China's Leading Memory Module Makers See Profits Surge Over 300x in First Three Quarters, but Q3 Shows Broad Sequential Decline
nashnova research
China's A-share memory-module sector delivered triple-digit-multiple profit growth through the first nine months of 2026, but Q3 profits dropped sharply across the board — the spread between cheap inventory and soaring contract prices is closing, and the windfall window has passed its inflection point.
How extreme were the first nine months?
BIWIN Storage Technology (佰维存储) posted net profit growth of 28,728% to RMB 8.77 billion; Techwinsemi Technology (德明利) grew 32,400% to RMB 8.75 billion — hundred-fold gains were the norm, not the exception.
Longsys Electronics (江波龙) led in absolute scale: net profit of RMB 14.32 billion, up 1,910%; revenue RMB 38.72 billion, more than doubling year-on-year.
This means → Over three quarters, module makers rode a rare "scissors spread" — they held inventory bought cheap while contract prices were surging.
Why did Q3 hit the brakes so hard?
Longsys's Q3 net profit fell from RMB 6.715 billion in Q2 to RMB 3.745 billion, down roughly 44% quarter-on-quarter. BIWIN dropped from about RMB 4.267 billion to roughly RMB 1.602 billion, down about 62.5%.
In plain terms = The money came from a time lag. Q1 conventional-DRAM contract prices rose 90%–95% quarter-on-quarter; Q2, 58%–63%; by Q3, just 13%–18% (TrendForce data). Prices were still rising — but the rate of increase collapsed, and margins compressed with it.
Shannon Semiconductor Technology (香农芯创) fell roughly 27% q/q; Toyou Feiji Electronics saw Q3 revenue drop 55.5% q/q — the pullback was sector-wide, not isolated.
Were there any exceptions?
Techwinsemi was one of the few that held: Q3 net profit of RMB 2.729 billion, up about 2% from Q2. Its first-nine-month enterprise-storage revenue reached RMB 8.237 billion.
Mason Technologies went the other way — revenue fell 8.9% y/y, net profit rose just 1.31%. It barely participated in the cycle.
This reflects a divergence already forming within the same upcycle: companies skewed toward enterprise-grade products held up better.
Can AI and enterprise storage fill the gap?
BIWIN's AI-edge storage products generated roughly RMB 4.823 billion in revenue, up 389.64% y/y, with customers including Meta, Google, Alibaba, and Xiaomi.
Longsys's enterprise-storage segment posted RMB 2.14 billion in H1 revenue, up 208.8% y/y. Its in-house controllers have shipped over 280 million units cumulatively; a 5 nm UFS 4.1 controller is now in mass production.
This means → The new growth vectors are real, but not yet large enough to fully offset the margin compression from a closing scissors spread — that gap is the key variable for whether sector profits can hold.
What does the balance sheet look like?
BIWIN's inventory ballooned from RMB 7.868 billion at end-2025 to RMB 24.053 billion. Operating cash flow was negative RMB 10.041 billion; the debt-to-asset ratio stood at roughly 68.3%.
Shannon's prepayments surged 1,427% y/y to RMB 9.461 billion. In plain terms = to lock in supply, the company is paying upstream chip vendors far in advance, at scale.
Longsys held RMB 28.456 billion in inventory but turned its operating cash flow positive at RMB 877 million in Q3 — relatively healthy within the sector.
Is the super-cycle over?
Kong Rong, deputy head of Guolian Minsheng Securities Research, told China Fund News: "As long as model iterations continue, memory demand will keep growing" — calling the super-cycle far from finished. Nomura Securities likewise expects the memory shortage to last through 2028.
Shannon chairman Huang Zewei sees the shortage persisting to 2028 at minimum, potentially extending to 2029 in an extreme scenario.
But Q3 data already makes one thing clear: the cycle may continue, but the phase of earning easy money on the spread is ending — from here, what matters is product mix, customer quality, and cash-flow discipline.
市场有风险，内容仅供研究参考，不构成投资建议。