Post

Cumulative revenue in the first half of the year exceeded 100 trillion KRW for the first time, a year-on-year increase of 257%. Second-quarter revenue was 79.32 trillion KRW, operating profit 60.54 trillion KRW, net profit surged 1242% year-on-year, and operating margin reached as high as 76%. Cash and cash equivalents stood at 88 trillion KRW (approximately $61.6 billion), a nearly 62% quarter-on-quarter surge, with a net cash position of 69.4 trillion KRW. HBM4 has entered mass production and shipment, and HBM4E samples have been delivered to customers. In Q1, HBM market share was 58%, firmly holding the global number one position. Each of these figures alone would be enough for any company to pop champagne in celebration. So what happened? On the day of the July 29 earnings release, SK Hynix’s Korean stock plummeted over 17%, marking the largest single-day drop in history. Since listing on the US stock market, the stock price has fallen about 21% from a high near $195. From the June peak, the retracement exceeds 50%, with market capitalization nearly halved. The most profitable earnings report became the worst catalyst for the stock price. What exactly is the market afraid of? It fears the infamous "death spiral" of the memory industry— Profit → Capacity expansion → Overcapacity → Price war → Losses → Bankruptcy. Over the past 30 years, this cycle has destroyed countless memory companies. Every industry boom marks the start of a capacity expansion frenzy; every expansion frenzy foreshadows the next crash. What is SK Hynix doing now? Expanding capacity at the fastest pace in history. In the first half of 2026, capital expenditure on tangible assets reached 18.33 trillion KRW, a 72.7% year-on-year increase. Full-year capital expenditure is expected to reach a high range of 40 to 50 trillion KRW. SK Group has announced a $720 billion investment over the next decade for AI memory capacity expansion, aiming to triple current capacity. In July, it raised another $26.5 billion through a Nasdaq listing, all funneled into expansion. What the market sees is a company throwing every penny earned, and even borrowed money, into new factories. And the memory industry’s history tells the market: when everyone thinks demand will never be enough, that’s usually when supply is about to become excessive. But this time, it might really be different. First, long-term contracts lock in demand. SK Hynix has signed long-term supply agreements with about 10 core customers, lasting 3 to 5 years. Among them, a $500 billion cooperation deal with NVIDIA has been finalized. The long-term contract coverage for HBM even reaches 100%. Previously, the industry signed one-year contracts; now they sign for five years. What does five years mean? It means that no matter how market prices fluctuate, this capacity already has buyers. Second, supply really can’t keep up. Goldman Sachs estimates a global DRAM supply-demand deficit of about 5% in 2026, expanding to 5.9% in 2027, with tightness continuing through 2028. JPMorgan predicts HBM supply-demand deficits of approximately -15%, -14%, and -22% for 2026 to 2028 respectively. SK Group Chairman Choi Tae-won put it more bluntly: "All customers are demanding nearly double the original supply volume, but supply is completely insufficient." He even predicts 2027 will be the worst year for the "memory chip shortage." Third, EUV equipment delivery times exceed two and a half years; capacity can’t just be expanded on a whim. It takes at least four to five years from semiconductor factory construction to mass production. Even if money is spent now, new capacity won’t be fully released until late 2027 to 2028. Demand surges instantly; supply response is measured in years. The market isn’t pricing "how good Hynix is," but rather "whether this industry will repeat its past mistakes." A 557% profit growth isn’t enough because the market fears that every dollar you earn now is ammunition for the next crash. But if you look further— Long-term contracts lock in five years of demand, supply-demand deficits persist through 2028, EUV equipment expansion takes four to five years, and AI demand is structural rather than cyclical— This time, the "death spiral" might really be broken. Choi Tae-won himself said: "Rather than saying the semiconductor industry cycle has disappeared, it’s more accurate to say the cycle has been greatly extended." What does this mean for the crypto market? First, don’t judge an asset’s value by "profit growth." The market prices "whether future cash flows can cover today’s capital expenditure." Hynix earned 557%, but invested all profits into new factories—the market says "not enough." Second, long-term logic and short-term pricing are different. Hynix’s long-term logic is impeccable—AI demand, locked-in contracts, supply-demand deficits. But the short-term market sees capital expenditure growth (72.7%) far outpacing profit growth (the 557% base effect can’t hide marginal slowdown). Third, "AI narrative" projects in the crypto market are undergoing the same valuation reset. When traditional capital markets are repricing AI hardware stocks, projects riding the AI hype will fall even faster. Final sentence: A 557% profit growth, in the face of "death spiral" fears, is only worth a limit-down. But when long-term contracts lock in demand and supply-demand deficits persist through 2028— today’s capacity expansion might be tomorrow’s deepest moat. $BTC $SKHYNIX $SKHY #海力士扩产提速，资本开支能否兑现回报