On October 13, the South Korean stock market experienced a "Black Tuesday."
The KOSPI plunged 5.46%, Samsung Electronics fell 7.39%, SK Hynix dropped over 10%, and foreign investors net sold 4.53 trillion KRW in a single day, directly triggering the sell-side circuit breaker mechanism.
The trigger was not war, nor the Federal Reserve, but a piece of news from China.
Changxin Memory's president Cao Kanyu announced: the 4F² architecture has made key progress, and by the end of the year, a new generation DDR5 RDIMM product integrating this architecture with hybrid bonding technology will be launched.
What is 4F²? Why are Koreans so scared?
The current mainstream architecture in the DRAM industry is 6F². F represents the minimum linewidth achievable by lithography. 6F² means each memory cell occupies 6 times the square of the minimum feature size. 4F² cuts this down to 4 times — reducing the cell area by 33%, allowing 50% more memory cells in the same area.
But this is not a simple "shrink." Moving from 6F² to 4F² is essentially a fundamental transformation where the transistor channel changes from a horizontal layout to a vertical standing one. The design and process flow of the 6F² era cannot be reused; the entire new process must be integrated from scratch.
Changxin's core weapon is called VSG.
Vertical Surrounded Gate — a vertical surround-gate transistor. The gate surrounds the vertical channel on all four sides, maximizing gate control capability, directly solving the two DRAM scaling bottlenecks: leakage and short-channel effects.
Changxin has not taken this path on a whim. Since its establishment, it has initiated 4F² research and patent layout, publicly holding over 300 patent families, leading both domestic and international manufacturers. Papers were published in 2021, with follow-ups in 2023 and 2025. In 2025, it completed the first experimental verification of DRAM and CMOS wafer-level bonding on a 12-inch wafer platform at IEDM.
Can't Samsung and SK Hynix do it?
They have long researched it. They have the technical reserves.
Samsung has identified working 4F² architecture chips at the 10a process node but has pushed mass production to 2028. SK Hynix announced a 4F² technology roadmap in Japan last year but has yet to disclose any mass production plans.
According to insiders in the Korean semiconductor industry: "Samsung and SK Hynix both have 4F² technology but have not applied it to mass production due to cost issues."
Why are cost issues so fatal?
The 6F² architecture has been highly profitable for many years. DRAM prices began soaring in the second half of 2025; in Q1 2026, contract prices for general DRAM rose 55%–60% quarter-over-quarter, with spot prices increasing over 300% cumulatively.
The profits from 6F² are too rich. Switching to 4F² means revamping production lines, reconfiguring equipment, and enduring yield ramp-up periods with reduced output. In a storage supercycle, no Korean manufacturer is willing to sacrifice current profits for "technological leadership."
It's not that Korea can't produce 4F²; they dare not do it.
Changxin dares to do it precisely because it has little to lose at the 6F² table.
Changxin's current mainstream process is around 16 nanometers and faces restrictions in acquiring EUV lithography equipment. On the 6F² architecture track, it has chased with "cost-effectiveness" for many years; the gap is narrowing, but the ceiling is clear.
Rather than chasing on someone else's track, it's better to switch tracks.
Changxin chose to forcibly enter 4F² during a window of soaring storage prices and expanding profit margins, exchanging technology for market share. The first-generation product directly targets server-grade DDR5 RDIMM — the data center and enterprise market, the most profitable and highest barrier sector.
Publishing technology does not equal mass production success. Yield, long-term reliability, ECC behavior, power consumption, platform validation — these are the real hurdles for server-grade products. Korean analysts point out that Changxin still needs to overcome the triple barriers of yield, performance, and customer certification, and EUV equipment limitations are a real constraint.
But the market has voted with real money. The KOSPI plunged 5.46%, foreign investors fled 4.5 trillion KRW in one day; this is not a reaction to "a piece of news" but a preemptive liquidation over "the possible redistribution of future pricing power."
Samsung and SK Hynix are guarding 6F² profits, waiting for a "safer time."
Changxin has no safe time to wait.
Korea is not incapable of producing 4F²; they dare not do it. Changxin dares because it has little to lose at the 6F² table.
This catfish has already entered the water. The oligopoly pricing system of storage chips is feeling the real heat from a "second-tier pursuer" for the first time.
$SNDK$SAMSUNG$CXMT#韩股急跌后反弹收窄，半导体买盘能否持续？
Disclaimer: OKX Orbit content is provided for informational purposes only. Learn more
