SK hynix is converting an AI-memory shortage into a contracted franchise, and the Nasdaq listing is how global capital now prices that shift. The second quarter after the American depositary debut shows a seller allocating scarce wafers to customers who pay for bandwidth rather than cheap bits. Operating margin reached seventy-six percent, which is the signature of that allocation, not of a cost-out program. The debate is not whether the cycle is strong. The debate is whether the new contract book and the cash-return program survive the first serious test of pricing power.
The cash engine is real, but the income statement is two stories stacked together. Memory operations produced operating profit of KRW 61 trillion as DRAM and NAND contract prices jumped and high-bandwidth memory began shipping the newest stack at scale. Net profit reached KRW 94 trillion only because a long-held Kioxia stake produced a sale and revaluation gain that does not repeat. Shareholders who treat the bottom line as the run-rate of the franchise are capitalizing a gift. The operating line is the argument. The investment gain is a balance-sheet event that later funded the buyback conversation.
Newest-stack mass shipments started, long-term agreements now cover about ten customers, and the board later authorized a KRW 40 trillion repurchase and cancellation after calling the Seoul price cheap relative to intrinsic value. Cash finished the quarter far above remaining debt, which is what made that authorization possible. The next several quarters resolve a single question: does contracted AI demand keep conventional memory tight enough for these margins to hold after the one-time gain rolls off?