has completely recovered from the cycle bottom.

The LTA Model Is Reshaping Memory Sector Valuation Logic

In the short term, LTA means Hynix lags Samsung and Micron in this round of spot price surges; in the long term, it might change the valuation approach for the memory sector.

Traditionally, the memory industry has been valued as a highly cyclical space: ASP fluctuations drive margin swings, which in turn determine share price volatility. But widespread use of LTA is smoothing out revenue curves—50% of Hynix’s revenue is under LTA, Micron has already signed 16 SCAs (strategic customer agreements) with a goal of putting over 50% of revenue under long-term deals, and Samsung is fast-tracking a similar model.

As ASP volatility is compressed, the market anchor point shifts from "single-quarter ASP increase" to "high earnings sustainability." For leaders like Hynix and Micron, this means the valuation anchor could shift from "cyclical stocks" to "AI infrastructure assets." The trade-off: in some super-cycle quarters, share price elasticity will look "less exciting"; investors need to adapt to profit curves that resemble cloud services or semiconductor equipment firms, not legacy commodity memory vendors.

Looking back at the financials, one number stands out: Hynix Q2 net profit was 93.92 trillion KRW with a net margin of 118%—over 100% because of 62.17 trillion KRW (pre-tax) in non-operating gains from selling Kioxia shares. At a company tax rate of about 23.5%, the post-tax effect is about 47.6 trillion KRW; after deduction, core operating net profit is about 46.3 trillion KRW. This anomaly doesn't affect core business judgment but should be noted to avoid mistaking the non-operating margin for operating net profit margin above 100%.

Closely Watch ASP Increase After HBM4 Ramp-up

The real highlight in Hynix’s report isn’t the 557% profit growth, but the change in industry pricing mechanisms as the AI memory bull market enters its second stage.

For next quarter, the focus shouldn’t be whether revenue breaks new records, but whether HBM4 ramp can return ASP growth to Samsung and Micron’s levels.

If Q3 DRAM ASP QoQ growth rebounds to 40-50%, then Q2's miss will be redefined as a "transitional discount"; if ASP stays around 30%, the market may start formally discussing whether LTA will be a long-term ceiling on memory valuations.

Either way, the memory sector’s valuation logic is already completely different from two years ago.

Risk Warning and DisclaimerThe market has risks, investments need caution. This article does not constitute investment advice and does not consider individual investors' specific objectives, financial status, or needs. Readers should consider whether any opinions, views, or conclusions in this article fit their circumstances. Investing accordingly is at your own risk. ```