Goldman Sachs: AI Storage Demand Underestimated, KOSPI Target of 12,000 Implies 80% Upside
nashnova research
Goldman Sachs Asia-Pacific chief equity strategist Timothy Moe holds his KOSPI target at 12,000, implying roughly 80% upside from current levels, arguing the market systematically underestimates how long the AI-driven memory-chip upcycle will last.
An 80% upside call — what's the case?
Timothy Moe set the 12,000 target about three months ago — already one of the most bullish calls on the street. KOSPI has since fallen 27% from its June high. He is not backing down.
His core logic rests on one thing: actual earnings delivery. This means → he is not betting on a sentiment bounce but on companies genuinely earning the money.
Samsung Electronics and SK Hynix both posted strong results, yet their shares barely moved. Moe argues the market has already digested the idea that earnings growth will eventually slow — that alone does not explain the current weakness.
Why will memory chips stay in short supply?
The global data-center buildout race has created a massive shortage of memory chips — chips that store and retrieve data — pushing prices steadily higher.
Moe expects U.S. big-tech capex to exceed $1.2 trillion next year, far above the prior consensus of roughly $800 billion. In plain terms = tech giants will spend nearly half again as much on data centers as people had assumed.
Hyperscale cloud operators "must keep investing even if they cannot yet turn a profit." This reflects the fact that AI compute demand is inherently storage-intensive — more compute power means more memory chips, directly benefiting memory manufacturers.
The valuation math behind 12,000 — is it really that aggressive?
The 12,000 target is based on a 7.5× forward P/E — price-to-earnings ratio, which measures how much the market is willing to pay for each dollar of profit.
KOSPI currently trades at roughly 5.3×, about half its seven-year average. This means → the market is applying a steep discount to Korean corporate earnings.
Moe's own words: "If Korean companies deliver on their earnings forecasts, 12,000 is not as aggressive as it looks." Put simply = the question is not whether the target is high — it is whether the profits actually materialize.
After 360% earnings growth — what comes next?
Moe projects KOSPI constituent earnings growth of roughly 360% this year, slowing to about 35% in 2027.
He acknowledges growth will decelerate but stresses this is not news — the market already knows it. This means → what really moves the stock price is not "growth will slow" but how far it slows and whether absolute profit levels can still justify the valuation.
The supply-demand squeeze is expected to tighten further in 2027, underpinning absolute earnings levels.
Where are the biggest risks?
The rise of China's CXMT (ChangXin Memory Technologies) — if its technology catches up faster than expected, it could erode Samsung's and SK Hynix's market share.
Political headwinds for U.S. data-center construction — site selection, energy permitting, and local opposition could all slow the buildout pace.
Moe's view: these risks are not enough to undermine the fundamental advantage of advanced memory-chip makers over the next several years. The gap between current valuations and historical averages will be the key variable in testing whether that call proves right.
市场有风险，内容仅供研究参考，不构成投资建议。