On July 28, SK Hynix delivered an epic-level set of earnings:

Revenue was 79.32 trillion won, up 257% year over year;
Operating profit was 60.54 trillion won, up 557% year over year, with an operating profit margin of 76%;
Net profit was 93.92 trillion won, with a net profit margin of 118%.
In the first half of the year, cumulative revenue exceeded 100 trillion won for the first time, and quarterly profit hit an all-time high.
So what happened? The stock price still crashed that day.

In the Asian trading session, trading opened and Hynix’s stock kept falling; over the past five trading days, it has already dropped by 20%.
The explanation given by the authorities—or, in other words, the first market reaction—is that it was "below expectations".
LSEG SmartEstimate (a weighted measure that gives more weight to analysts with higher prediction accuracy) had previously forecast operating profit of 64 trillion won. The actual result came in at 60.5 trillion won—about 5.5% lower.

But the actual reason is quite simple. The storage cycle is still in the downturn stage; we’re nowhere near the bottom yet—this is still the “buying the dip” phase for the crowd.
The essence of the memory cycle is this: when a new narrative emerges (right now it’s AI; before it was big data, mobile internet, etc.), prices of memory chips rise rapidly and manufacturers’ profits explode.
Then the industry starts wildly expanding capacity. Once the new capacity concentrates into the market, prices will fall back again, and manufacturers’ profits will plunge dramatically.

Micron, Hynix, and Samsung—the three big memory manufacturers—started losing money and controlling production from the end of 2022, at the bottom of the last cycle.
From 2024 to early 2025, the industry saw almost no new capacity investment, and the coordinated production cuts directly led to a tight supply of DDR4.
After 2025, all three will shift capacity to HBM with higher profit margins. In effect, the supply of traditional DRAM/NAND passively contracts, so the storage industry’s overall gross margin stays high.
And right now, it’s clearly the phase of capacity expansion:
Micron raised its capital expenditures for fiscal 2026 from $20 billion to $25 billion, guiding that it will exceed $35 billion in 2027—almost doubling in two years.
Samsung plans to spend over 110 trillion won (about $7.33 billion) on R&D and infrastructure in 2026, a 21.7% year-over-year increase. The previously shelved P5 plant in Pyeongtaek has also been restarted.
A bigger move is that Samsung and SK Group have joined forces to announce a decade-long investment plan with a total scale of 20 quadrillion won—about $1.3 trillion—spanning semiconductors, AI infrastructure, and data centers.
What does that mean? The planned expansion investment from the three major storage original equipment manufacturers combined is expected to exceed $1.5 trillion. Historically, this scale of capital expenditures hasn’t happened before.

The real reason is something everyone in the storage industry knows: China’s storage capacity is about to enter the scene and flip the table~
Xin’s global DRAM market share surged from 3% to 8% within a year. It has already secured more than $7 billion worth of server DRAM orders from ByteDance over five years, plus a $3 billion deal with Tencent. The share of server business revenue jumped from 8.4% in 2024 to 26.5%.
The IPO prospectus planned to raise RMB 30 billion, with nearly 70% going to upgrades of conventional DRAM production lines and R&D for the next generation. The capacity target is to ramp to 300,000 wafers per month by the end of 2026.
On the higher-margin battlefield of HBM, ChangXin is also preparing—HBM3E mass-production targets are set for 2027.
When memory makers start shouting that they’re losing money and begin cutting production, then everyone can go bargain-hunting again~
