CXMT Books 1,688% Profit Jump as Memory Crunch Funds Shanghai IPO

ChangXin Memory Technologies booked a 1,688% jump in net profit and a 719% jump in revenue for the first quarter, the Hefei-based DRAM (dynamic random-access memory, the working memory used in phones, PCs, and AI servers) maker disclosed in a Shanghai Stock Exchange prospectus filed Monday, setting up the largest mainland China chip listing in nearly six years. Revenue reached 50.8 billion yuan (about $7.5 billion) for January through March, and net profit cleared 20 billion yuan, against a base of roughly 1.1 billion yuan a year earlier.

Samsung, SK Hynix, and Micron have spent eight straight quarters rotating wafer capacity into high-bandwidth memory (HBM, the stacked DRAM that feeds AI accelerators), leaving the conventional DDR5 and LPDDR5 aisle thin. CXMT, blacklisted by the US Bureau of Industry and Security (BIS, the export-control arm of the Commerce Department) in December 2024 and three process generations behind the oligopoly, is the cheapest seller in the market the leaders walked away from.

The Numbers Behind the 1,688% Jump

Monday’s filing is an update to a draft prospectus CXMT submitted earlier this year. Four figures carry the weight.

• 50.8 billion yuan: Q1 2026 revenue, against roughly 6.2 billion yuan in Q1 2025
• 20-plus billion yuan: Q1 2026 net profit, against a 1.1 billion yuan base
• 110 to 120 billion yuan: forecast first-half revenue, or about $17.6 billion at current rates
• 36.65 billion yuan: cumulative accumulated losses still on the books at end-2025

Q1 on its own cleared what CXMT booked across the first nine months of 2025, when revenue reached 32 billion yuan, up 97.8% year on year. Twelve months ago the company had never logged an annual profit. Its first profitable full year arrived in 2025, with net income of 2 to 3.5 billion yuan, a rebound credited to a DRAM price recovery that began in the second half of last year.

The earlier losses came from CXMT building out fabs and writing down inventory through the 2022 to 2023 trough. The same inventory, repriced into a 2026 shortage, is now part of the upside the prospectus puts in front of Shanghai investors.

Why Samsung and SK Hynix Built CXMT’s Quarter

Three of the four DRAM majors account for more than 95% of global supply. Since 2024 each has aggressively shifted capacity toward HBM, where revenue per wafer runs three to five times higher than commodity DDR5. By the start of this year, HBM had absorbed roughly 23% of total DRAM wafer output, per market trackers cited by analysts at S&P Global.

The arithmetic on the other side of that pivot is what shows up in CXMT’s prospectus. Spot DDR5 32-gigabyte module pricing climbed from about $149 to $239 between summer 2025 and the end of September, a 60% move reflected in trade-channel data. Contract DDR5 pricing more than doubled across the same window. The US Commerce Department’s December 2024 entity-list package on Chinese advanced computing had been designed to slow precisely the kind of expansion CXMT now plans to fund with public capital.

HBM, DRAM, and NAND capacity is essentially sold out for 2026.

The line is from SK Hynix’s October 2025 earnings call, where management told analysts that book commitments already covered the year ahead. Both major Korean and US producers have warned that AI-driven memory shortages may extend into 2027 and beyond, with hyperscale customers reserving supply years in advance.

The HBM race itself remains years away for CXMT. What the prospectus shows is volume DDR5 and LPDDR5 shipped into a market the Big 3 no longer prioritize, at prices set by the leaders’ decision to point their best wafers elsewhere.

The $4.2 Billion Shanghai Raise

The IPO targets 29.5 billion yuan, or roughly $4.2 billion at the current rate. At an implied valuation of about 300 billion yuan (close to $42 billion), the deal would be the second-largest listing on Shanghai’s Sci-Tech Innovation Board (STAR Market, the Nasdaq-style technology venue launched in 2019) since SMIC’s 2020 offering raised 53.2 billion yuan. CXMT is the first issuer to use the STAR Market’s pre-review confidentiality mechanism, introduced in mid-2025 to let strategically sensitive firms hold technical disclosures back from rivals during the review window.

Up to 10.62 billion new shares would be issued. Government-linked vehicles already hold close to 40% of the company before any dilution. Hefei municipality’s chip-investment platform owns 21.67%, the National Integrated Circuit Industry Investment Fund Phase II (known in the industry as “Big Fund II”) holds 8.73%, and Anhui Investment Group adds 7.91%. Alibaba, Xiaomi, China Life Insurance, and China Taiping fill out the strategic register.

Use of proceeds, per the prospectus, splits across three buckets:

1. 13 billion yuan to Phase II wafer fabrication capacity at the Hefei main site
2. 7.5 billion yuan to upgrade the existing memory wafer line, including the HBM3 production capability targeted for year-end 2026
3. 9 billion yuan to forward-looking research on next-generation DRAM, including the move toward sub-15-nanometer process nodes

The Shanghai Stock Exchange has accepted the filing for review. Its listing committee has not yet scheduled the registration hearing that would clear the path to pricing.

The DRAM Oligopoly Adds a Seat

By production volume, CXMT is now the world’s fourth-largest DRAM maker. The leaders still hold a comfortable technology margin, but the gap on volume has narrowed faster than US controls were designed to allow.

Sources: company filings; S&P Global Market Intelligence memory supply research.

Output composition matters as much as node geometry. The Big 3 derive an outsized share of profit from HBM and the most advanced DDR5. CXMT books almost all of its revenue from conventional DDR4, DDR5, and LPDDR5 destined for Chinese-brand smartphones, PCs, and second-tier servers. That mix is the reason the 719% revenue jump translated into a 1,688% profit jump: spot and contract pricing was a tailwind across the basket, and CXMT’s cost base did not move proportionally with it.

The Equipment Ceiling Has Not Moved

Beijing’s victory lap will be qualified. The December 2024 BIS package added 140 firms to the US Entity List, including CXMT, and tightened licensing for the tools needed to push DRAM into more advanced nodes and credible HBM stacking. The post-IPO capital does not buy those tools.

The constraints CXMT still faces include:

• EUV lithography access: ASML extreme ultraviolet scanners, required for sub-10 nm DRAM, remain prohibited for export to Chinese fabs under coordinated Dutch and US controls
• HBM stacking and TSV equipment: several US suppliers cannot ship the through-silicon-via and hybrid-bonding tools that make HBM3E and HBM4 economically viable at scale
• Spare parts and field service: existing Applied Materials, KLA, and Lam Research tools at CXMT face restrictions on service contracts and replacement components, raising the cost of running aging fabs

CXMT’s stated roadmap is to mass-produce HBM3 in Shanghai by the end of 2026, three generations behind SK Hynix’s HBM3E volume and four behind the HBM4 cohort. Industry analysts at Counterpoint and TrendForce have put the gap on advanced nodes at roughly three years. The Peterson Institute analysis of the 2024 export-control package tracks how Washington has periodically tightened the screw on Chinese memory specifically. Funded capacity is one variable. The technology gap is the other, and no amount of yuan raised in Shanghai shrinks it on its own.

Pricing the IPO Against a 2027 Shortage Window

Demand-side, CXMT is filing into the strongest memory cycle in a decade. Samsung’s revenue per bit on traditional DRAM is forecast to rise 116% year on year in 2026 to $0.79; the second Korean leader is projected up 78% to $0.70; Micron 54% to $1.06. CXMT’s blended ASP (average selling price) has trailed those numbers but moved in the same direction. The first-half forecast of 110 to 120 billion yuan implies a second quarter that lands flat to slightly down from the Q1 print, consistent with what the Big 3 have signaled for the HBM-led ramp through year-end.

What the listing committee, retail bidders, and Big Fund II are really pricing is whether that strip holds into 2027. If the leaders are correct that AI-driven shortages run that long, the IPO becomes the funding leg of a credible HBM3 ramp at a moment when the conventional DRAM market still pays well for second-tier output. If Big 3 capacity rebuilds faster than guided, or if Chinese-brand smartphone and PC demand softens into a tariff cycle, the deal will have priced the cycle’s peak.

The Shanghai listing committee will set the registration hearing within weeks. If third-quarter DDR5 contract pricing holds the strip the Big 3 have signaled, the prospectus reads as a base for CXMT’s first profitable cycle. If the strip cracks earlier than guided, the Q1 print is the peak the IPO priced.