Earnings surge by 556%, but the stock price falls anyway into a bear market—this logic for memory stocks has been彻底 changed.

Today in the Korean stock market, blood is spilled.

SK Hynix falls more than 10% straight at the open, breaking below the 2 million won level. Samsung Electronics drops more than 5%. The KOSPI triggers the circuit-breaker mechanism.

Guess what happened? A profit bomb? Did the industry collapse?

On the contrary.

SK Hynix’s Q2 operating profit is expected to be 60.4 trillion KRW, up 61% quarter-on-quarter and up 556% year-on-year.

A 556% surge. One of the most explosive earnings in history.

But the market doesn’t buy it. Not only doesn’t buy it—on the contrary, it crashes hard.

From the June 25 historical high, SK Hynix’s pullback over three weeks is 33%, officially slipping into a technical bear market.

This isn’t an isolated case. A week ago, Samsung Electronics released a Q2 earnings outlook—operating profit up 18 times year-on-year, to 89.4 trillion KRW. What happened then? The stock price fell nearly 7% that day.

The better the earnings, the harder it falls.

The cycle logic of memory stocks is undergoing a brutal reversal.

Why did earnings explode but the stock price crumple instead?

The core is just four words: expectations priced in.

Over the past half year, by how much has SK Hynix’s stock price risen? By more than 300%. Samsung Electronics is up nearly 116% within the year. The market has already chewed up, swallowed, and even fully digested the story of “AI storage demand exploding.”

At this point, when you publish earnings—even with a 556% year-on-year increase—if you don’t beat the number the market has already priced in, it’s a negative.

Where is SK Hynix’s problem this time? KIS, a Korean local brokerage, forecasts 60.4 trillion KRW, while the market consensus is 65 trillion KRW. An 8% shortfall leads to a 10% stock drop.

Capital markets are just that cruel: what matters isn’t how well you perform, but whether you’ve done “so well” that you meet or exceed the level “already expected.”

Even more painful: the reason for this “below expectation” outcome is precisely the biggest selling point Hynix had before—HBM has too high a share.

The logic is like this: HBM uses long-term supply contracts to lock in prices. Contract prices are relatively fixed and won’t rise sharply along with market conditions. Meanwhile, the spot prices for ordinary DRAM and NAND have higher price elasticity—so during this price hike wave, their gains are even larger.

KIS predicts Q2 DRAM average price up about 30% quarter-on-quarter, and NAND up about 50%. But because Hynix’s HBM share is too high, its overall ASP gain is “held back” by the contract prices.

Your bragging moat, at some point in the cycle, becomes a drag instead.

Bulls and bears are arguing like crazy now: is this a short-term pullback, or a cycle top?

The bears say: Memory is a cyclical stock. The thicker the profits, the more aggressive the capex, and the supply-demand gap will eventually be filled. Samsung’s R(F) production ramps up early, Micron rushes to catch up—supply is being released. Plus Meta starts renting compute power; the biggest buyer is signaling it has “bought enough.” The cycle top is here—time to take profits.

The bulls say: You don’t get it. This is an AI-driven supercycle, completely different from past memory cycles.

The Bank of Korea even stepped in to speak: “There are no signs yet that the semiconductor industry cycle has slowed.” AI infrastructure investment has boosted demand significantly, but the pace of supply expansion remains slow.

Bank of America is even more aggressive: it directly raises its storage chip price and revenue forecasts for 2026–2027, judging that the 2026 market size will be close to $604k and that 2027 will break through $1.2 trillion. BofA says the market’s selling of Samsung is “mispriced.”

The CEO of SK Hynix, Kwak Ro-jung, said the key lines as follows: “2027 may become the most supply-tight year in the history of the storage industry. Even with continued capacity expansion, demand after 2030 may still be in shortage.”

Bulls also have another card: HBM4. Industry sources say that the price of the next-generation HBM4 could jump from about $2 per gigabit in the second half of 2026 to $4–$5 or even higher—basically doubling. The HBM4 production cycle lasts 4 to 6 months, and the wafer capacity it consumes is three times that of standard DDR5. Capacity simply can’t keep up.

On one side, panic and stampede from “the cycle is already topped.” On the other, firm bullishness that “the supercycle is only just starting.”

The contradiction you see is exactly the essence of the cycle

Memory stocks have never been growth stocks. They are cyclical growth stocks—having growth attributes, but fundamentally, they’re cycles.

When they rise, the market buys them as AI growth stocks and assigns extremely high valuations. Once even a little wind blows—profit-taking flows out in droves—then it reverts to that cyclical stock where “profits move with prices.”

The deep reason behind this crash can be boiled down to one sentence:

In the past half year, it’s risen too much. Any news that isn’t “perfect enough” will be magnified into a negative.

Last Friday, SK Hynix successfully listed on Nasdaq, rising 12.8% on the first day. Good news got cashed in—some funds naturally chose to take profits. Also, Korean local brokers cut expectations, there were too many profit-taking positions, and trading was extremely crowded—three forces stacking together is today’s stampede.

But is this a signal that the cycle is ending?

I don’t think so.

The Bank of Korea’s assessment, BofA’s upgrades, the CEO’s warning, and the HBM4 pricing expectations—none of these are short-term noise; they’re bottom-layer signals from the fundamentals.

The business model of the memory industry is undergoing deep changes: in the past it was like a commodity, with contracts for quarterly and annual terms; now, cloud providers and AI data centers sign 3-to-5-year long-term agreements to secure key supply. What does that mean? It means capacity for the next few years is already locked in, and supply flexibility is greatly reduced.

This isn’t a cycle that can be easily “fixed” by ramping up production.

“The top of a cycle stock is never at the worst earnings—it's when earnings are the best and everyone thinks, ‘This time is different.’”

But this time, maybe it really is different. #PreIPOs第二期OpenAI认购 #LAB两日腰斩53% #伊朗宣布关闭霍尔木兹海峡 $BTC $SKHY $MU