CXMT's 4F² DRAM cell comes out 0.8% larger than Korea's 6F² cell on one professor's line widths
CXMT said it will unveil a 4F² server DDR5 module before year-end but did not disclose capacity, performance or a mass-production date. A Sungkyunkwan professor expects meaningful volume no earlier than end-2027, against a projected 2028 for Samsung.
The Investor · Invest desk
Three cell areas. A 4F² cell at a 15 nm line width covers about 900 square nanometers. A 6F² cell at 12.2 nm covers about 893. A 4F² cell at 12.2 nm would cover about 595.
| Figure | Value | Note | Claim |
|---|---|---|---|
| 4F² cell at a 15 nm line width | 900 sq nm, about | 14 | |
| 6F² cell at 12.2 nm | 893 sq nm, about | 15 | |
| 4F² cell at 12.2 nm | 595 sq nm, about | 17 |
What happened
- Stock markets in South Korea and the United States were rattled by CXMT's plan, according to en.sedaily.com, which did not say how far the shares moved.
- Kwon Seok-joon of Sungkyunkwan University noted in a social media post that the announcement refers to neither mass production nor samples.
- Samsung presented a working 16-gigabit DRAM prototype with a 4F² structure at this year's International Solid-State Circuits Conference.
- Last month CXMT announced mass production of fifth-generation DRAM made with self-aligned quadruple patterning and no EUV, a separate achievement from 4F².
Why it matters
- contradiction The 50% theoretical density gain and the near-identical cell areas at actual line widths give opposite answers, so the comparison chosen decides whether Korea looks behind.
- constraint Server DRAM still has to clear yields and customer qualification, so on Kwon's timeline a visible share shift is unlikely before the end of 2027.
- exposure Korean makers' EUV advantage is less protective if cell geometry can substitute for line width, and the source says the switch cuts into the time export controls bought.
Put Kwon Seok-joon's line widths through the cell formulas and the density claim shrinks. A 4F² cell at CXMT's 15 nanometers covers 4 x 15 x 15, or 900 square nanometers [14]. A 6F² cell at the 12.2 nanometers Korean makers use covers 6 x 12.2 x 12.2, or about 893 [15]. CXMT's cell comes out 0.8% larger [16]. Kwon, a chemical engineering professor at Sungkyunkwan University, explained that the two cell areas are similar in size [6].
The theoretical case is real, and some in the industry fear the structural change could leapfrog Korean technology, according to en.sedaily.com [13]. At the same line width a 4F² cell is one-third smaller and theoretical density rises 50% [5]. Give CXMT's structure Korea's 12.2 nanometers and the cell falls to about 595 square nanometers, a third under 893 [17]. Whether CXMT can take 4F² below the 15 nanometers Kwon cites is the open question. If it can, we are wrong. If its first part ships at 15, the stock move is pricing a press statement.
Timing weakens the leapfrog reading further. Kwon does not see meaningful volumes before the end of 2027 at the earliest, and thinks the first half of 2028 is the realistic date [4]. That is at least a year after the unveiling CXMT promised for this year [18]. Projections put Samsung's mass production around 2028 [19], so on those dates any CXMT lead is at most about a year [20]. Kwon said it was a stretch to interpret China as being two years ahead [19].
Asked by reporters at the Semiconductor Exhibition on the 14th whether Samsung was developing 4F² technology, its CTO, Song Jai-hyuk, answered "yes" and said only that the company was "working hard at it" [7].
The cell change is the export-control workaround. Samsung and SK hynix use EUV, while CXMT relies on multi-patterning with DUV equipment [9]. Hwang Cheol-seong, a chair professor at Seoul National University, said that "for China, which has difficulty securing EUV, switching the cell structure may be a realistic alternative" [11]. The article adds that the 4F² structure leaves room for comparable density at a relatively wide line width, and that this cuts into the time the controls bought [12].
The module could appear by year-end at 15 nanometers with density near Korea's, slip past 2026, or arrive at a narrower line width. We think the first is likeliest, and that the market reaction is pricing an announcement.
What to watch
- Whether CXMT shows the 4F² RDIMM before year-end with capacity and performance figures attached.
- Any CXMT 4F² line width below 15 nanometers; at Korea's 12.2 the cell would be about 595 square nanometers.
- Yield and customer-qualification news on server parts, and any Samsung 4F² schedule beyond 'working hard at it'.
Clarity's read
What the record supports and how the coverage leans. The claims behind it follow.
Reality
- Evidence40
- Adoption
- Insufficient
- Hype gap+35
- Incentives
- Insufficient
- Confidence50
Claim ledger
Ranked by verification strength, evidence, and original report placement.
- [1]
CXMT said on the 10th that it plans to unveil a server-grade DDR5 memory module (RDIMM) combining a 4F² structure with hybrid bonding before the end of the year.
- [2]
CXMT did not disclose the product's capacity or performance, nor a specific release date or mass-production schedule.
- [3]
According to a social media post by Kwon Seok-joon, a professor of chemical engineering at Sungkyunkwan University, CXMT's announcement contains no reference to either 'mass production' or 'samples', and should be read as a launch announcement rather than the start of full-scale mass production.
- [4]
Kwon expects meaningful production volumes of CXMT's 4F² DRAM no earlier than the end of 2027, and realistically in the first half of 2028; mass-producing server DRAM requires securing yields and winning customer qualification.
- [5]
A 4F² cell measures 2F in each direction, against 3F wide by 2F deep for the prevailing 6F² structure; at the same line width the cell area shrinks by one-third and theoretical density rises 50%.
- [6]
Kwon explained that CXMT's 4F² cells use a 15-nanometer line width while Korean memory makers use 12.2-nanometer 6F² cells, leaving the two cell areas similar in size.
- [7]
Samsung Electronics president and CTO Song Jai-hyuk, at the Semiconductor Exhibition on the 14th, answered 'yes' when reporters asked whether the company was developing 4F² technology; on progress he said only that the company was 'working hard at it.'
- [8]
Samsung presented a working prototype of a 16-gigabit DRAM chip with a 4F² structure at this year's International Solid-State Circuits Conference (ISSCC).
- [9]
Samsung Electronics and SK hynix use EUV to shrink circuit patterns, while CXMT relies on multi-patterning, etching circuits repeatedly with DUV equipment.
- [10]
Last month CXMT announced mass production of fifth-generation (G5) DRAM using self-aligned quadruple patterning (SAQP) without EUV, a separate achievement from mass-producing 4F² products.
- [11]
Hwang Cheol-seong, a chair professor of materials science and engineering at Seoul National University, said that "for China, which has difficulty securing EUV, switching the cell structure may be a realistic alternative."
- [12]
Applying the smaller-area 4F² structure instead of 6F² opens room to achieve comparable density even with a relatively wide line width, which the article says cuts into the time U.S. export controls bought.
- [13]
Some in the industry fear CXMT's change to the structure of the DRAM cell could leapfrog South Korean technology, while many analysts are more cautious, noting that announcing a product and mass-producing it are two different things.
- [14]
A 4F² cell at a 15 nm line width covers about 900 square nanometers.
- [15]
A 6F² cell at a 12.2 nm line width covers about 893 square nanometers.
- [16]
CXMT's 4F² cell at 15 nm is about 0.8% larger than a 6F² cell at 12.2 nm.
- [17]
A 4F² cell at 12.2 nm would cover about 595 square nanometers, roughly one-third below the 893 of a 6F² cell at the same width.
- [18]
Kwon's earliest estimate for meaningful volume (end of 2027) falls at least about 12 months after the year-end 2026 unveiling CXMT announced.
- [19]
Comparing CXMT's plan to launch this year with projections that Samsung will reach mass production around 2028, Kwon said it was a stretch to interpret China as being two years ahead.
- [20]
Setting Kwon's CXMT volume window (end 2027 to first half 2028) against projected Samsung mass production around 2028, any CXMT lead is at most about one year, not two.
- [21]
CXMT's plans to roll out a next-generation DRAM product this year using 4F² technology have rattled stock markets in South Korea and the United States.
Get this ledger inside ChatGPT. Ask about any of our stories and the answer brings its figure, each claim’s verdict and the sources. Free, no account.
Sources
1 independent publisher whose own reporting we read for this story.
- en.sedaily.comChina's CXMT Touts Next-Generation DRAM With No Specs or Launch Date
1 article · October 14, 2026
Topics and entities
Follow any of these and your For You feed starts watching them — no settings page required.
Topics
- Memory chipsFollow
- DRAM cell architectureFollow
- US-China Chip Export ControlsFollow