What Actually Crashed Samsung Wasn't Earnings — It Was the Moat. Here's My Read on the Meltdown.

## Short-Term Outlook (Next 1–6 Months)

The nearest checkpoint is the earnings call at 10 a.m. on July 30, 2026, just two days from now. That's where we'll get a first confirmation of where the July 7 preliminary figures — 171 trillion won in revenue, 89.4 trillion won in operating profit — land after external audit. I don't think the confirmed numbers are likely to diverge sharply from the preliminary ones, but I think it's more important that the market's attention has already moved elsewhere. The real variable in this call isn't the numbers — it's the language. How management characterizes China's equipment self-sufficiency and commodity DRAM pricing, and how much visibility they're willing to give into HBM supply contracts, is what I expect to drive the stock that day.

The second checkpoint is the Q3 DRAM contract price. Per TrendForce, Samsung is negotiating with customers targeting up to a 20% quarter-over-quarter increase in Q3 DRAM average selling prices, while TrendForce's own forecast sits lower, at 13%-18%. Compared with the more-than-90% rise in Q1 and the estimated 50%-60% rise in Q2, a meaningful deceleration is already baked in. If the actual contracted prices land close to 20%, that strengthens the case that yesterday's decline was excessive; if they come in below 13%, the peak-earnings debate will follow this stock through the entire third quarter. I think this single line item will end up being a stronger signal over the coming months than any single analyst comment.

The third factor is how the information vacuum gets filled. The price targets on record before the crash were Morgan Stanley's 248,000 won, Macquarie's 340,000 won, and Citi's 460,000 won, and whether these have been adjusted since yesterday's crash is not officially confirmed. In other words, the market currently has no official post-event valuation baseline. In vacuums like this, rumor and guesswork tend to set the price, which is why volatility tends to run higher than usual. I think the point at which the first revised analyst reports emerge is likely to be the real dividing line for confirming a short-term bottom.

The fourth factor is follow-on reporting out of China. As TrendForce laid out, the first shipments of domestically produced immersion DUV equipment are expected to reach SMIC, CXMT, and Hua Hong before year-end. If those deliveries are actually confirmed, yesterday's shock could repeat itself; conversely, if the timeline slips or early yield problems surface, some of that repricing could reverse. I think Korean semiconductor stocks are going to be more sensitive to equipment-related headlines than to earnings for the next few months, simply because yesterday taught the market that sensitivity.

## Medium-Term Outlook (6 Months to 2 Years)

The core medium-term variable compresses down to one thing: whether China actually hits its roughly-20-machine production target for 2027. If the slope from five to 20 holds, then at some point in 2027 it becomes arithmetically plausible for a legacy DRAM line to be filled predominantly with domestic Chinese equipment. Conversely, if component-import dependence and yield issues mean actual shipments fall short of that target, a meaningful share of yesterday's repricing has room to reverse. I expect data to settle which of these two branches is correct sometime around the first half of 2027, and until then, I think this stock's valuation multiple is likely to sit lower than it used to.

The second medium-term variable is CXMT's financial firepower. The company raised 57.919 billion yuan through its Shanghai STAR Market IPO and, on its debut day, reached a market cap of roughly 3.3 trillion yuan (about $488 billion), the largest on China's A-share market. Given that its 2025 global DRAM share, per its own IPO prospectus, was 7.67%, that kind of capital could translate directly into capacity expansion. The moment that expansion couples with domestic equipment supply, commodity DRAM starts to face oversupply pressure. I expect the price curves for commodity DRAM and HBM to visibly diverge sometime around 2027.

The third variable is Samsung's own investment cycle. Some brokerage estimates put the company's total 2025 capital expenditure in the low 50-trillion-won range, with memory-focused investment reportedly continuing to expand in 2026. The trouble is that bigger investment means bigger depreciation, which makes profit fall faster during a price downturn. I take Samsung Securities analyst Cho A-in's comment yesterday — that the market is starting to focus on financial soundness amid expanding AI-infrastructure investment — as pointing at the same thing. Going forward, I expect capex and depreciation to matter almost as much as operating profit itself in this company's quarterly results.

Fourth, the bullish forecasts that existed before the crash will each be tested one by one over the medium term. Citi projected 2026 DRAM and NAND average selling prices rising 171% and 127% respectively, with Samsung's operating profit reaching 331 trillion won; Macquarie projected memory strength lasting through 2028. These forecasts represent the picture before the China-equipment variable was priced in. That doesn't mean they should all be thrown out, either. I think the actual outcome will most likely land somewhere between these bullish forecasts and yesterday's fear, and where exactly that landing point falls will explain most of this stock's price action over the next two years.

## Long-Term Outlook (2–5 Years)

Over the long term, the thing I care about most is where the moat moves next. In the three-layer structure I described earlier, if layer one — equipment access — weakens, what's left as a line of defense is yield, packaging, and customer qualification. That line of defense is still powerful, but it's a different kind of powerful. An equipment barrier is a wall that keeps competitors out of the market entirely; a yield barrier is merely a cost advantage over a competitor who is already in. I expect the memory industry of the late 2020s to gradually shift from a business built on exclusive position to one built on cost competitiveness, and I expect average margin levels and margin volatility to get re-rated together as part of that shift.

The second long-term picture is polarization of product mix. As China's presence grows in commodity DRAM, Korean makers will have little choice but to shift their center of gravity further toward HBM and premium server products. HBM is still reported to be split roughly evenly between SK Hynix and Samsung, which gives it real near-term defensibility. But because the absolute scale of revenue still comes from commodity products, how much mix improvement can offset volume decline becomes the key question. I don't think that offset will be complete, which is why I think it's safer to anchor long-term profit expectations below the peak of the current super-cycle rather than at it.

The third piece is a structural problem with the Korean stock market as a whole. The scene from yesterday — the KOSPI falling 10.84% in a single day, with Samsung SDI, LG Innotek, Seoul Semiconductor, and LG Chem all collapsing at once — laid bare just how concentrated this market is around a single narrative. An index that's effectively synchronized with one industry's cycle acts as leverage on the way up and as contagion on the way down. I think this concentration itself could get priced into the risk premium foreign capital demands over the next several years. It's not a single-stock problem — it's a market-design problem.

Looking back at history sharpens where yesterday sits. Per multiple media reports, Samsung's -13.39% is the largest single-day decline since -13.76% on October 24, 2008, and it broke the -12.31% record set as recently as June 23 of this year. That day in 2008 was a moment when the financial system itself was shaking, while yesterday was a day when a question was raised about industrial structure, not finance — the two are different in nature. But the two events share one thing in common: a premise long taken for granted got put up for re-examination within a single day. I have rarely seen that kind of re-examination wrap up within a few weeks.

The knock-on effects are also worth noting. First, commodity DRAM pricing comes under pressure; second, memory makers' capex plans get adjusted; third, orders at semiconductor equipment and materials suppliers start to wobble. I think LG Innotek's 16.29% drop and Seoul Semiconductor's 8.78% drop yesterday are evidence that the market had already run this domino chain through its head. Push one step further and regional economies, employment, and even national tax revenue get affected. That's why I think this issue needs to be framed not as one stock's price problem, but as a question about the positioning of Korean manufacturing as a whole.

## Scenarios — Framed as Conditions, Not Probabilities

Let's lay out the bullish picture first. If confirmed earnings on the July 30 call land near the preliminary figures, HBM supply visibility gets reconfirmed, Q3 DRAM contract prices settle close to Samsung's 20% target, and on top of that Chinese equipment deliveries slip this year or early yield problems surface — the conditions would be in place for a substantial reversal of yesterday's decline. Morningstar's Kai Wang's comment yesterday that "we believe the sell-off today is largely a knee-jerk reaction and overdone" fits closer to this picture. That said, I'm not going to attach a specific probability or price target to it. The moment I invent a number that isn't in my source material, that stops being analysis and becomes fiction.

The neutral picture looks like this: Q3 contract prices settle in the 13%-18% range TrendForce forecasts, Chinese equipment ships in small volumes as planned but its actual contribution to mass production stays minimal. In that case, the stock likely trades in a wide range, swinging hard on every headline, caught between the support of record earnings and the pressure of an eroded entry barrier. The bearish picture involves confirmed earnings coming in below the preliminary figures, the Q3 price increase falling short of double digits, or signs emerging that the 2027 target of 20 machines is being pulled forward. If two or more of these three overlap, the interpretation that yesterday's decline was the starting point rather than the endpoint gains real force.

## Where I Could Be Wrong, and What I'd Actually Do

To be intellectually honest, I need to write down the conditions under which my view breaks. First, domestic production of a single category of lithography equipment doesn't automatically mean the collapse of memory's entry barrier. Factor in the cost burden of multi-patterning, dependence on imported components, and the time required for qualification, and the point at which real competitive pressure arrives could be much later than I expect. Second, if AI infrastructure demand keeps beating expectations, prices could hold up even as supply grows somewhat. If both of those hold true together, yesterday's crash will, a few months from now, just be recorded as noise — and in that case, I'll admit my reading here was overly cautious.

Finally, the practical suggestion I can offer readers is simple. Instead of watching the price chart, I'd recommend marking three verifiable facts on a calendar and checking them off one by one: the Q3 DRAM contract price, whether Chinese equipment is actually delivered, and whether price-target revision reports come out after the crash. Let me repeat once more: this piece is meant to share information and perspective, not to recommend buying or selling any specific stock, and the Q2 earnings cited here are preliminary, pre-audit figures subject to change. Responsibility for any investment decision and its outcome rests entirely with the reader, and I'd strongly encourage confirming official disclosures and a qualified professional's opinion before making any actual decision. I don't have the answer — I've simply tried to transcribe, as accurately as I can, what the market was asking yesterday.