SK Hynix shares slid from 2,959,000 won on June 22 to 1,291,000 won on the morning of July 30 — down 56.4% from the peak, and 29% in four sessions from the July 27 close of 1,819,000 won. In the same week, the company reported second-quarter 2026 revenue of 79.32 trillion won and operating profit of 60.54 trillion won, its best quarter ever.
A record quarter and a halved share price arriving in the same month tells you what kind of episode this is. The market was not reacting to the already-settled second-quarter figures, but to the fact that those figures missed expectations and to how little the company said about HBM pricing and volumes from the third quarter onward. What matters here is not the size of the earnings, but the gap between expectation and result — and the sentences that created that gap.
Share prices move not on the absolute size of earnings, but on the moment consensus about their direction breaks.
Three numbers: record quarter, consensus miss, 29% in four days
Revenue rose 256.8% year on year and operating profit 557.2%, with an operating margin of 76.3% — 4.8 percentage points higher than the prior quarter. Even so, as Seoul Shinmun summarised, both revenue and operating profit landed roughly 5% below the brokerage consensus. The gap column below is calculated directly from the reported figures against consensus.
| Item | Q2 2026 reported (trillion won) | Consensus (trillion won) | Gap (%) |
|---|---|---|---|
| Revenue | 79.32 | 83.62 | -5.1 |
| Operating profit | 60.54 | 63.55 | -4.7 |
| Operating margin | 76.3 | - | +4.8pp QoQ |
A 5% miss does not explain a 56% decline. Most of the fall came from the size of the expectations stacked up through June. After a run that more than tripled the stock this year, the price already embedded the premise that HBM volumes and unit prices would keep climbing. When that premise wobbled, the price came down at the speed required to erase it. The sequence of double-digit down days shows the same character.
| Date | Close / intraday (won) | Daily change (%) | vs June 22 peak (%) |
|---|---|---|---|
| June 22 | 2,959,000 | - | 0.0 |
| July 27 | 1,819,000 | - | -38.5 |
| July 28 | - | -16.16 | - |
| July 29 | - | -7.54 | - |
| July 30 (a.m.) | 1,291,000 | -7.85 | -56.4 |
From the July 27 close to the morning of July 30, the decline was 29.0%. Three sessions around the July 28 earnings release produced about half the total drawdown, which is why this episode hangs on the content of the release and its interpretation rather than on industry indicators.
Why the best quarter arrived with the crash
Three threads overlapped. First, reports said the company would moderate the pace of its HBM4 capacity ramp and redirect that capacity to conventional DRAM; the market read this as a signal of softening AI memory demand. Expectations of a cut to next-generation GPU volume forecasts pushed the interpretation further in one direction.
Second, the product mix worked against the company this time. According to Seoul Shinmun, HBM accounts for a larger share of output than at rivals, so rising conventional memory prices fed through to total earnings less powerfully than hoped. In a quarter when commodity prices rise, an HBM-heavy structure takes only part of the upside.
Third, the earnings call offered little concrete detail on HBM price negotiations and long-term supply agreements. When settled results are strong but next-quarter pricing goes unmentioned, the market fills the blank conservatively. The opposite reading exists too: brokerage views relayed by Newsway argued that an HBM-centred production strategy and expanding long-term contracts would prolong the memory shortage, making July's fall an excessive correction. That two opposite conclusions come from the same facts is the crux of the debate. Read through a framework separating EPS, guidance and consensus, this release classifies as strong actuals with empty guidance.
What the target-price cuts reveal
Collecting the revised target prices shows what the market recalculated. The change and gap columns below are computed from the targets compiled by Seoul Economic Daily against the July 30 morning price of 1,291,000 won.
| Brokerage | Previous target (mn won) | Revised target (mn won) | Change (%) | Gap vs 1.291mn (%) |
|---|---|---|---|---|
| Shinhan Investment | 4.20 | 2.70 | -35.7 | +109.1 |
| NH Investment | 4.10 | 3.40 | -17.1 | +163.4 |
| Daishin Securities | 3.90 | 3.20 | -17.9 | +147.9 |
| Samsung Securities | 3.50 | 3.00 | -14.3 | +132.4 |
| Korea Investment | 3.80 | 4.70 | +23.7 | +264.1 |
Two things stand out. One, the houses that cut trimmed targets by 14–36% while the stock fell 56% — the price moved first and further than analyst expectations. Two, direction split: Korea Investment & Securities raised its target, attributing the weak quarter to shipment timing slipping rather than to demand deterioration. A target price is not a forecast but a summary of the assumptions a house is running, so the reason attached to it — lower unit prices, or deferred recognition — is more useful than the number itself.
Conditions that would break this reading
The view that this is an expectations-versus-results gap rather than a collapse in the cycle fails if any of the following shows up.
- Conventional DRAM and NAND spot prices turning down on a quarterly basis — meaning the commodity demand that justified the capacity switch is itself weakening.
- The customer count or volume under HBM4 long-term supply agreements coming in below prior guidance — deferral confirmed as contraction.
- Major customers cutting AI server capital plans in won or dollar terms — unlike a single GPU volume adjustment, that is a cycle signal.
- Inventory growing faster than revenue — more stacking up than shipping out.
The symmetrical case deserves writing down too. If third-quarter guidance holds HBM unit prices or adds contracted customers while conventional prices hold, July's repricing ends as a reset of expectations. Placed beside the conditions that separate a bounce from a trend change, what needs checking is not the size of any rebound but the type of evidence behind it.
What to watch
- The company's first sentence on HBM price negotiations at the third-quarter release — the box that was left blank this time.
- Changes in the number of HBM4 long-term supply customers (around ten by current guidance) and contract durations.
- Monthly direction of DDR5 and NAND contract and spot prices — where the conventional-capacity strategy is judged.
- When target-price cuts stop, and whether the stated reason shifts from unit prices to timing deferral.
- Whether foreign and institutional flows turn net buyers, and how large that is relative to the drawdown.
- Whether the 76.3% operating margin holds — mix shifts show up in margin first.