Goldman Sachs Maintains KOSPI 12,000 Target, Citing Valuations at Historically Extreme Low Percentiles
nashnova research
South Korea's KOSPI fell roughly 20% over seven straight sessions, triggering a circuit breaker. Goldman Sachs kept its 12-month target at 12,000 — implying a doubling from the current ~5,700 level — but the call rests on a memory-chip earnings recovery, not a near-term bounce.
How rare is this sell-off?
KOSPI dropped for seven consecutive sessions, with a peak drawdown of about 20% and an intraday circuit-breaker halt.
This means → in the past two decades, only the March 2020 Covid crash (−20.5%) and the August 2011 U.S. credit-rating downgrade (−15.1%) produced comparable drawdowns. This ranks in the top three.
In plain terms = South Korea's market is living through the kind of shock that comes two or three times in a generation.
Why did it crash — did fundamentals collapse?
Goldman's diagnosis: the root cause is a structural absence of buying support, not an earnings breakdown.
Samsung Electronics completed its share-buyback program on October 6; SK Hynix wound down its late-session purchases around the same time. Both stocks' "floor bids" exited at once, leaving the market without key price support.
Foreign investors sold for 8 straight sessions, dumping $11.6 billion in tech alone — roughly 80%–87% concentrated in Samsung and SK Hynix.
In plain terms = the companies aren't broken — the buyers all disappeared at once. Buybacks stopped; foreigners kept selling.
Why does Goldman still see 12,000?
Rock-bottom valuation: KOSPI's forward P/E has fallen to 5.1×, sitting 2.7 standard deviations below the historical mean. This means → statistically, the index is cheaper than it has been in the vast majority of its history.
Outsized earnings growth ahead: Goldman forecasts Korean market earnings growth of 320%, 35%, and 20% for 2026–2028, driven by AI hardware demand — Samsung and SK Hynix are among the world's top memory-chip suppliers and direct beneficiaries of rising data-center capex.
This reflects Goldman's core logic: the current price treats the memory-chip earnings recovery as if it won't happen, yet the AI demand cycle is far from over.
What role does the won play in this story?
Samsung and SK Hynix must convert 70%–80% of their trade surplus into won to cover taxes, capex, dividends, and employee pay.
This means → as long as both companies earn dollars from exports, they must keep selling dollars and buying won — a persistent downward pull on USD/KRW.
Bank of Korea Governor Shin said further won appreciation would boost household purchasing power and reiterated a restrictive monetary stance until inflation converges to target.
Which three risks could derail the 12,000 call?
Memory-cycle peaking: if HBM — high-bandwidth memory, a critical component for AI training — or DRAM demand shows signs of topping out, the earnings-growth path faces a direct downward revision.
Retail-leverage stampede: if leveraged-ETF deleveraging shifts from orderly to forced liquidation, it would pile selling pressure on top of the foreign outflow.
Chinese open-source AI competition: if Chinese open-source models keep dampening global hyperscalers' willingness to spend on compute, memory-chip shipment expectations would weaken from the demand side.
What are the next dates to watch?
October 27: SK Hynix reports Q3 earnings — the market's first hard check on actual memory demand, and the first real test of the 12,000 thesis.
November 18–20: Samsung Electronics pays out roughly ₩40 trillion in dividends. The resulting FX conversion demand should provide a short-term floor under the won and offer a window to see whether foreign net selling is fading.
市场有风险，内容仅供研究参考，不构成投资建议。