SemiAnalysis: CXMT Prepares Record China Semiconductor IPO as Memory Supercycle Drives 6x Revenue Growth
June 23, 2026 – Deep dive on China's leading DRAM maker ahead of landmark STAR Market listing
China's ChangXin Memory Technologies is positioning for what could become the country's largest semiconductor IPO in years, with a STAR Market listing expected in the coming months following formal CSRC registration submitted on May 27. The timing could hardly be better. CXMT reported revenue of approximately $8.6 billion for 2025, up 156% year-over-year, and first quarter 2026 revenue of $7.3 billion, representing roughly 700% year-over-year growth. The company turned its first annual profit in 2025 at $1 billion after carrying an accumulated deficit of roughly RMB 36.65 billion built up over nearly a decade.
SemiAnalysis projects CXMT's full-year 2026 revenue could exceed $50 billion, driven almost entirely by explosive pricing rather than dramatic market share gains. The firm's bit shipments increased only 11% in first quarter 2026, while average selling prices rose approximately 57%, following sequential ASP increases of 63% and 68% in third and fourth quarter 2025 respectively. This pricing strength is not unique to CXMT but reflects what SemiAnalysis has characterized as a "once-in-four-decades shortage" across the DRAM market, with pricing expected to double again this year.
Pricing Supercycle Masks Technology Gap
What challenges conventional assumptions about Chinese memory is CXMT's pricing power. In first quarter 2026, CXMT's DRAM ASP was only 5% to 10% below Samsung, SK Hynix, and Micron in the same period. SemiAnalysis expects this gap to widen through 2027, driven not by inherent pricing differences but by product mix. Leading suppliers continue to benefit from a higher server DRAM mix in their bit shipments, and by the end of 2027 the firm expects server DRAM and HBM to account for well over 50% of total DRAM end-market demand. Given that server DRAM and HBM carry higher dollars per gigabyte than other memory end markets, this should allow leading suppliers to widen their ASP gap versus CXMT as the mix shifts.
The pricing environment has materially improved CXMT's margin profile. The company's 2025 gross margin reached 37.8%, moving closer to Samsung at 39.4% and Micron at 39.8%, though it remains far below SK Hynix at 60.4% due to SK Hynix's much higher HBM mix. CXMT's operating margin reached 70% in first quarter 2026, compared with SK Hynix at 73%, Samsung at 81%, and Micron at 84% in the same period. This represents a dramatic swing from negative 113% in 2023 and negative 4.7% in 2024.
The margin improvement is almost entirely pricing-driven rather than cost-competitive. SemiAnalysis cost-per-bit analysis on DDR5 products shows CXMT's cost per bit remains more than 30% higher than the three leading suppliers. Because DDR5 pricing was already exceptionally strong in first quarter 2026, this still lifted CXMT's gross margin above 70%. Nearly all of the company's bit sales are conventional LPDDR and DDR products, with HBM remaining a minimal contributor to revenue and earnings.
Capacity Expansion Without Market Disruption
By the end of 2026, SemiAnalysis expects CXMT to reach roughly 350,000 wafer starts per month, only modestly below Micron's estimated 385,000 wafer starts per month. This would position CXMT close to becoming the industry's third-largest memory supplier if ranked by wafer capacity alone. The firm remains materially behind Samsung and SK Hynix, estimated at approximately 720,000 and 595,000 wafer starts per month respectively. By 2027, with the initial ramp of Shanghai Phase 1 and the full ramp of Hefei and Beijing, CXMT capacity could reach 420,000 wafer starts per month by year-end, representing roughly 17% of global DRAM capacity, up from approximately 13% in 2025.
Despite this expansion, SemiAnalysis argues that concerns about supply-demand disruption from Chinese players are overplayed for at least the next two years. Factoring in incremental wafer capacity and bit shipments from CXMT and other memory suppliers, and assuming utilization rates in the high-90% range, the firm continues to see DRAM as extremely supply constrained. Looking solely at CXMT's wafer additions, the firm expects roughly 85,000, 70,000 and 80,000 wafer starts per month added each year from 2026 to 2028, versus Samsung at 15,000/50,000/110,000, SK Hynix at 60,000/60,000/90,000, and Micron at 30,000/90,000/115,000. Even with these wafer additions, SemiAnalysis expects DRAM to remain undersupplied by a high-single-digit percentage this year, widening to a low- to mid-teens bit undersupply next year.
HBM Remains the Achilles Heel
CXMT's wafer allocation to HBM has been extremely limited. By the end of 2025, SemiAnalysis believes only approximately 5,000 wafer starts per month of CXMT's roughly 265,000 wafer starts per month capacity was allocated to HBM. The firm thinks this figure will increase to nearly 30,000 wafer starts per month and 55,000 wafer starts per month by the end of 2026 and 2027 respectively. This capacity trajectory aligns with CXMT's filings indicating that roughly 99% of revenue consists of DDR and LPDDR products in 2025.
The limited HBM allocation reflects both strategic economics and technical challenges. Commodity DRAM currently offers materially higher margins than CXMT's HBM products, while also delivering more than 3x the bits per wafer on a like-for-like basis. On technology readiness, SemiAnalysis believes CXMT is still struggling to stabilize supply for HBM3 8-high, with even greater challenges in 12-high. On the front end, the company appears to have made progress stabilizing production of its Gen 4, or 1z-equivalent DRAM, which the firm believes will represent the majority of CXMT's DRAM output this year. However, front-end wafer-sort yield remains materially lower for the core DRAM die used in HBM, given its larger die size, more demanding cell performance, and overall performance requirements versus commodity DRAM.
Die stacking remains the major obstacle for CXMT's HBM. HBM stacking introduces significant technical hurdles, including thermal stress, die cracking, warpage, bonding defects, and yield loss across multiple stacked dies. These issues become even more severe as the company attempts to move from HBM3 8-high to HBM3 12-high, and eventually HBM3E. SemiAnalysis models CXMT's HBM3 8-high front-end and back-end yields at roughly 35% and 70% respectively, implying an overall yield of only around 25%. The firm thinks this number should be lower when the company attempts to produce HBM3 12-high or HBM3E 12-high given higher difficulty in die stacking and bonding.
SemiAnalysis notes there is an increasing possibility that CXMT will skip HBM3 and focus instead on HBM3E 8-high and 12-high. The firm believes this potential roadmap change is driven by customer demand for more competitive HBM products in the 2027 timeframe and the fact that mainstream accelerators will be equipped with HBM3E, HBM4, and HBM4E. Importantly, Huawei and CXMT will have custom HBM that is not based on the slow JEDEC standards and physical interfaces, which should allow them to close the bandwidth disadvantage.
CXMT's HBM struggles continue to be reflected in its limited product presence and slow penetration within China's AI accelerator market. SemiAnalysis thinks only Huawei, Cambricon, and select emerging Chinese AI chip startups are likely to adopt CXMT's HBM, though the firm suspects adoption rates will be large. The firm actually believes domestic AI accelerator vendors would still prefer foreign HBM3, or even HBM3E, if they can secure supply through any available channel or stockpiled inventory before the export controls implemented in December 2024.
The Qimonda Inheritance and Talent Foundation
CXMT's technological foundation traces back to Qimonda, the European DRAM player that went bankrupt in January 2009 during the Global Financial Crisis. In December 2019, Polaris Innovations and CXMT signed an agreement for a license to a large set of DRAM patents. CXMT leaders once publicly claimed they obtained roughly 2.8 terabytes of Qimonda technical documentation, which became the basis for the company's DRAM business.
A major technology CXMT inherited from Qimonda is its 46nm-class buried wordline cell, which CXMT scaled toward the 10nm class. Buried wordline is the load-bearing idea. Instead of routing the access transistor's gate across the wafer surface, buried wordline sinks it into a trench beneath the bitline. This pulls the gate off the surface so the cell collapses to a 6F2 layout versus 8F2, lengthens the channel without spending surface area to suppress short-channel leakage that destroys retention, and cuts gate-to-bitline parasitic capacitance. Buried wordline plus stacked capacitor is the architecture all three leading players run today.
Beyond patents, the more durable asset CXMT extracted from Qimonda's collapse was its engineers. CXMT successfully attracted Karl-Heinz Kuesters, a senior engineer from Qimonda's German site, to Hefei. Kuesters had spent 24 years as a VP of technology and pre-development across Siemens, Infineon, and Qimonda. EE Times called Kuesters the company's "ace in the hole." What Kuesters brought was the part of Qimonda's legacy that neither the patents nor the 2.8 terabytes of documents carry: the tacit know-how of integration and yield judgments that no patent records.
CXMT has also recruited heavily from Korea and Taiwan. Korean prosecutors have charged former Samsung employees with leaking technology to CXMT, and dozens of Korean engineers are reported to have worked there. The Qimonda patents were always a finite, expiring asset. What lets CXMT keep moving, from Gen 4 to Gen 5 and now into HBM, is the assembled capability of domestic talent, Chinese nationals who worked at foreign companies and returned, as well as some experts from foreign firms, not the documents.
Hefei's Patient State Capital Model
It is difficult not to attribute at least part of CXMT's success to the strong support of local and central government in China. Hefei municipal government is one of the great examples of patient state-venture capital nurturing successful companies over the past two decades, from BOE to NIO and now to CXMT. The Hefei government did two things for CXMT. First, it helped build a local supply chain around the fab. The city took a large equity stake in CXMT as the anchor and then drew the rest of the chain in around it.
Surrounding CXMT's plant in Hefei's airport-zone industrial park, the government produced a dense local cluster. Two packaging-and-test houses, Peyton and Xinfeng, are located within a street or a wall of CXMT's fab, with Xinfeng booking over 99% of its revenue from CXMT. An on-site bulk-gas plant run by Guanggang supplies most of CXMT's needs, while wafer-reclaim capacity from Zhiwei Semiconductors, a subsidiary of Zhichun Technology, is located in Hefei's Xinzhan Hi-Tech District. The state-venture-capital vehicles have also taken outright control of an upstream chip-molding equipment maker, Wenyi Technology.
In addition, Hefei's state-venture capital could afford to lose money for a very long time. Unlike a private venture-capital fund answerable to limited partners that expect a return on a fixed timetable, Hefei's state-venture capital, ultimately backed by the city's municipal and development-zone state entities, faced no such clock. They kept funding a company that, even after turning its first annual profit in 2025, still carried an accumulated deficit of roughly RMB 36.65 billion built up over nearly a decade. The original project launched in 2016 began with Hefei's state-venture capital funding about 80% of the project's first phase, or RMB 14.4 billion of RMB 18 billion. By the IPO, the largest holder, Hefei's Qinghui Jidian, held 21.67% and state-venture capital vehicles together held over 30%.
IPO Structure Reveals Control Reality
CXMT could become one of China's largest semiconductor IPOs, and its ownership structure matters more than the headline financials. CXMT reported RMB 7.14 billion of consolidated 2025 net income, yet only RMB 1.87 billion was attributable to parent shareholders, with 74% attributable to minority interests. The reason is the ownership architecture. CXMT holds 30.68% of the economics of Changxin Xinqiao and 31.72% of Changxin Jidian Beijing while controlling 73.01% and 75.32% of the votes through long-term acting-in-concert arrangements. This lets the company consolidate fabs it mostly does not own, so the consolidated figure overstates what public shareholders will actually receive by roughly four times.
That same voting structure undercuts the company's declaration that it has no controlling shareholder and no actual controller, which the prospectus lists as a formal governance risk. CXMT exercises majority voting control of its fabs through acting-in-concert pacts, and state vehicles including National IC Fund Phase II, Hefei, and Anhui together hold well over 30% even after the listing. The arrangement looks designed to manage export-control and foreign-investor perception at a moment when CXMT's ties to the Chinese state draw the most scrutiny.
CXMT plans to deploy RMB 29.5 billion, or about $4.1 billion, while issuing 10% to 15% of its post-IPO shares. Fully funding those uses through the IPO implies a price of roughly RMB 4.41 at 10% dilution or RMB 2.78 at 15%, versus RMB 2.63 in the June 2025 financing. At RMB 2.78, CXMT would be valued at about RMB 197 billion, or $27 billion, equivalent to just 1.8 times annualized first-half 2026 parent earnings. SemiAnalysis believes this arithmetic floor sits well below a realistic book-building valuation and is too cheap.
Of the RMB 29.5 billion in planned net proceeds, RMB 20.5 billion, or 69.5%, funds wafer-production-line and DRAM technology upgrades, while RMB 9 billion, or 30.5%, supports forward-looking DRAM research. The prospectus discloses no dedicated HBM project and does not mention HBM. The IPO therefore primarily strengthens CXMT's core DRAM manufacturing and technology base, with no disclosed funding commitment to a near-term HBM expansion.
Alibaba's place on the cap table changes how to read CXMT's demand. Alibaba Cloud is at once the anchor hyperscaler customer, a near-4% holder, and an endorser, sitting alongside GigaDevice, chairman Zhu Yiming's own fabless house, at 1.8%. Domestic volume is effectively guaranteed in a way the Korean incumbents did not have in their home markets.
Export Controls Accelerate Equipment Localization
Successive export controls have progressively narrowed CXMT's access to the lithography, deposition, etch, metrology, service, spares, and process support required for advanced DRAM scaling. As a result, CXMT's long-term competitiveness increasingly depends on how quickly it can qualify domestic tools, redesign process flows around equipment constraints, and reduce reliance on restricted foreign suppliers.
The first and most important constraint is lithography. Since 2019, restrictions on ASML EUV exports to China have prevented CXMT from accessing the lithography platform that leading DRAM makers rely on for advanced nodes. Samsung and SK Hynix have already used EUV for the most critical layers at 1a, 1b, and 1c, and EUV usage will likely expand further at 1d and subsequent nodes. CXMT, by contrast, must attempt to pattern comparable features using DUV immersion combined with increasingly complex self-aligned multipatterning schemes such as SADP and SAQP. This path requires more masks, more deposition and etch passes, more metrology loops, longer cycle time, tighter overlay budgets, more defect opportunities, and higher cost per bit. Long-term, scaling will require either a domestic advanced lithography tool or alternative scaling approaches like 3D DRAM.
The October 2022 BIS package tightened the constraint further by controlling wafer fabrication equipment capable of producing DRAM at or below 18nm half-pitch, directly capturing CXMT's Gen 4 and future Gen 5 node. It also restricted U.S. persons from supporting such activities. This effectively limited CXMT's access to new leading-edge deposition, etch, metrology, and process-support capability from suppliers such as Applied Materials, Lam Research, and TEL. The October 2023 controls added even more pressure by adding node-agnostic tool restrictions, expanding controls on components and parts, tightening technical parameters, and widening the country scope to reduce trans-shipment.
The December 2024 control focused on limiting China's access to foreign HBM by restricting HBM shipment to China. But for CXMT, what is more important is the control on wafer fabrication equipment. The package added 24 types of semiconductor equipment and three software tools, introduced a new semiconductor manufacturing equipment foreign-direct product rule, and added 140 entities, including domestic toolmakers such as NAURA, ACM Research, and Piotech. These measures directly affect the process modules CXMT needs for advanced DRAM and HBM, including DUV multipatterning, TSV formation, wafer thinning, bonding, stacking, inspection, and advanced packaging process control.
In oxidation and diffusion, CXMT is working with domestic equipment leader NAURA, which has oxidation and diffusion equipment in its core product portfolio. In rapid thermal processing, Mattson is likely the domestic supplier given that its parent E-Town has disclosed CXMT as a customer, while on the foreign side the global vertical furnace duopoly of Kokusai and TEL are probable suppliers. Etching is one area where China has made notable improvement. CXMT is confirmed as choosing AMEC as a domestic etch supplier, along with BEST focused on 14nm and below etch, based on public disclosure. On the foreign side, Lam Research, TEL, and Applied Materials very likely remain as the pre-restriction installed base, with possible residual supply of non-controlled tools and service.
On deposition, Piotech and NAURA are CXMT's two main domestic suppliers, with Korea's Jusung Engineering reported to supply ALD. A large pre-restriction Applied Materials CVD/ALD/PVD installed base is likely still a major part of CXMT's production line. Ion implantation most likely runs on the pre-restriction installed base of Axcelis and Applied Materials. Domestic vendors appear to be moving quite slow in this space, mostly stuck at the development and qualification stage. This is probably one of the most difficult wafer fabrication equipment obstacles for domestic vendors to overcome besides lithography.
For TSV and advanced packaging, which are critical for HBM, China's domestic equipment chain is led by NAURA and AMEC. NAURA is the broadest supplier, covering deep-silicon TSV via-etch, barrier/seed PVD, ALD, furnace anneal, cleaning, and its own copper-electroplating tool for TSV fill. AMEC is a major CXMT supplier centered on etch, especially providing the high-aspect-ratio dielectric etch used on CXMT's DRAM array and fine-pitch TSV-last process on the roadmap. Under the broader policy goal of balancing market dynamics, SemiAnalysis believes AMEC will gain more share at CXMT while NAURA will continue to reinforce its leadership in logic and interposers by working with SMSC and high-performance computing customers.
China's Structural HBM Constraint
Looking at China's HBM supply environment, both internally and externally, the situation appears to be worsening. Internally, CXMT's progress in HBM still looks limited. Externally, the supply backdrop is also deteriorating due to existing U.S. restrictions on HBM exports to China, declining foreign HBM inventory, and tightening HBM supply across the three major memory suppliers. SemiAnalysis believes China's HBM constraint will only worsen over the next 18 months unless there is a major technology improvement in domestic HBM or a change in export control rules, which will likely impact the country's pace in deploying compute domestically.
However, HBM re-export and smuggling could complicate this conclusion. SemiAnalysis understands that some Chinese companies continue to obtain HBM3 from memory suppliers, a dynamic the firm reported on last year and that has since been corroborated by other leading media outlets. Based on conversations with industry participants, re-export through foreign offices or partner companies located in third countries remains one pathway through which Chinese companies can access HBM. In addition, some downstream outsourced semiconductor assembly and test providers or intermediaries in third countries appear to facilitate these flows. Some entities may ship partially assembled systems or modules in forms that are not treated as fully manufactured GPUs or ASICs and are therefore still permitted to be shipped into China. The HBM can then be recovered and repackaged onto domestic Chinese GPUs or ASICs.
On commodity DRAM, however, Chinese suppliers are making material progress. CXMT's DDR5 and LPDDR5 products can already be found and are appearing more frequently, especially in consumer products from both domestic and foreign brands, spanning smartphones, wearables, and other end devices. SemiAnalysis has learned that CXMT's memory products are either in qualification, with some already approved, or under consideration by several other major foreign PC and consumer electronics brands. Even so, CXMT's market presence remains largely concentrated in China. Nevertheless, considering the global constraints on memory supply and the capacity allocation shift by leading memory suppliers toward server DRAM and HBM, the firm believes CXMT could further gain share in China's consumer DRAM space.
SemiAnalysis is also observing incremental momentum in the company's server DDR5 products, which have historically been a weaker part of its portfolio. The firm believes one reason CXMT's DDR5 mix has materially increased to over 30%, from the low-20% range previously, is broader tightness in global server DRAM supply, alongside incremental demand growth from domestic hyperscalers such as Alibaba, ByteDance, and Tencent. As the firm noted in an institutional report a month ago, CXMT is currently negotiating 3-plus year DRAM long-term agreements with domestic cloud service providers, as these customers seek to secure as much server DRAM supply as possible beyond volumes available from foreign suppliers.
In the long term, SemiAnalysis believes competitive pressure from CXMT on leading memory suppliers will only intensify, both at the product level and from a capacity perspective. Although CXMT is becoming more competitive gradually rather than rapidly, it continues to gain ground. While the company remains meaningfully behind in HBM, leaving the big three memory suppliers with a comfortable lead, competitive pressure in commodity DRAM is already being felt in the Chinese market and could eventually extend globally. In addition, continued improvement in Chinese wafer fabrication equipment could make equipment less of a gating factor for CXMT's DRAM capabilities over the long run, especially as its process technology matures and improves.