Memory Stocks Rally With Korea Closed: Nanya and CXMT Hit Records While Kioxia and SNDK Sit 45% Below Theirs
Monday’s memory tape looked like one trade and was actually two. Kioxia closed up 15.07% at ¥61,840 in Tokyo, inside a Nikkei that gained 0.74% to 69,220.25. CXMT closed up 12% at ¥61.80 in Shanghai, a record, lifting its market capitalisation to ¥4.13 trillion and holding the title of China’s most valuable listed company, which it took from Tencent on Thursday. Nanya extended a month that has now delivered more than 30%. Micron traded up 3% pre-market, SanDisk up 5%, and the SK Hynix ADR up 4%. South Korean markets were shut for a holiday.
That last detail is not colour. With Samsung Electronics and SK Hynix untradeable, roughly two-thirds of global DRAM supply had no listed expression on the day. Every dollar that wanted memory exposure had to route through the names that were open: Kioxia, CXMT, Nanya, the ADRs, and the US pre-market. Smaller float, same flow, larger prints. The correct read of a 15% move in Kioxia and a 12% move in CXMT on a day when the two largest producers are closed is that some unknown share of it is plumbing rather than repricing, and the only way to size that share is to watch what Seoul does when it reopens. If SK Hynix and Samsung confirm on Tuesday, the move was real. If they open flat and the Asian names give back half, the move was a liquidity artefact wearing a thesis.
The divergence underneath is the part worth owning. Nanya and CXMT printed fresh all-time highs. Kioxia did not. At ¥61,840 the stock is roughly 45% below its 52-week high of ¥112,700, which is almost exactly where SanDisk sits against its June record. Two groups, one headline, opposite positions on the chart. The names making new highs are the mature-node DRAM stories and the China domestic-substitution story. The names bouncing are NAND. Anyone reading Monday as broad confirmation that the memory cycle has resumed is aggregating a breakout and a retracement into a single number.
What Nanya and CXMT are actually pricing is the legacy-node squeeze, not the AI pull. Nanya’s July revenue was NT$43.87 billion, up 49.3% month over month and 719.6% year over year, driven by contract price increases concentrated in DDR4 and DDR3. Its gross margin ran 79.5% and net income NT$50.19 billion, up 1,324%. Those are not the economics of a company winning an advanced-node fight. TrendForce put Nanya fourth among global DRAM suppliers in the first quarter of 2026 at 1.6% share, against Samsung at 39%, SK Hynix at 26%, Micron at 25% and CXMT at 7%. A 1.6%-share producer earning 79.5% gross margin on DDR4 is a measurement of how far the top three have walked away from mature nodes, not a measurement of Nanya’s competitive position. The moat here is temporary and it belongs to someone else: it exists because Samsung, SK Hynix and Micron reallocated wafers to DDR5 and HBM, and it closes the moment any of them decides the legacy premium is worth serving. Counterpoint had Nanya’s second-quarter DRAM revenue up 690% year over year, second only to CXMT’s 716%. Both of those numbers are ASP, not bits.
The NT$346.6 billion Fab 5A commitment is being read as a bullish supply-cycle signal, and it is a bullish signal about equipment orders rather than about DRAM pricing. The board approved it on 5 August, alongside lifting 2026 capital expenditure from NT$52 billion to NT$69.7 billion, a 34% increase that goes mostly to equipment prepayments. It was not Monday’s catalyst. More importantly, it is not near-term supply. Wafer production at Fab 5A starts in the second half of 2027, reaches 30,000 wafer starts per month in 2028 and 35,900 in 2029, with 45,000 as the eventual ceiling and total project cost estimated near $16 billion. The fab introduces 10nm-class 1b, 1c, 1d and 1e nodes with EUV. Every one of those dates falls outside the window in which current contract pricing is set. The immediate cash consequence of the announcement lands on the lithography, deposition and hybrid-bonding suppliers taking the prepayments, and the company has separately signalled it intends to quadruple 2027 capital spending to roughly $6.2 billion. A 1.6%-share producer adding 45,000 wafers per month by the end of the decade is a 2029 supply question. It is a 2026 order book.
The asymmetry between the DRAM breakouts and the NAND drawdowns has a structural explanation, and it cuts in the opposite direction to how the tape is treating it. Nanya and CXMT are levered to spot and to short-tenor contracts, which is why their revenue lines move 700% and their share prices make new highs on a monthly print. SanDisk and Kioxia have been converting the opposite way, toward contracted volume at floor prices, which is exactly what the 13 August investor day set out. Contracted revenue is worth more and reprices less. That is why the NAND pair lags on a sentiment day and why it should hold better if DDR4 contract prices roll. The market is currently paying the higher multiple to the more cyclical exposure.
The number that decides which of Monday’s two trades was correct is the direction of DDR4 and DDR3 contract prices into September. Nanya’s entire re-rating, and a meaningful part of CXMT’s, rests on legacy DRAM holding a premium to DDR5 that exists only because the top three producers chose to vacate the node. Watch for the first indication that any of them is reallocating capacity back.