KRW Stabilized at 1,481, So Why Did SK Hynix Crash 38%?

This piece is the verification chapter of the series, an after-the-fact fact-check. The structure chapter, Memory's Power Reversal: When the AI Cycle Flips Back, covered "why it rose"; the event chapter, KOSPI's 8,000 Breaks: Real Crisis or Just a Fit?, covered "why everything fell together in June"; the ADR report, SK hynix ADR: How High Can the Stock Go?, covered the listing event; and the downside chapter, Foreigners Sold 150T Won: Why SK Hynix Falls First, covered "why it kept cracking on FX through early July." On top of that, a July 6 FX-stabilization note argued that "the won can stabilize on ADR dollar inflows and the like."

Two weeks ago (July 2 and July 6) I made two predictions. First, the won can stabilize on ADR dollar inflows and similar factors. Second, whether the current decline is a "fit (liquidity correction)" or a "real drop (downcycle)" depends on the leading indicators, so watch them. In the meantime the market moved into an entirely new phase. This piece verifies, with data, how far the past scenarios held, where they went wrong, and what the true axis of the crash has shifted to. Figures were cross-checked against local data and reporting from July 16-17, 2026, and estimates are marked as "estimates." Jargon is collected in the glossary at the bottom.

The Won Stopped, So Why Did Chips Fall Even Harder

Two weeks ago this series was titled "chips fall every time FX hits." But now the won has stopped, has actually strengthened, and chips fell even harder. The late-June to early-July crash and the mid-July crash only look alike on the surface (a screen full of blue); the engines are completely different.

On July 16 alone the KOSPI fell -6.37% (6,820), Hynix -11.53% (close 1.842M), and Samsung -8.77% (255K). Foreigners sold -1.39T won and institutions -2.37T won, while retail absorbed all of it with +3.66T won. That day marked the 37th sell-side sidecar of the year.

And Hynix broke the 1.85M-won line, closing around 1.84M on July 16. That 1.85M was the floor and last line of defense of Scenario 1 (liquidity correction), the "next support if 2.0M breaks" that the piece two weeks ago pinpointed. Its collapse on a closing basis is technical confirmation of a decline: the price has left the "simple liquidity correction" band and moved down into the 2/3 zone.

The intraday move was even more decisive. On July 14 Hynix crashed intraday to roughly 1.67-1.68M won (down 9%-plus, breaking the "1.7M Hynix" line) before rebounding V-shaped to 1.913M (+3.69%). That 1.67M is the top of the piece's Scenario 3 (a real downcycle, 1.3-1.7M). In other words, the market had already tagged the downcycle price band once intraday and reversed. The meaning of this "floor test" is handled below, and the downside-and-rebound map is addressed head-on in the support-and-rebound section.

Yet the same day the US was roughly flat (Nasdaq -1.5%, S&P -0.5%). Over the month Korea is down -23% and the US about -3%. This is not a global correction in tandem; it is a shock specific to Korea. And the won held firm even as the index plunged (from 1,550 to 1,481). The "FX-and-stock feedback loop" of the last two pieces has snapped.

These three decouplings (FX, Korea-US, and flows) reveal the true nature of this crash. Let us unpack them one by one.

What Changed: Three Decouplings

FX Decoupling: the FX Loop Switched Off

The core mechanism of the July 2 and July 6 pieces was a feedback loop: "weak won (1,540) → foreign selling → weaker won → more selling." But the actual data moved in exactly the opposite direction.

The won swung about 70 won stronger in two weeks. Yet over that span the KOSPI fell another 11%, from 7,648 to 6,820. If "FX were the culprit," this could not happen. If the won is strengthening while stocks fall harder, the force pulling stocks down is no longer FX.

Korea-US Decoupling: This Time Only Korea Collapsed

The trigger of the late-June crash (the event chapter) was the US. When "Philadelphia Semiconductor -6%, Micron -10%" hit overnight, Samsung and Hynix took the blow at the open the next morning. The US led, and Korea followed.

Mid-July is different. The US (Nasdaq, S&P) is roughly flat while Korea alone fell -23%. Look at July 16 alone: the big US indices were -0.5% to -1.5% while the KOSPI was -6.4%. Korea fell more than 5x the US, a domestically sourced shock. That said, even in the US the chip sector was weak, with Micron and others plunging and the Philadelphia index down -2%. In other words, it was not the whole market but "only the chip group" that was weak even in the US, and because chips are half of Korea's index, Korea took the hit wholesale.

Flow Decoupling: Foreign Selling Actually Eased

The most counterintuitive part. The July 2 piece cited "three straight days of 2T-plus won of foreign net selling" as hard evidence of the crash. Yet foreign flows on the KOSPI in mid-July (in trillions of won) looked like this.

The market plunged even though foreign selling eased and there were more net-buying days. In particular, on July 14 (intraday low 6,448) foreigners bought a net +1.9T won. In other words, this decline was not caused by foreigners "dumping." While retail absorbed +44.7T won over the month (nearly a mirror image of foreigners' -47.5T won), the market collapsed inside a retail-led defensive tape. The old "foreign selling → crash" formula was not the lead actor this time either.

So What Was the Real Trigger: Five Newly Switched-On Forces

If the earlier pieces saw macro-liquidity (FX, rates, flows) as the culprit, the culprit of this crash moved to "the fundamentals narrative of chips themselves (peak-out)." And layered on top were domestic monetary policy and the ADR aftermath.

Semiconductor Peak-Out: the Narrative Finally Cracked

The biggest axis. In the June event chapter I noted that "Broadcom put the first crack in the 'infinite AI demand' story," and in mid-July that crack spread to Korean chips' profit outlook.

First, the analyst estimate cut was the decisive blow. On July 13 Korea Investment & Securities cut SK Hynix's 2026 and 2027 operating-profit estimates by -9% and -11% respectively (reflecting long-term supply agreements, or LTAs). It was the first time the "direction of earnings," which the earlier pieces told you to watch, bent in the numbers. The basis for peak-out worries changed from "a feeling" to "a report." On top of that, on July 16 a brokerage report warning that "Q2 results could come in below market expectations" pulled earnings expectations down further.

Second, the "earnings peak-out" frame appeared. The view that earnings growth slows in 2027 became the market's dominant narrative. The moment the word "peak" is attached to a supercycle winner, a stock that was priced for perfection flips into sell-on-news. The very mechanism from the SanDisk chapter now applies to Hynix and Samsung.

New York's One-Year Data-Center Pause: the First Institutional Brake on the Root of Demand

On July 14, New York's governor (Kathy Hochul) signed an executive order halting new permits for large data centers of 50MW or more for one year (through July 2027). It is the first data-center moratorium at the US state level. The backdrop is resident backlash over rising electricity bills and water depletion. Fourteen states including Georgia, Michigan, and Pennsylvania are reviewing similar rules.

This is the second leading indicator (a hyperscaler capex slowdown) that the July 2 piece told you to watch, arguing that "if the wallet of the side that buys memory (the hyperscalers) dries up, the side that sells shakes too," now made real through regulation. Beneath it lies the outlook for slower capex growth at Google, Amazon, Microsoft, and Meta. Overnight, on this news, the Philadelphia Semiconductor index fell -2%, and Korean memory took it straight at the open.

China's CXMT Listing: the Seed of Oversupply Ignites

On July 16, CXMT, China's largest DRAM maker, lists on Shanghai's STAR Market. The raise is about $8.5B (up to 14-15T won), one of the largest chip IPOs on record. Most of the proceeds go to commodity-DRAM capacity expansion.

This is the first leading indicator (accelerating expansion by the three memory makers plus China's CXMT), which the July 2 piece flagged as "4,700T won of capex, the seed of future oversupply," now lit up by a fundraising event.

It strikes precisely at Korea's vulnerability. Korea concentrates capacity on high-margin HBM and AI servers, and in the gap China moves into commodity DRAM for PCs and general servers. The fact that both Hynix's ADR ($29B) and China's CXMT ($8.5B) are money adding supply on both sides stoked the "2027-2028 oversupply" fear.

That said, there is a lag. Counterpoint and others believe that "from capex to equipment installation, yield ramp, and customer qualification takes time, so it does not translate directly into next-quarter oversupply." In other words, right now it is the "narrative," not the price, that moved first. This distinction is the heart of the leading-indicator re-check.

Bank of Korea Rate Hike: the Double Edge That Fixed FX and Killed Growth Stocks

Here is the variable the earlier pieces underrated the most. At the July 16 meeting the Bank of Korea raised its benchmark rate from 2.50% to 2.75% (the first hike since early 2023). The justification was FX defense and inflation. Experts even see additional hikes in August and October (entry into a tightening phase).

That single move split the market two ways.

It was medicine for the won. By capping the widening Korea-US rate gap, it turned the won stronger (from 1,550 to 1,481, five straight days of gains). This is the real driver of the FX stabilization seen earlier.

It was poison for growth stocks. A rate hike raises the discount rate on high-valuation growth stocks (chips and AI) that lean on distant future earnings. In the event chapter I called rates "the master switch," and this time that switch was pressed not by the Fed but by the Bank of Korea, domestically. This is the core reason for the decoupling in which only Korea fell.

In short, the BOK hike is a single variable that explains both the "FX decoupling" and the "Korea-US decoupling" at once. The FX stabilizing and only the Korean market collapsing share the same root.

ADR Flowback: a Hit Deal That Still Pressed the Local Shares

On July 10 SK Hynix's ADR debuted successfully on the Nasdaq (offer price $149, close $168.49, +13.1%). But a hit deal did not mean a rising local share.

A 16% US premium created flowback pressure. Once the ADR traded about 16% richer than the domestic close, the reverse flow of arbitrageurs selling the cheaper Korean local share (shorting) and buying the ADR grew. Indeed, from June 23 to July 8 Hynix's stock-loan balance surged +31.4% (3x Samsung's +11.7%), circumstantial evidence that foreigners preemptively shorted the local share to target ADR arbitrage.

There was also "event burnout." On the July 13 plunge, analysts diagnosed it as "not a deterioration of the industry, but a volatility correction as the short-term event of the ADR listing burned out and coincided with elevated earnings expectations and the unwinding of leveraged positions." That is exactly the shape of Scenario 2 (ADR event volatility) from the July 2 piece.

Here is what that day actually looked like. July 13 was what the market called "Black Monday." With the KOSPI at -8.95%, a circuit breaker (a 20-minute halt of all trading) triggered at 1:28 p.m., and Hynix's local share, back from its US trip via the ADR, was pushed intraday to 1.87M won (-15%) the moment it returned, breaking the "1.9M Hynix" line. Sidecars then repeated across all four trading days that week, all the way to the 37th sell-side sidecar of the year (July 16).

Verifying the Past Scenarios: How Far Did We Get It Right

This is the heart of the piece. I compare the predictions of the July 2 and July 6 pieces against reality, one by one. Honestly, dividing what was right from what was wrong.

Prediction Scorecard

What Was Right: the Framework Survived

FX prediction (hit). The July 2 piece said "the ADR does not remove FX risk, but $30B of inflows itself is favorable to the won," and the July 6 piece said "the only short-term card Korea holds is ADR dollars." Both were right. That said, the decisive blow that actually turned the won was the BOK rate hike rather than the ADR. The July 6 piece weighted "dollar supply," but in practice "defending the rate gap" mattered more.

Leading-indicator methodology (hit, the biggest win). The July 2 piece said "watch the leading indicators, not the price," and listed five. Of those, capacity additions (the three memory makers plus CXMT) and hyperscaler capex cuts lit up precisely on July 14-16. The CXMT listing and the New York data-center pause are textbook realizations of that checklist. The framework of deciding in advance "what to watch" actually worked.

Support roadmap (hit, and below that floor). The July 2 piece marked two supports in order: "if 2.0M breaks, next is 1.85M." Indeed, amid the July 13 circuit breaker the 2.0M line crumbled to 1.87M intraday (breaking the 1.9M Hynix line), and on July 16 it closed at 1.842M, breaking through 1.85M. The location and order of the supports were exactly right. However, it went one step further than expected, leaving even the floor of Scenario 1 (1.85M) on a closing basis. That is both a hit and, at the same time, a signal that "price went below 1 and lower." ADR flowback (Scenario 2) also materialized with the stock-loan balance +31%.

What Was Wrong: the Headline Was Disproven

Reversal of causation (the most important reflection). The series' flagship claim, "chips fall every time FX hits," was right in late June and early July but broke in mid-July, because the won strengthened while chips fell harder. The lesson is clear: FX was not the "cause" of that phase but a "co-symptom." The real cause (deleveraging and peak-out) shook FX and stocks at the same time. Read a correlation as causation, and the moment the cause changes, your prediction flips. This is that case.

The actor behind rates (missed). The July 2 piece saw rates only as "the Fed is exogenous, Korea cannot control it." But the actual trigger was the BOK's active hike. The master switch Korea supposedly could not control was pressed by Korea itself (for the purpose of FX defense), pressing growth stocks down. The earlier piece underrated this "domestic monetary-policy risk."

Underrating the "bad" side of the ADR's two faces (missed). The July 2 piece treated the ADR somewhat lightly, as "an FX helper plus one layer of temporary overhang." In reality, even though the ADR was a hit (+13%), a 16% premium drove local-share shorting (stock-loan balance +31%), and the downside pressure was larger than expected.

The Most Important Thing: 1.85M Broke

This is the core you cannot miss in the verification. The July 2 piece drew Hynix's support structure as "first 2.0M, next 1.85M," and set the band of Scenario 1 (liquidity correction, the main case at 40% probability) at 1.85M-2.15M. The 1.85M was the floor of that band, the last line marking "up to here it is a simple correction."

That 1.85M broke on the July 16 close (1.842M). Why is that decisive?

First, the shelf life of the "simple liquidity correction" reading ran out. While the price sat inside band 1, you could read it as "a healthy correction giving back an expensive spike." But once it leaves that floor on a closing basis, only two readings remain: 2, a peak-out de-rating, or 3, the onset of a downcycle. The break of 1.85M is an event in which the price itself confirmed that the center of gravity had moved "from 1 to 2/3."

Second, support flips to resistance. In the July 2 piece's terms, a broken support (1.85M) now becomes a wall of supply (resistance) on the way back up. Reclaiming 1.85M becomes the first gate of the next rebound.

Third, clear support below is thin. The 1.85M was the floor of the June base (1.85M-2.16M, June 8 low 1.853M). Break it and the area below is essentially empty until the next meaningful support (Scenario 3, 1.3-1.7M). With a thin cushion, if stop-losses and mechanical selling attach, it can slide fast (the 37th sidecar is the evidence of that volatility).

To sum up, the coordinates of the supports (2.0M and 1.85M) were exactly right, but the market broke through even that last line and came down. The "map" of the prediction was correct, and the actual price took the darkest path on that map. Further, on July 14 it was pushed intraday to 1.67M, rehearsing even the top of Scenario 3 (1.3-1.7M), though it reversed to 1.91M within the day. So the question now is not "does 1 hold?" but "does it stop at 2, or reach 3?"

A Fit or a Real Drop: Re-checking the Leading Indicators

I re-grade the July 2 piece's five-item leading-indicator checklist today, two weeks later. This is what divides the nature of this crash.

The diagnosis is this. Of the five, the first and second have switched on, and a new warning light, estimate cuts, has been added. But the "hard evidence of price," the third through fifth (book-to-bill, spot price, HBM ASP), has not rolled over yet.

Here is what that combination means.

We have stepped up one level, from "a phase where only the price fell (a fit)" to "a phase where the narrative and expectations have started to bend." The lighting of the first and second is exactly the turning point the July 2 piece described as "once these switch on, it moves from 1 to 3."

But it is not yet confirmed as a "real downcycle (3)." The decisive evidence of a downcycle is the spot price, HBM ASP, and order backlog, and those still have not bent. What has moved is the expectation that "future supply rises (CXMT)" and "future demand slows (data centers)," not the measured fact that "today's price fell."

In short, right now it is a gray zone straddling the border of "a fit" and "a real drop." The narrative has tilted toward 3 and the price broke the floor of 1 (1.85M), but the final evidence (spot price, HBM ASP, order backlog) still has not bent. It is a state where "the narrative, flows, and chart have conceded the decline, and only the fundamental hard data is unconfirmed."

There is one technical confirmation too. In a local screener, the July 17 "bullish alignment" (uptrend) list was swapped wholesale for financials, insurers, inverse ETFs, and defensives. Samsung and Hynix dropped out, and in their place an "SK Hynix Inverse 2X" ETF entered the bullish-alignment list. An inverse Hynix trending up means the local share's trend has clearly bent (the "bullish alignment captures the trend" signal from the data chapter switched on in reverse). Flows and the technicals have already conceded the downtrend, and the remaining question is whether the fundamentals (price) follow.

So How Far: Updated Scenarios

I recalibrate the July 2 piece's four scenarios to reality two weeks on. With 1.85M broken, the core question itself has changed. It is no longer "does 1 (liquidity correction) hold?" but "does 1.842M stop inside the 2 (peak-out de-rating) band, or does the spot price roll over and take it to 3 (1.3-1.7M)?"

Here is the direct answer to "how far."

The price is already inside the 2 band (peak-out de-rating, 1.5-2.0M) at 1.842M. Once 1.85M broke, the "pure liquidity correction" logic of 1 lost force (the FX driver of 1 has also stabilized and faded), and now a re-rating that says "the peak is past" (a de-rating) presses the stock. The 1 that was the main case two weeks ago has now been demoted to "a retrace scenario that only holds if 1.85M is reclaimed."

The real downcycle of 3 (1.30M-1.70M) still opens only if "the hard price indicators (3-5) bend." That signal is not here yet. CXMT is a narrative, not yet supply. But the 1.85M break has thinned the cushion to 3. This is the only gate dividing 2 and 3, so watch whether spot DRAM and HBM contract prices actually roll over.

The 4 (re-rating higher) has lower odds but is not dead. Peak-out is still an "expectation," not a "measurement." If August earnings confirm HBM demand and 2027 guidance as firm and CXMT's real supply is delayed, today's -38% can reverse into "a correction that shook out expensive expectations." But for that, 1.85M must be reclaimed first.

Watch triggers (next two weeks). First, spot DRAM and HBM contract prices (a roll confirms 3). Second, the 2027 guidance in August earnings from Hynix, Samsung, and Micron. Third, whether the BOK hikes again in August (added pressure on growth stocks). Fourth, a reversal in the ADR-vs-local premium/discount (a sign flowback is exhausting). Fifth, the spread of more data-center-regulating states.

How Far It Falls and When the Rebound Comes: a Support-and-Rebound Map by Name

I answer the two most-asked questions head-on. First, the true floor test of this phase was not the July 16 close (1.84M) but the July 14 intraday low (1.67M).

Here is the trajectory of this week's crash, day by day.

Intraday It Already Tagged the Downcycle: the July 14 Capitulation

The cause that day was a temporary geopolitical shock: a re-ignition of Middle East (Hormuz) risk (a US-Iran military clash and a maritime blockade). The key point is that Hynix was pushed intraday to 1.67M (about -43% from the 2.96M peak) and reversed to 1.91M within a single day. That is the classic fingerprint of a selling climax. In other words, the market tagged the "real downcycle price (3)" once and rejected it.

How Far Does Hynix Fall

These are the support-and-resistance coordinates based on the current price (about 1.84M, intraday low 1.67M).

The direct answer is this. The most probable downside support is 1.67-1.70M won (the July 14 low already defended). Ranging inside this box (1.67-1.85M) and building a base is the main picture (2). If 1.67M breaks on a closing basis, the real downcycle (3) of 1.48M to 1.30M opens, but that is a conditional scenario requiring spot DRAM and HBM prices to actually roll over (not yet).

Samsung Electronics and the Chip Sector

Samsung (currently about 255K, -30% from the 363.5K peak) has a shallower drawdown than Hynix. It is diversified across foundry, mobile, and appliances, so its pure exposure to the memory cycle is lower (even in the 2018-19 downcycle, Samsung's operating profit fell -53%, shallower than Hynix's -87%).

At the sector and index level, the KOSPI's front line of support is the July 14 intraday low of 6,448, and below that is 6,000 (psychological). Because Korean chips move in sync with US Micron and the Philadelphia index, if US memory names roll over further, Hynix's 1.67M line gets retested.

Can a Rebound Come

You have to distinguish two kinds of rebound.

First, a short technical rebound (a bear market rally). It is already underway and likely to repeat. The July 14 V-shape (1.67M to 1.91M, +14%) and July 15's +6.2% are the evidence. With oversold conditions, a capitulation wick (1.67M), and institutional and foreign dip-buying underneath, short and sharp rebounds can come at any time. But with high beta and leverage, they are violent both ways (the 37th sidecar).

Second, a trend reversal (a real bottom). Not confirmed yet. The decisive point is that the July 15 +6.2% rebound was undone by July 16's -6.4% (a failed follow-through). Right now it is "a rebound within a downtrend," not "a bottom and re-rating higher."

Here are the signals, in priority order, that must switch on for a rebound to become a "trend."

1. A close back above 1.85M won. You must reclaim the broken first resistance to begin a technical rebound (Samsung: reclaim 280K).
2. Confirmation that DRAM spot and HBM ASP stabilize. The only fundamental signal proving peak-out was "overdone fear" rather than a "measurement."
3. Firm 2027 HBM guidance in August earnings. If demand is confirmed in Hynix, Samsung, and Micron results, the de-rating stops.
4. An end to further BOK hikes and a delay in CXMT's real supply. The growth-stock discount pressure and the oversupply narrative ease at once.
5. Exhaustion of the ADR premium and flowback. Once the stock-loan balance peaks and rolls over, local-share short pressure eases.

To sum up, short trading rebounds repeat with high probability (inside the 1.67-1.85M box), but for a "V-shaped re-rating" to solidify, the fundamental confirmation of items 2 and 3 above (spot price and earnings) is essential. Until then, it is a phase that splits into a W-shape (double bottom) attempt if 1.67M holds, or an L-shape (further decline) if 1.67M gives way on a close. The most probable picture is "a high-volatility range trade inside the 1.67-1.85M box, swinging hard while waiting for August earnings," and the direction is set by August earnings and DRAM spot prices.

These are high-beta, high-leverage names, so they move about ±10% a day even against the trend. The coordinates above are probabilistic support and resistance, not a "bottom declaration."

In One Line

Two weeks ago I wrote "chips fall every time FX hits." Yet in mid-July, FX stabilized just as predicted (1,550 to 1,481) while chips fell even harder (Hynix -38%, KOSPI -23% for the month). The reason is that the engine of the crash changed. The axis moved from "FX and foreign selling (macro-liquidity)" to "the chip peak-out narrative + a BOK rate hike + CXMT oversupply + the New York data-center brake + ADR flowback." The past piece's flagship claim (FX is the culprit) was disproven, but its method (watch the leading indicators, support 2.0M then 1.85M) nailed it: the two indicators it told you to watch switched on, and the support coordinates were right. And the market broke even that last support of 1.85M on a closing basis (7/16, 1.842M). Right now it is the border of "a fit" and "a real drop." The narrative, flows, and chart have conceded the decline, but the final evidence, spot prices and HBM ASP, still has not bent. Hynix at 1.842M has left the 1 floor and sits inside band 2 (peak-out de-rating, 1.5-2.0M), and ahead is a fork: 1.30M-1.70M (3) if spot prices roll over, or a re-rating (4) if peak-out proves to be overdone fear (reclaim 1.85M first). FX is no longer the answer. What to watch now is DRAM spot prices and August earnings.

Glossary

Disclaimer

This article is research material for informational purposes, compiled and cross-checked from public reporting, statistics, and local data. It is not investment advice recommending the purchase, sale, or holding of any specific security. The companies mentioned (SK Hynix, Samsung Electronics, Micron, CXMT, and others) are analytical examples, and the author may hold some of these securities. Hynix's July 16 close ranges by data source from 1,830,000 to 1,842,000 won (-9.5% to -11.5% on the day, intraday low around 1.80M), and Samsung around 255K, with slight differences by reporting and data source. The July 14 intraday low was about 1.67-1.68M won (after breaking the "1.7M Hynix" line, it rebounded to close at 1.913M), and the July 13 intraday low was about 1.87M (breaking the "1.9M Hynix" line and triggering a circuit breaker). Samsung's July 14 intraday low (240K range) is an estimate. Support and resistance coordinates, scenario price bands and probabilities, and the leading-indicator status calls are subjective estimates based on precedent and mechanism, and they change frequently. All investing carries the risk of principal loss, and investment decisions and their consequences rest with the investor.

References

Mid-July crash market action and causes

• KOSPI plunges 7%, breaks 7,000 for the first time in two months, SK Hynix down 12%-plus · Etoday (7/13)
• SK Hynix plunges 15% in a direct hit, "Black Monday" KOSPI drops to the 6,800s · Hankookilbo (7/13)
• Chips plunge back to "6,000-KOSPI," peak-out and war fears press the market · Kyunghyang (7/13)
• KOSPI circuit breaker triggered, "1.9M Hynix" broke intraday · YTN (7/13)
• KOSPI circuit breaker on Samsung-Hynix plunge, 6,800 line · Financial News (7/13)
• SK Hynix Tumbles 15% in South Korea, Leading Semiconductor Drop · The Motley Fool (7/13)
• SK Hynix down 11%, the 37th KOSPI sidecar of 2026 · BeInCrypto (7/16)
• SK Hynix breaks even 1.7M won, KOSPI plunges 5%-plus to the 6,400s · Herald Economy (7/14)
• "Samsung-Hynix" rollercoaster again on the 14th, down 9% then rebounding 3%-plus (Hynix intraday 1.678M) · Getnews (7/14)
• SK Hynix rebounds 4% to reclaim 1.9M won, Samsung up 5% · News1 (7/14)
• "Samsung-Hynix" defend the index, KOSPI shakes off the plunge to close up 0.7% · Edaily (7/14)
• KOSPI down 6.4% on chip-peak worries, rate hike pressures growth stocks · Hankookilbo (7/16)
• Back to "1.8M Hynix and 255K Samsung" in three days, KOSPI -7% · Seoul Shinmun (7/16)
• Buy to sell, sidecars at the drop of a hat, the fallen Samsung-Hynix · Money Today (7/16)

Bank of Korea rate hike

• BOK raises benchmark rate 2.50 to 2.75% (7/16 meeting), the first hike since early 2023 · Bank of Korea
• Tightening begins, experts see "an August or October hike after July" · Financial News (7/12)

China CXMT listing

• China's CXMT starts a record 15T-won IPO, commodity-DRAM "oversupply" fears · The-Sen News
• China's largest DRAM maker CXMT lists, a "watershed" for chip self-reliance · Hankookilbo (7/16)
• CXMT targets the Big 3 with an $8.5B IPO · Counterpoint Research

New York data-center pause

• New York brakes AI data centers, the first US halt to new construction for a year · Seoul Economic Daily (7/15)
• New York executive order halts large data-center construction for a year · Electronic Times (7/15)

SK Hynix ADR and flowback

• SK Hynix ADR debuts successfully on the Nasdaq ($168.49, +13.1%), can the Korean local share ride the wave? · Seoul Economic Daily
• Implications and investment strategy of the SK Hynix ADR listing · Shinhan Financial

Data were compiled by cross-checking local daily/weekly candles for SK Hynix and Samsung Electronics and the KOSPI index (7/16 close 6,820.6, USD/KRW 1,481.6, daily foreign flows).