Post

Samsung fell, then Hynix fell, but this time it's really different This morning, the South Korean stock market crashed again. The South Korean KOSPI index dropped more than 5% at one point, triggering a circuit breaker. SK Hynix fell more than 10%, Samsung Electronics fell more than 6%. In one month, SK Hynix dropped 35%, Samsung dropped 23%. Do you have friends like this around you—who chased Korean chip stocks at the beginning of the year, shouting "Memory is king in the AI era," and now looking at their accounts, feeling terrible? Don't rush to sell. Goldman Sachs released a report today with one sentence: "The South Korean stock market has been oversold." They set the KOSPI target price for the next 12 months at 12,000 points. The logic is simple—memory cycles may be stronger and last longer than before, and the current stock prices do not reflect this at all. Goldman Sachs also reiterated "buy" ratings for Samsung Electronics and SK Hynix, with target prices of 490,000 KRW and 3,500,000 KRW respectively. You might think Goldman Sachs is bullish and unloading. But look at two sets of data: First, inventory. As of the end of Q2, Samsung and Hynix's DRAM and NAND inventory is only 2 to 4 weeks. The normal level is 4 to 5 weeks. Historically, before every downturn cycle, inventory was above 10 weeks. Current inventory is even below normal levels. Second, capacity. The three major manufacturers' DRAM and HBM capacity for 2027 is completely sold out. Customers ultimately receive only 60% to 70% of their initial orders. In 2027, the industry is already calling it "the year of the greatest storage shortage." Even more striking is one thing— Apple tried to pressure ChangXin Memory on price but was rejected. Apple, the world's most powerful buyer, who could refuse Apple's price cuts in the past? ChangXin Memory directly said: prices cannot be lower than Samsung and Hynix. No backup options left. The bargaining power for DRAM has completely shifted from end manufacturers to storage companies. This is not a cycle. This is a structural shortage. So the question arises: with memory supply continuously tight, who truly benefits? Chip companies? Samsung and Hynix are raising prices—July DRAM contract prices rose 14.3% month-over-month, hitting a record high. But stock prices are falling because the market worries "AI capital expenditure has peaked," treating chip stocks as derivatives of the AI cycle and selling them off. Equipment companies? Tokyo Electron, ASML, those selling the "shovels," have orders booked through 2028. But equipment companies' valuations already fully reflect expectations; further gains require surprises. Or the AI industry chain? HBM is fully booked by Nvidia, and the three major manufacturers' 2027 HBM capacity is sold out. But AI chip companies' own stock prices are also adjusting. My view is simple— In this memory cycle, the real beneficiaries are not any single segment but the entire "storage" asset class itself. DRAM is shifting from a "component" to a "strategic material." Memory used to be a commodity with a three-year price cycle. Now memory is a bottleneck; whoever has the supply calls the shots. Apple can't push prices down, capacity is sold out two years in advance, inventory is only two weeks—this is not how a cycle peak looks. To be frank: The market is treating Korean chip stocks as AI concept stocks and selling them off, but the fundamentals of memory and the AI narrative are two different things. AI can cool down, but servers need memory, phones need memory, cars need memory. DRAM demand is rigid, supply is monopolized. Samsung and Hynix have dropped 23% and 35% in one month. Goldman Sachs says this is overselling. Do you think it's a wrong kill or the cycle peak? $SKHYNIX $SKHY $SAMSUNG #内存卖方市场延续，韩股能否迎来反转？