SNOWSHAGALThis note's analysis stops at the October 12 close. The KOSPI fell another 5.46% on October 13 to 6,035.35. Samsung Electronics lost 7.39% to KRW 232,000 and SK hynix 10.09% to KRW 1,443,000.8, 9 The KOSPI touched 5,911.11 intraday, and a sell-side program-trading sidecar was triggered at 9:47 a.m. Foreign investors net sold roughly KRW 4.5tn of KOSPI shares (based on Seoul Shinmun and Edaily closing reports; other outlets reported different totals).9, 10, 11
| Asset | Two sessions through Oct 12 | Oct 13 alone | Three sessions through Oct 13 |
|---|---|---|---|
| KOSPI | −6.2% | −5.46% | −11.3% |
| Samsung Electronics | −6.7% | −7.39% | −13.6% |
| SK hynix | −6.8% | −10.09% | −16.3% |
Three-session returns are calculated from closing prices. Holding the same forward EPS estimate of 1,293.5 index points constant gives a forward P/E of roughly 4.7× at the October 13 close. The main analysis covers only the first two sessions.CALCULATED
iM Securities analyst Yang Hyun-kyung cited concerns about financing and profitability of AI infrastructure investment, China's CXMT catching up technologically, and pressure from foreign selling as factors behind the October 13 decline.12 In the preceding US session, the Philadelphia Semiconductor Index fell by about 2.6% and Micron by roughly 3.9%.8, 13
CXMT · Year-end product launch planned; mass-production viability unverified
At an October 10 event, CXMT's president said the company would introduce server DDR5 using a 4F² cell architecture by year-end. A Morningstar analyst, cited by Bloomberg, expects a substantial gap between unveiling a product and commercializing it at economically viable yields. Korean industry sources say Samsung Electronics and SK hynix also possess 4F² technology but do not use it for volume production because of yield and cost constraints.14, 15, 16
The statement came on October 10. An October 12 closing-market report also cited the possibility that Chinese memory makers could gain share, but did not establish whether it referred to the same development.7, 14
Developments discussed on October 13 should not be retroactively presented as explanations for the preceding two-session selloff.
THE SHORT ANSWER
- Earnings did not fall. Q3 operating profit was up 20.0% even from the previous quarter, and Hana Securities' KRW 116tn Q4 forecast is about 8% higher again. Yet over the two sessions following the release, the KOSPI fell 6.2%, Samsung Electronics 6.7%, and SK hynix 6.8%, all measured from the October 7 close.CALCULATED3, 4, 5, 7, 17
- Against expectations, the surprise was modest. The result slightly exceeded some pre-release estimates of KRW 105–106tn, but missed FnGuide's October 1 consensus of KRW 110.2803tn. Hana Securities notes that the Q3 estimate had already been revised down from a peak of KRW 112tn toward KRW 106tn.2, 3, 5, 6
- The next test is 2027–2028 estimates. Around November 5, the first scheduled checkpoint (calculated), compare 2027 operating-profit consensus against its October 7 baseline—KRW 1,330tn for the KOSPI, as cited by Kiwoom—using the same data provider and fiscal-year definitions wherever possible.5
Both Samsung Electronics and SK hynix lost nearly 7% in two sessions
Samsung Electronics announced preliminary Q3 operating profit of KRW 107.4tn on the morning of October 8. Over the following two trading sessions, its shares fell 6.7% and SK hynix fell 6.8%, measured from the October 7 close.CALCULATED4, 7, 17 The broader decline, however, began before the earnings release. The KOSPI had fallen for four consecutive sessions from October 6 to 12, shedding 8.85%; on October 7, the day before the announcement, it was already down 1.98%.5, 19
KRW 107.4tn is up 782.5% from a year earlier and 20.0% from the prior quarter.4 Hana Securities' Q4 forecast of KRW 116tn is about 8% above Q3.CALCULATED3
The release-day fall was also attributed to profit-taking, options expiry and ETF rebalancing, risk reduction before the holiday, and the withdrawal of share-buyback demand.17, 20
A record profit is not the same thing as a big earnings beat
Whether KRW 107.4tn exceeded expectations depends on the forecast vintage and data provider. The result ranges from a small beat to a miss.
| Benchmark | Forecast operating profit | Beat/miss vs KRW 107.4tn |
|---|---|---|
| Immediately pre-release consensusCited by Daishin Securities (Oct 8) | KRW 105tn | +2.3% |
| Immediately pre-release consensusCited by Kiwoom and Hana Securities | KRW 106tn | +1.3% |
| FnGuide consensusPublished October 1 | KRW 110.2803tn | −2.6% |
| Peak Q3 estimateCited by Hana Securities | KRW 112tn | −4.1% |
Calculated as KRW 107.4tn ÷ benchmark − 1. Hana's quant team placed the pre-release media consensus at KRW 107–114tn and characterized the result as near the bottom of that range. Hana's semiconductor team, by contrast, called it "in line with the lowered consensus." This comparison also appears in NO.09.2, 3, 5, 6, 21, 22
Hana Securities strategist Lee Jae-man classifies this as an earnings beat delivered while the Q3 estimate was being lowered from a KRW 112tn peak toward KRW 106tn. Historically, Samsung shares showed a muted response four weeks after releases in that category.3
Figure 2 · Beats amid falling estimates: historical four-week share return averaged +2.5%
Average four-week outcomes after Samsung's preliminary Q3 releases, 2014–2025 · Hana Securities classification · Historical observations, not a forecast for this release
Hana Securities Figures 3–4, four-week averages. Categories reflect whether Q3 forecasts were rising or falling before the announcement; bars show changes in next-year operating-profit forecasts and stock returns. Sample counts and the return basis (absolute versus market-relative) were not disclosed.3 · Chart: Snowshagal
Over the previous month, annual earnings consensus for Samsung had also declined by 4.83%, while SK hynix estimates were largely unchanged (FnGuide, Oct 8).18
| Metric | Value on reference date | One-month change |
|---|---|---|
| Samsung operating profitFnGuide 10/8 | KRW 370.3756tn | −4.83% |
| Samsung net income attributable to ownersFnGuide 10/8 | KRW 302.9573tn | −5.49% |
| SK hynix operating profitFnGuide 10/8 | KRW 263.3430tn | −0.67% |
| SK hynix net income attributable to ownersFnGuide 10/8 | KRW 256.3226tn | +0.89% |
| KOSPI operating profitHana Securities · −1.3% over one week | KRW 968tn | −1.9% |
FnGuide figures are as reported by Aju Business Daily on Oct 12. Hana data come from an Oct 12 note with no underlying data-as-of date. Most of the KOSPI's one-month downward revision occurred in the latest reported week (exact interval unspecified).18, 21
Hana's semiconductor analysts say memory-makers' earnings forecasts stopped rising after late June and were revised lower from the Q3 preview onward because of the exchange rate (for details, see NO.10); the same team says Samsung's memory prices beat expectations, shipment volumes fell short of assumptions, and the tightest supply was in LPDDR.22
- Depending on the benchmark, the result ranged from a +2.3% beat to a −4.1% miss; Q3 estimates had been falling ahead of the release.
- The historical reaction in this category was modest (Hana; sample count undisclosed). It does not explain this particular two-session drop.
Profits are still rising. Forecast growth rates are falling.
A slowdown in growth does not mean earnings are falling. All three providers expect growth to decelerate from 2026 to 2027. The two with 2028 projections—Eugene and I/B/E/S—expect growth to drop into single digits. Their forecasts nevertheless imply higher absolute earnings.
| Provider | Coverage · Measure | 2026 | 2027 | 2028 |
|---|---|---|---|---|
| Eugene SecuritiesOct 12 report | KOSPI EPSQuantiwise·Refinitiv | 236% | 32.6% | Around 6.3% |
| Daishin SecuritiesI/B/E/S as of Oct 6 | Korea EPSI/B/E/S | 333.5% | 36.0% | 4.8% |
| Hana SecuritiesOct 12 report | KOSPI net incomeFnGuide · KRW 795tn → 1,049tn | 265% | 32% | n/a |
EPS is earnings per share. Eugene's 2026/2027 values use chart labels (236%/32.6%; text rounds to 237%/33%); 2028 uses around 6.3% from the report text. Provider coverage and EPS versus net-income definitions differ, so earnings levels are not compared. The companies and analyst counts underlying the 2028 forecast could not be verified.1, 2, 3
Illustrative EPS level index (2025 = 100), derived from Eugene's growth rates1
Samsung Electronics: quarterly growth rates
Samsung operating-profit growth slowed from 1,814% YoY in Q2 (Eugene Securities) to 782.5% in Q3. If the Q4 forecast of KRW 116tn holds, profit would still be roughly 478% higher than the year-earlier KRW 20.0737tn.CALCULATED1, 3, 4, 23
Semiconductors account for 56.7% of KOSPI market capitalization. Sector net-income growth is forecast to slow from 592.9% in 2026 to 37.5% in 2027.3 The slowdown is not confined to chips. Eugene forecasts KOSPI operating-profit growth excluding semiconductors at 39% in 2026 and 12% in 2027.1
Figure 3 · Korea's growth falls from 36.0% to 4.8% in 2028—the sharpest drop in this table
EPS growth forecasts, I/B/E/S · Smaller figures beneath each name show 2026 growth
Figures from the I/B/E/S table cited by Daishin Securities (Oct 6). Korea also has the lowest 2028 rate among the listed markets. Emerging markets slow from 23.7% to 10.8%, and emerging Asia from 27.3% to 11.2%. Taiwan is absent; Eugene separately projects 20–30% for Taiwan in 2027–2028. The two S&P 500 markers overlap because the projections are identical.1, 2 · Chart: Snowshagal
Korea is not the only market where earnings growth is forecast to slow. Hana's Bloomberg data show S&P 500 EPS growth declining from 35% in 2026 to 16% in 2027 (34.2% to 17.2% in I/B/E/S), even as the S&P 500 continued making new highs during this period, according to iM Securities' Oct 12 report.2, 3, 24 The bigger divergence comes in 2028. Global growth at 14.1%, the S&P 500 at 17.2%, and Japan at 10.4% remain close to their 2027 levels; Korea drops from 36.0% to 4.8%.2
Daishin Securities argued in its Oct 12 weekly report that the market was discounting the durability of earnings from 2027 onward (as reported by Aju Business Daily).18 iM Securities similarly argues that SK hynix will need evidence that profits can endure beyond 2028 before investors reduce the discount applied to potential peak-cycle earnings. It attributes the recent widening of that discount, however, to Korea-specific factors such as foreign selling and reports of a possible Solidigm listing.24
- It is the growth rate that slows; earnings levels are still projected to rise (based on the two datasets with 2028 estimates).
- Korea has the sharpest 2027→2028 slowdown in the I/B/E/S table. The US also decelerates from 2026→2027, while the S&P 500 has been reaching new highs.
What the earnings trajectory cannot explain
To attribute the two-session fall solely to a slowdown that was known before the announcement, we would need separate evidence that expectations deteriorated further after the release.
A stock with steady estimates fell, too
SK hynix's 2026 operating-profit consensus had slipped just 0.67% over the preceding month, according to FnGuide.18 Nevertheless, its shares fell 6.8% over two sessions, much like Samsung's.CALCULATED7, 17
One-month changes in 2026 consensus estimates are not enough to explain the two stocks' simultaneous decline. If prices were newly reflecting 2027–2028 profitability or a higher risk premium, this comparison would not capture it. Nor do unchanged analyst estimates necessarily mean investors' expectations have remained unchanged.
Foreign flows, rates, AI and China
Yuanta Securities analyst Lee Jae-won called foreign investor flows, rather than semiconductor earnings, the domestic market's biggest immediate concern. Foreign investors net sold KRW 2.0186tn of KOSPI stocks on October 12, including more than KRW 1tn of SK hynix.7, 19 Net selling is an observed market outcome. These data cannot tell us whether it was an independent cause or the mechanism through which earnings and interest-rate concerns were reflected in prices.
The US 10-year Treasury yield peaked at 5.31% in the October daily series on Oct 5, then edged down to 5.28% on Oct 7 and 5.24% on Oct 9. It remained elevated throughout the period (US Treasury daily closing data).25 The Philadelphia Semiconductor Index was down 4.3% for the week in iM Securities' Oct 12 report.24 An October 8 Financial Times report put OpenAI's annualized revenue run rate at roughly $50bn as of late September, prompting questions about a previously reported figure near $70bn. Bloomberg's $70bn figure the following day was not a current run rate but the company's projected year-end 2026 annualized revenue, as presented to investors. Hana's quant team saw the dispute as one of revenue-recognition conventions rather than a clear change in underlying AI demand.21, 22, 26 Reports on October 12 also raised the prospect of Chinese memory suppliers gaining market share.7
Current operating conditions, seen through exports
Slower forecast earnings growth need not mean deteriorating operations. September exports reached a monthly record $120.94bn, up 83.5% year on year and above June's previous record. Semiconductors drove the gain at $60.3bn (+262.8%), but exports outside chips also rose 23% (as reported from the trade ministry's release).27, 28 Exports over October 1–10 increased 102.0% in total value, partly reflecting more working days; the average per working day rose 28.5% (Korea Customs Service).29 Export figures are nominal values combining prices and volumes, not measures of corporate earnings growth.
At Micron's late-September earnings call, CEO Sanjay Mehrotra said fiscal 2027 should be stronger than fiscal 2026 and that memory and storage supply-demand conditions in calendar 2027–2028 should be much tighter than in 2026.30 That is management's outlook. Tight supply can persist while growth rates slow against an already enlarged earnings base. The comments therefore do not necessarily contradict the forecast slowdown in Korean EPS growth.
A 5.1× forward P/E
Daishin Securities wrote in its Oct 12 weekly report that the KOSPI's 12-month forward P/E (current price divided by expected EPS over the next 12 months) had fallen to 5.1×, with no underlying price date specified. It judged that prices already reflected considerable concern about the semiconductor cycle peaking and earnings deteriorating.18
Holding the forward EPS figure of 1,293.5 index points from Daishin's Oct 8 report constant, a 6× P/E implies a KOSPI level of 7,761 (the framework discussed in NO.07); applying the same fixed EPS to the Oct 12 close yields 4.9×. Both are mechanical calculations using an earlier EPS estimate; any fair-value multiple requires separate assumptions.CALCULATED2
NOT ESTABLISHED: These data cannot show whether the already anticipated slowdown directly caused the two-session decline, or how much of the move each factor explains. The fact that the index started falling before the announcement makes isolating the earnings reaction harder still.
Some sectors rose even as earnings growth slowed
In two periods selected by Hana Securities, some sectors rallied strongly while index-level earnings growth was decelerating. These are examples of what has happened—not a sector-selection rule.
| Period | KOSPI net-income growth | KOSPI return | Hana's selected sectors · Net-income growth · Return |
|---|---|---|---|
| End-Q3 2010 to end-Q1 2011 | 65.4% → 9.1% | +12.5% | Energy 14.3% → 244.8% · +48.7%Automobiles 30.7% → 101.7% · +36.5% |
| End-Q3 2015 to end-Q1 2016 | 18.7% → 6.8% | +1.7% | Energy: loss to profit → 98.2% · +57.9%Steel 48.0% → 168.1% · +20.3% |
Hana Securities Figures 9–10. Outcomes cover only these two six-month windows; later performance is not shown. Sector growth figures are realized ex-post values and were not available to investors at the time. Two examples only.3
Figure 4 · Forecast 2027 net-income growth by sector
KOSPI sectors (%) · Colors distinguish changes in growth rates versus 2026 · Right-hand values show shares of KOSPI market capitalization
Selection from Hana Securities Figure 11: nine sectors plus KOSPI (the seven highest forecast growth rates, with autos and banks for reference). Sectors that turned profitable in 2026 are shown separately from acceleration because the prior-year growth rate is undefined. IT hardware and steel still show high growth, but slower than in 2026. Of seven sectors where growth accelerates, together representing 14.7% of market capitalization, only insurance and autos are shown in this chart.3 · Chart: Snowshagal
Hana Securities Figure 11 identifies seven sectors with faster numerical growth in 2027: insurance, utilities, transportation, trading companies and capital goods, software, cosmetics/apparel/toys, and automobiles. Their combined market-cap weight is 14.7%. Batteries and media/education, where profit grows after a move out of losses, account for another 5.4%. Semiconductors alone weigh 56.7%.CALCULATED3
Three figures to check by early November
The question now is whether 2027–2028 forecasts will be revised upward from earlier expectations—and whether share prices respond if they are.
2027 operating-profit consensus versus the Oct 7 baseline
The baseline is KRW 1,330tn for KOSPI companies, KRW 554tn for Samsung Electronics, and KRW 401tn for SK hynix (cited by Kiwoom Securities; data aggregator unspecified).5 The first checkpoint is around November 5, four weeks after the release (calculated). As context, Hana's historical 2014–2025 four-week averages show next-year estimates rising 0.4% for a "beat amid downward revisions" and 7.6% for beats overall (sample sizes undisclosed).3
SK hynix Q3 operating profit versus its final pre-release consensus
FnGuide's October 1 figure was KRW 78.0577tn.6 Keep that as an earlier benchmark, then obtain the same provider's immediately pre-release estimate for the actual comparison. Hyundai Motor Securities argues that the key test is whether Q3 results provide evidence of a memory shortage continuing through 2028.31
2028 EPS estimates · I/B/E/S Korea growth baseline: 4.8%
Check whether 2028 estimates rise alongside 2027 estimates. If the 2028 growth rate rises only because the 2027 base has been cut, that does not count as an improvement.2
COMPARISON RULE Compare the same data provider, constituent universe, fiscal year, and earnings measure wherever possible. If estimates rise but stocks do not follow, revisit pricing factors outside earnings forecasts, including flows and interest rates.
Schedule sources: Hana Securities, US Bureau of Labor Statistics, TSMC, Samsung IR, Federal Reserve, and news reports · No SK hynix company date announcement as of Oct 1321, 22, 32, 33, 34, 35, 36
- Samsung preliminary Q3 operating profit: KRW 107.4tn, +20.0% QoQ (Oct 8)
- Over two sessions after release: KOSPI −6.2%, Samsung −6.7%, SK hynix −6.8% (calculated)
- Foreign KOSPI cash-equity net selling around KRW 2.0tn on Oct 12 (reported)
- Record September exports; average daily exports +28.5% for Oct 1–10
- Forecast earnings-growth paths (Eugene: 236%→32.6%→~6.3%; I/B/E/S: 333.5%→36.0%→4.8%)
- Consensus and Q3 estimates lowered from KRW 112tn to KRW 106tn; classified as a "beat amid downward revisions" (Hana)
- Samsung Q4 forecast KRW 116tn · 2027 sector growth forecasts
- Micron management's 2027–2028 supply-demand outlook
- What directly drove the two-session decline, and each factor's contribution
- Whether investors actually cut 2027–2028 expectations after the release
- Memory supply-demand balance and profit durability in 2028
- When foreign investor flows might turn
This note examines stock-price moves through the lens of earnings projections; it is not a trading recommendation. Source dates, chart-reading choices, and calculations are documented below.
Glossary
- EPS · Earnings per share
- Net income divided by the number of shares. Growth rates here are year-on-year.
- Consensus · Expectations · Estimates
- Consensus is the average of multiple analysts' forecasts. In this note, "expectations" generally denotes the immediately pre-release quarterly consensus; "estimates" refers more broadly to quarterly or annual earnings projections.
- I/B/E/S · FnGuide · Quantiwise
- Data aggregators that compile analysts' forecasts. Their coverage and definitions differ, so the same market can have different estimates.
- 12-month forward P/E
- Current share price divided by expected earnings per share over the next 12 months.
- Preliminary vs final earnings
- A company's initial summary earnings announcement versus its later, more detailed release including segment results.
- Return to profitability
- A move from a loss to a profit. With a loss in the comparison year, a conventional YoY growth rate is undefined, so the following year's rate cannot be compared as an acceleration.
- LPDDR
- Low-power DRAM used in mobile devices and servers, among other applications.
Sources · Reference dates 36 entries
Main analysis uses information available at the Korean close on Oct 12, 2026 (before the Oct 12 US close); only the update at the top uses Korean closing reports from Oct 13.
- AS-OF DATESKOSPI and flows: Oct 12 closing reports · Samsung and SK hynix FnGuide consensus: Oct 8 · I/B/E/S: Oct 6 (Figures 1 and 3, and table) · 2027 operating-profit consensus: Oct 7 · US 10-year: Treasury daily series through Oct 9 · Eugene, Hana and iM reports dated Oct 12 (underlying data dates unspecified)
- CALCULATEDTwo-session performance following the release (from Oct 7 closes): KOSPI 6,803.90 → 6,384.24; Samsung KRW 268,500 → 250,500; SK hynix KRW 1,723,000 → 1,605,000. The Oct 7 stock closes were reconstructed by adding the Oct 8 declines of KRW 6,500 and KRW 42,000 to Oct 8 closing prices. Earnings surprise: KRW 107.4tn ÷ benchmark − 1. Q4/Q3 forecast growth: 116 ÷ 107.4 − 1. Samsung Q4 YoY growth: KRW 116tn ÷ KRW 20.0737tn − 1. EPS level index: 100 × 3.36 × 1.326 × 1.063. Sector market-cap weights: summed from Hana Figure 11. Forward P/E 4.9×: 6,384.24 ÷ 1,293.5. Index at 6×: 1,293.5 × 6 = 7,761. First checkpoint: Oct 8 + 28 days. Three-session return through Oct 13, relative to Oct 7 closes: KOSPI 6,803.90 → 6,035.35; Samsung KRW 268,500 → 232,000; SK hynix KRW 1,723,000 → 1,443,000.
- CHART INTERPRETATIONFigure 1's 2028 Eugene bar uses the approximately 6.3% figure from the report text; all other chart figures are taken from published chart labels or tables.
- SOURCE DISCREPANCIESEugene says Q4 operating profit of KRW 120–130tn would imply YoY growth in the 480% range; against finalized Q4 2025 profit of KRW 20.0737tn, the calculation instead gives +498–548%. Eugene labels its EPS-chart source Quantiwise on p.1 but Refinitiv on p.3. Hana Figure 4's prose, "under 1%" and "around 2%," approximates the plotted +0.4% and +2.5%. Hana's quant note incorrectly dates the preliminary earnings release to before trading on Oct 7; the company announced on Oct 8. The S&P 500's 2027 EPS growth is 16% in Hana/Bloomberg data versus 17.2% in I/B/E/S. Hana Figure 1 has SK hynix Q3 operating profit at KRW 77tn versus FnGuide's Oct 1 consensus of KRW 78.0577tn. Hana gives SK hynix 2027 operating profit at KRW 402tn versus Kiwoom's KRW 401tn. iM reports a weekly Philadelphia Semiconductor Index fall of 4.3%, Hana 3.8%. US 10-year yields are Treasury end-of-day readings and may differ from intraday highs. Daishin's 5.1× P/E does not specify the underlying price date. Aju Business Daily reproduced Daishin's assessment as discounting profit durability "from 2027 onward rather than this year and next"—internally overlapping since next year is 2027—so the main text retains only "from 2027 onward." Hana uses KRW 116tn in the Q4 forecast narrative but KRW 115–116tn in the Figure 1 note (KRW 115tn would be +7.1% QoQ). Reports of Oct 13 foreign net selling range from KRW 4.5128tn in Seoul Shinmun/Edaily (3:30 p.m.) and KRW 4.4926tn in K News to KRW 5.1532tn in Jungang Economy News; differences in collection timing or methodology could not be confirmed. The text uses roughly KRW 4.5tn. Seoul Shinmun's 3:35 p.m. Oct 13 report gives Samsung at KRW 232,500 (−7.19%) and SK hynix at KRW 1,442,000 (−10.16%); we use the Newsis/K News figures, which match the reported day-on-day declines of KRW 18,500 and KRW 162,000. For the Oct 12 US session, Newsis and K News report a 2.67% Philadelphia Semiconductor Index loss, while Jungang Economy News reports 2.60%.
- HOW TO LOCATE REPORTSBroker research reports are identified by title, date, page and figure number.