SK Hynix, Samsung and the KOSPI's 21% Rout: What Peak Buyers Should Do Now

Assessment as of October 16, 2026, after the Seoul close. Prices are Korea Exchange closes in won: SK Hynix ₩1,272,000, Samsung Electronics ₩215,500, KOSPI 5,668.05 (Investing.com; Korea Times). Percentage moves and break-even figures are the author's calculations.

The Short Answer

The memory boom is intact. Investors who bought near the top paid too much for it, and the fundamentals do not change that. During seven sessions from October 6 through October 15, the KOSPI lost 20.7%, rounded to 21%, from its October 2 close of 7,003.74 to 5,554.22. Over the same stretch, SK Hynix fell 31.5% from its October 2 close of ₩1,841,000, and Samsung fell 23.9% from ₩276,000 (Korea Times; Investing.com). Brokers kept raising next-year earnings forecasts while the shares fell (Korea Times), though a stronger won has trimmed SK Hynix's third-quarter consensus. That makes this mainly a positioning and valuation shock, not an earnings collapse. It also does not mean the June highs are coming back.

My verdicts for investors who bought near the highs: SK Hynix, REDUCE if the position exceeds your current risk budget; otherwise HOLD through October 28 results. Samsung Electronics, HOLD. KOSPI, HOLD unleveraged index exposure and cut leveraged ETFs. The HOLD calls assume diversified positions and the ability to bear further losses without needing to sell for cash. The same criteria apply to June and September buyers. Neither stock is a BUY for investors who are averaging down.

Key Takeaways

• SK Hynix sits 57.4% below its June 25 intraday high of ₩2,987,000. A buyer at that peak needs a 135% gain to break even. Samsung sits 42.5% below its June 19 high of ₩374,500, so a peak buyer needs 74%.
• On October 15, the KOSPI fell 7.57% to 5,554.22, which triggered circuit breakers. Foreign investors sold a net ₩18.07 trillion over eight sessions (Seoul Economic Daily).
• Forward P/E multiples of about 3x look cheap. Morgan Stanley reportedly warns that memory stocks often reach their lowest multiples at peak earnings.

Five Phases, One Round Trip

Phase	Period	SK Hynix	Samsung	KOSPI	What buyers were paying for
1. Early optimism	Sept. 2025–Apr. 2026	Crossed ₩1.1M in mid-April	First $1T valuation, Feb. 26	6,307.27 close, Feb. 26	Recognition of a memory upcycle
2. Peak euphoria	May–June 2026	₩2,987,000 intraday, June 25	₩374,500 intraday, June 19	First close above 7,000, May 6; 9,063.84 record close, June 18	A supercycle with no visible end, financed with leverage
3. Leverage crash	July 2026	Down more than 46% from the high after a Q2 miss	Sold off with the market	5,593.56 close, July 29	Forced deleveraging
4. Secondary rally	July 31–Oct. 2	₩1,868,000 close, Sept. 21	₩285,500 close, Sept. 23	7,003.74 close, Oct. 2	Record exports, share buybacks
5. Correction	Oct. 6–15	₩1,262,000 (−31.5%)	₩210,000 (−23.9%)	5,554.22 (−21%)	Inflation, AI financing fears, foreign selling

Sources: TradingKey, Korea JoongAng Daily, Yahoo Finance, TradingKey Q2 coverage, KED Global, Seoul Economic Daily, Investing.com.

Phase 2 matters most for today's losses. Borrowed money helped build the run to 9,000: program trading nearly doubled after single-stock leveraged ETFs on Samsung and SK Hynix launched, and regulators later capped retail exposure to them (KED Global). When SK Hynix's second-quarter operating profit of ₩60.54 trillion missed the roughly ₩64 trillion consensus on July 29, the shares fell further (TradingKey), and the KOSPI hit circuit breakers in consecutive sessions on July 28 and 29 (KED Global). The Phase 4 rally had help: from August through September, buybacks at Samsung and SK Hynix absorbed foreign selling of ₩31.7 trillion. By October 13, Samsung had finished its buyback, and SK Hynix's was due to end that week (Korea Times). Once that support disappeared, Phase 5 followed.

What Actually Broke

I rank the triggers by how much each one changed the outlook for earnings:

1. Market structure (largest effect on prices, smallest on fundamentals). Since September 23, foreign investors have net sold ₩23.08 trillion of KOSPI shares, and 79% of that selling was in these two stocks (Seoul Economic Daily). With the buybacks over, nobody large was buying. Yuanta's Lee Jae-won put it bluntly: "There are not enough buyers to absorb the selling pressure" (Herald Business).
2. Macro. U.S. September CPI rose 3.7% year on year, and PPI rose 1.3%, more than double the 0.6% consensus. Briefly, futures put the odds of an October Fed hike above 50% (Seoul Economic Daily). The Fed had already raised rates to 3.75%–4.00% on September 16 (Federal Reserve). The 10-year Treasury yield touched 5.342% on October 1 (Reuters via Investing.com); we covered that move in our Treasury yields analysis.
3. AI financing doubts. A Financial Times report put OpenAI's annualized revenue at $50 billion, against a previously reported $70 billion (Korea Times). Gartner warned that $3.5 trillion of AI infrastructure spending faces impairment risk, and CoreWeave fell 15.5% (Seoul Economic Daily). The debt behind the buildout is a real risk, as our CoreWeave analysis argued. Yuanta's Lee notes, however, that no actual capex cut has been confirmed (Herald Business).
4. Competition from China. CXMT says it will ship 4F² server DRAM by year-end, while Samsung targets 2028 for the same cell design. CXMT has disclosed no yields or specifications (Seoul Economic Daily). Korean brokers do not expect meaningful volume before late 2027 (Seoul Economic Daily). The bigger long-term risk is CXMT's DRAM market share, which Counterpoint puts at 10% in the second quarter, up from 4% a year earlier (Seoul Economic Daily).
5. Fundamentals (smallest effect). On October 8, Samsung reported preliminary third-quarter operating profit of ₩107.4 trillion, a 55.1% margin (CNBC; Seoul Economic Daily). Just before the crash, TSMC's results showed no sign that AI demand was slowing (Seoul Economic Daily).

Is the Boom Over? No. Is It Accelerating? Also No.

TrendForce expects conventional DRAM contract prices to rise 10–15% in the fourth quarter. It also notes that "the pace of contract price increases is expected to moderate," while consumer devices struggle to absorb higher memory costs (TrendForce). Prices are still rising, but more slowly. For cyclical stocks, a slowing rate of improvement can weigh on the shares even while profits are still at records.

SK Hynix's second quarter showed the same pattern. DRAM average selling prices rose about 30% from the prior quarter, half of the first quarter's pace of roughly 60%. Long-term agreements with about 10 major customers also limit how much of the spot-price spike SK Hynix captures (TradingKey). As Korea's exporters face a firmer won, consensus for SK Hynix's third-quarter operating profit has slipped from ₩81.51 trillion to ₩76.97 trillion over three months (Korea Times).

The 3x P/E Trap

On today's figures, SK Hynix trades at 3.1 times forward earnings and Samsung at 3.5 times (StockAnalysis, SK Hynix; StockAnalysis, Samsung). Daishin Securities estimated that October 13 prices implied a 35–51% drop from forecast earnings (Korea Times).

Morgan Stanley's counterpoint deserves the same attention. I could not verify the full note directly, so its points here come from secondary summaries of an October 13 report (TokenPost; BigGo Finance). As reported, the bank argues that in past cycles memory share prices peaked four to eight months before earnings did and that new capacity starts arriving in the second half of 2027.

One inference is mine, not the bank's: four to eight months after the June peaks falls between October 2026 and February 2027. If the reported pattern holds, earnings could keep rising while the shares stay below their highs. This is a reason for caution, not a forecast of when earnings will peak.

The two stocks differ here. SK Securities puts Samsung's 12-month forward P/E at 3.3 times, against a post-2010 low of 5.0 times. SK Hynix is at 3.0 times, which equals its own floor (Herald Business). Samsung trades well below its historical minimum. SK Hynix is at its minimum, with no further discount to its history. Our forward P/E explainer explains why these comparisons require consistent earnings periods.

The Peak Buyer's Arithmetic

Losses are asymmetric. A 57% loss needs a 135% gain to recover. Consensus targets of ₩3,141,025 for SK Hynix and ₩478,906 for Samsung (StockAnalysis, SK Hynix; StockAnalysis, Samsung) would cover even the June highs. During the crash, brokers held their targets. NH Investment & Securities kept ₩530,000 on Samsung and SK Securities kept ₩4 million on SK Hynix. Even the lowest targets, ₩270,000 and ₩1.48 million, sit above today's prices (Seoul Economic Daily). Targets that did not move during a 30% fall measure sentiment better than they forecast prices.

Investors should ask whether they would buy at today's price with today's information. The price they originally paid does not help answer that.

Rebound and Downside Scenarios Through 2027

Horizon	Base scenario	Bull scenario	Bear scenario
Weeks (through Oct. 29)	Volatile retest; KB sees a possible W-shaped second bottom	Fed holds Oct. 27–28; earnings beat	KOSPI sinks to 5,100–5,200, the zone analysts say could draw buyers
1–3 months	Partial recovery: SK Hynix ₩1.5–1.7M, Samsung ₩245–270K, KOSPI 6,000–6,500	Back to secondary highs (₩1.84–1.87M; ₩276–285K; ~7,000)	A Fed hike plus a Q3 miss or an AI credit event: SK Hynix ₩0.95–1.1M, Samsung ₩175–190K
6–12 months	Range-bound below secondary highs as markets price 2H 2027 supply	New highs need continued 2027 upgrades and lower rates	Capex cuts or CXMT volume: peak earnings get repriced

Scenario price ranges are the author's illustrative judgments, not outputs of an earnings-and-multiple valuation model. No probabilities are assigned; the ranges show possible outcomes, not expected returns or limits on losses. The 5,100–5,200 zone comes from analysts cited by Seoul Economic Daily. Catalysts: Bank of Korea on Oct. 22 (Seoul Economic Daily); FOMC Oct. 27–28 (Federal Reserve); SK Hynix earnings Oct. 28; Samsung earnings and shareholder-return update Oct. 29 (Korea Times; MoneyToday).

A partial rebound is my base scenario; a return to the June highs is a more demanding outcome. It would require higher earnings, higher valuation multiples, or both.

Verdicts for Buyers Near the Highs

SK Hynix: REDUCE oversized positions; otherwise HOLD through October 28

If you bought between ₩2.5 million and the ₩2,987,000 peak, recovering would take a 97–135% gain. Reduce to a size that keeps your portfolio exposure and potential further loss within your risk budget; otherwise, HOLD through October 28 results. A rebound toward ₩1.6–1.7 million, where the stock closed from October 7 to 12 (Investing.com), offers a possible trimming opportunity. If risk needs cutting now, do not wait for a rebound.

• Downside: ₩0.95–1.1 million, or 14–25%, if the Fed hikes and the third quarter misses or an AI credit event hits.
• Rebound scenario: ₩1.5–1.7 million, an 18–34% gain.
• What would change the call: The case for holding strengthens if third-quarter results meet the ₩77 trillion consensus, fourth-quarter guidance shows DRAM and HBM prices rising faster, and foreign buying lasts for weeks. Downgrade to SELL if a hyperscaler cuts capital spending or CXMT's server DRAM is confirmed to be shipping in volume.

Buyers around ₩1.84–1.87 million need a gain of 45–47%. The same HOLD/REDUCE criteria apply. A return to those prices within three months belongs to the bull scenario and would need support from earnings and demand.

Samsung Electronics: HOLD (both cohorts)

On my scenario ranges, the two stocks have similar upside-to-downside ratios. Samsung has the stronger case on valuation and catalysts. Its forward P/E sits well below its post-2010 floor, while SK Hynix's sits at its own. Its preliminary third-quarter profit beat consensus of about ₩105.7 trillion (Xinhua). Its shareholder-return update on October 29 is a specific catalyst that SK Hynix lacks. Its October 15 close of ₩210,000 was its lowest since April (Seoul Economic Daily), so the stock has already given back the entire secondary rally.

• Downside: ₩175–190K, or 12–19%.
• Rebound scenario: ₩245–270K, a 14–25% gain. That rebound recovers 40–80% of a secondary-peak buyer's loss. A June buyer would need a 74% gain to break even, which is outside my base scenario for the next 12 months. My HOLD judgment rests on today's valuation and the upcoming shareholder-return update.
• What would change the call: Downgrade to REDUCE if the October 29 return plan includes no new buyback, if fourth-quarter guidance points to falling memory prices, or if the KOSPI closes below 5,100, beneath the zone where analysts expect buyers, while foreign selling accelerates.

KOSPI: HOLD unleveraged exposure; cut leveraged ETFs

The index is a diluted version of the same trade. At the June peak, Samsung and SK Hynix made up more than half of its market value (Yahoo Finance). Their share is smaller now, because both have fallen further than the index. On October 13, the KOSPI traded at a record-low forward P/E of 4.54 times (Korea Times). A buyer at the June 18 record close needs a 60% gain to break even, which is outside my base scenario for the next year.

• Downside: 5,100–5,200, or 8–10%.
• Rebound scenario: 6,000–6,500, a 6–15% gain.
• What would change the call: Downgrade to REDUCE if foreign selling resumes at a pace of more than ₩2 trillion a day after the October 29 earnings, or if the index closes below 5,100, beneath both the analysts' buyer zone and its July 29 intraday low of 5,262.77 (KED Global), and does not recover within two weeks.

If you hold leveraged KOSPI or single-stock ETFs, cut them first. In July, forced selling from these products deepened the crash.

The memory cycle is still strong, but a low forward P/E does not remove earnings or portfolio risk. Holding should depend on the outlook from today's price and the size of the position. Peak buyers should not count on a higher multiple to restore their original investment.

---

General investment analysis, not personalized financial advice. Current prices are October 16, 2026 Korea Exchange closes; historical figures are dated in place, with intraday highs identified. Share prices are in won; KOSPI levels are index points. Drawdowns, break-even gains and scenario ranges are the author's calculations and judgments. Analyst data reflect October 13–16 snapshots, and linked pages may change. For background on how rate decisions affect equity valuations, see our CPI, PPI and Fed guide; for earlier coverage of the memory trade, see our Micron valuation phases and Nvidia earnings analyses.