CXMT's Q3 revenue jumped 613% as China's DRAM challenger eyes HBM3E
CXMT, China's homegrown memory maker, reported its first quarterly results since its July listing: revenue of ¥118.71 billion, up 613% year over year, with an 89.53% gross margin. Supply-chain reports say limited HBM3E production starts within two months — a direct challenge to the memory giants.
📊 Scorecard: our Bearish call on $MU is scored from $690.00, the last close before it ran. Its first result comes after the next close.
The News
China's homegrown memory maker just posted numbers that would turn heads in any industry. CXMT — ChangXin Memory Technologies — reported its first quarterly results since listing in July, filing on October 30: third-quarter revenue of ¥118.71 billion, about $17.8 billion, up 613.18% from a year earlier and 19.30% from the second quarter. Gross margin — the share of revenue left after the direct cost of making the chips — was 89.53%, up 1.94 points from Q2, and net profit attributable to shareholders was ¥54.43 billion, about $8.2 billion. CXMT is listed on Shanghai's STAR Market (ticker 688825.SS), which is outside the tickers this site tracks, so the public read-through is its listed rivals: Micron, Samsung, SK hynix and SanDisk.
The filing itself explains the surge in plain terms. A global DRAM shortage — DRAM is the working memory in phones, PCs and servers — left standard memory in short supply because the big suppliers tilted their factory capacity toward HBM, the high-bandwidth memory that AI processors need. CXMT, which kept building standard DRAM capacity, sold into the gap, and its own capacity build-out and shift toward higher-value products did the rest. DDR5, the current standard for PC and server memory, and LPDDR6, the low-power version for phones, are now in mass production at CXMT.
The forward signal is the one the incumbents will watch. Reuters reported on October 28, citing sources, that CXMT will begin limited production of HBM3E — the latest generation of high-bandwidth memory — within about two months, with yields reported at 20–30%, against roughly 70% at SK hynix. CXMT itself has not confirmed the timing or the yields. Yields are the share of chips made that actually work, so a 20–30% yield means most of the silicon is wasted — but it is also how every new memory maker starts. The figures are reported by supply-chain sources, not confirmed by the company. On technology direction, the big names have now all declared: CXMT is pushing its 4F² chip design toward commercial production, with engineering samples validated and DDR5 RDIMM samples — test memory modules for servers — going to customers by the end of 2026; Micron said on October 21 it will extend its existing 6F² design and go straight to 3D DRAM instead; Samsung confirmed on October 14 it is developing 4F², with media reporting a volume target of 2028; and SK hynix said on October 28 it will move to 4F² but it is not the current focus.
Revenue up 613% — the shortage that made the memory giants' year is now funding their challenger.
Why It Matters
A 613% revenue jump with an 89.53% gross margin is not a niche story — it is a new supplier getting rich on the same shortage that has made memory the hottest corner of the chip market. CXMT's growth came from the gap the big three left when they tilted capacity to HBM: standard DRAM got scarce, prices rose, and the Chinese supplier that kept building standard capacity collected the difference. That is the bearish read for the incumbents. Every point of share CXMT takes in standard DRAM is a point Micron, Samsung or SK hynix does not get, and the shortage that made their year is the same shortage funding their challenger.
The HBM3E plan is earlier and shakier, and it should be read that way. Reported yields of 20–30% against SK hynix's roughly 70% mean CXMT cannot yet make HBM profitably at volume, and the reports are supply-chain talk, not a company announcement. But limited production within two months would still be a milestone: it would put a fourth name on the HBM supplier list, and HBM is the product where the incumbents earn their fattest margins. The 4F² split is equally telling — CXMT and Samsung are betting on the new chip design, Micron is skipping it for 3D DRAM, and SK hynix is waiting. The technology map of memory is being redrawn, and the new lines are being drawn in China.
Analyst's Take
Our call is Bearish — for the incumbents' share, not for memory demand. Nothing in CXMT's quarter says the memory boom is ending: demand is so strong that a challenger grew 613% and still posted an 89.53% margin. What it says is that the boom is now funding competition. Micron at $690.00 (Friday close), Samsung at ₩189,900 and SK hynix at ₩1,136,000 (Monday closes), and their valuations assume the shortage stays theirs to harvest. A rival growing sixfold off the same shortage is the first crack in that assumption.
What to watch: whether CXMT's limited HBM3E production actually starts within two months, and whether reported yields improve from 20–30%; the company's DDR5 RDIMM samples to customers by the end of 2026; and the incumbents' next capacity decisions — whether they tilt back toward standard DRAM to defend share, or keep betting everything on HBM. If CXMT's yields climb, the HBM club gets a fourth member and the pricing power of the big three gets its first real test.
Are you bullish or bearish on this story?
By The Numbers
FAQ
What is CXMT?
ChangXin Memory Technologies, China's homegrown DRAM maker. It listed on Shanghai's STAR Market in July 2026 (ticker 688825.SS) and reported its first quarterly results on October 30. This story's tickers — Micron, Samsung, SK hynix and SanDisk — are its listed rivals, whose business it affects.
Why did CXMT's revenue jump 613%?
Two reasons, per its filing: a global DRAM shortage left standard memory scarce because the big suppliers tilted factory capacity to HBM for AI chips, and CXMT kept building standard DRAM capacity and shifting to higher-value products, so it sold into the gap.
What is HBM3E, and why does it matter?
HBM3E is the latest generation of high-bandwidth memory, the fast memory stacked next to AI processors. It is the highest-margin product in memory, and today only SK hynix, Samsung and Micron make it at volume. CXMT reportedly starting limited production would add a fourth supplier.
What is 4F²?
A new way of laying out DRAM chips that packs more memory into the same silicon. CXMT and Samsung are pursuing it, Micron is skipping it to go straight to 3D DRAM, and SK hynix says it will move to 4F² eventually but it is not the current focus.
Are the HBM3E yields confirmed?
No. The 20–30% yield figure, versus roughly 70% at SK hynix, comes from supply-chain reports, not from CXMT itself. Yields are the share of chips made that actually work, so low yields mean most silicon is wasted — but they are also normal for a new production line.
👁
Source: Digitimes / Reuters (via TrendForce)