Consensus has already chosen the easy story: South Korean retail traders piled into Samsung and SK hynix with leverage, the stocks cracked, and the whole rally must have been a froth bomb waiting to pop. That story feels complete because it has a villain. It just skips the part where the underlying businesses keep doing the only thing that matters: shipping memory into a structurally tight AI cycle.
CNBC’s report on the unwind is the clue, not the conclusion. The piece described leveraged retail bets in Samsung and SK hynix unraveling after the selloff, which tells you the first casualty was position size, not product demand. If a forced unwind can hit a stock hard enough to make the headline, that proves someone got crowded; it does not prove wafer demand disappeared or HBM orders collapsed.
Now add the numbers the market is lazily waving past. Samsung Electronics reported KRW 300.9 trillion in revenue and KRW 32.7 trillion in operating profit in 2024, while SK hynix reported KRW 66.2 trillion in revenue and KRW 23.5 trillion in operating profit, according to their annual results. That is not a pair of fragile story stocks. That is a memory duopoly with industrial-scale earnings and a market cap that can absorb a retail margin panic without the business itself blinking.
Here is the deadpan fact bomb: retail leverage can erase months of gains in hours, but it does not automatically change one shipment, one server build, or one purchase order. The market loves to confuse pain with signal. Reality is more boring and more useful: if the companies keep guidance intact, the panic is a price event, not a business event.
The proof chain gets stronger when you look at the operating side. SK hynix has been the cleanest beneficiary of the HBM cycle, and that matters because HBM is where pricing power has lived. Reuters reported in January that SK hynix expected HBM sales to account for a mid-20% share of total DRAM bit output in 2025, up from the low double digits the prior year. That is not the language of a business rolling over; that is the language of a vendor moving volume into the highest-value part of the stack.
Samsung is not immune to cyclical ugliness, but it is also not standing still. Reuters reported in January that Samsung planned to expand investment in advanced memory and AI-related capex after a tough 2024, and the company’s own annual results showed the scale of the turnaround already in the machine. If the market wants to argue that a retail unwind equals a fundamental break, it needs to show where Samsung’s memory and HBM roadmap changed. So far, it hasn’t.
The market’s mistake is simple: it is using the stock chart as if it were a customer survey. It isn’t. Forced liquidation can magnify a dip into a rout, especially when leverage is involved, but the only thing that turns that into a real thesis break is a visible operational crack. Until you see a guidance cut, a margin reset, or a shipment slowdown, the selloff is telling you who got forced out, not what the chips are worth.
Screenshot this: Samsung 2024 revenue: KRW 300.9 trillion; SK hynix 2024 revenue: KRW 66.2 trillion; SK hynix 2024 operating profit: KRW 23.5 trillion. That is the entire argument in one frame: huge businesses, not tiny trading toys.
The cleanest way to kill the bull case is also the simplest. If Samsung or SK hynix cuts 2025 revenue or operating profit guidance at the next earnings update, the story changes immediately. If either company reports sequential gross-margin compression tied to weaker memory pricing, or discloses lower HBM shipment volumes, then the market was right to look past positioning and at the business itself.
Until that happens, the selloff looks like the same old market habit: punish the levered buyer, then pretend the leverage itself was the thesis. It wasn’t. It was the wrapper. The actual asset is still a memory cycle with real demand, real earnings, and real scale underneath it.
Verdict: Buy the selloff. This is a leverage flush, not a chip-cycle top, and I want proof from the next earnings update before I flip.