CXMT Corporation (688825) Equity Research Report - Domestic DRAM Scarcity Is Real, but a RMB3.37 Trillion Valuation Has Already Priced In Peak-Cycle Earnings

Disclaimer: This report is compiled and analyzed from publicly available information and is intended solely for informational and research purposes. It does not constitute investment advice, a securities recommendation, a trading instruction, or any guarantee of returns. The company's operations, valuation, market price, and consensus expectations discussed herein may change over time. Readers should independently verify the data and make their own decisions.

1. Core Conclusion: An Earnings Surge and Valuation Overreach Are Happening at the Same Time

1. CXMT is the clear domestic DRAM winner, but it has not yet won a place in the industry's highest-end profit pool. The company operates three 12-inch DRAM wafer fabs in Hefei and Beijing. Its global revenue share reached 7.67% in 4Q25, ranking first in China and fourth worldwide. DDR5 and LPDDR5/5X are in mass production, while server revenue increased from ￥0.395 billion in 2023 to ￥15.970 billion in 2025. Localization, leapfrog product development, and capacity ramp-up have created genuine competitiveness. However, the prospectus discloses no HBM product line, and AI-compute server revenue remains relatively low. CXMT currently benefits more from AI products crowding out conventional DRAM supply than from directly participating in HBM's highest-margin profit pool.
2. The 2026 earnings surge is being driven jointly by price increases, volume growth, and mix upgrades, with cyclical pricing making the largest contribution. In 2025, DDR average selling prices rose 61.00%, unit volume increased 282.22%, and revenue grew 515.36%; LPDDR average selling prices rose 24.46%, unit volume increased 65.18%, and revenue grew 105.59%. This transmission chain continued as follows: higher-generation products and tight supply led 2025 consolidated gross margin to rise from 5.58% to 40.99%; 1Q26 revenue then increased 719.13% year over year to ￥50.800 billion; and attributable net profit swung from a ￥1.559 billion loss to a ￥24.762 billion profit. Management estimates 1H26 revenue of ￥110-120 billion and attributable net profit of ￥50-57 billion. The earnings trajectory is exceptionally steep, but the prospectus also explicitly warns that continued sharp price increases from already elevated levels are unsustainable.
3. Earnings have turned positive, but cash returns have not yet overcome the capital intensity of the business. Operating cash flow was ￥36.520 billion in 2025, up sharply from ￥6.897 billion in 2024. Purchases of long-term assets totaled ￥49.739 billion over the same period, leaving simplified free cash flow, defined as operating cash flow less long-term asset purchases, at negative ￥13.219 billion. At year-end 2025, cash and cash equivalents were ￥51.990 billion, while short-term borrowings, current maturities of long-term borrowings, and long-term borrowings totaled approximately ￥143.900 billion. Net IPO proceeds of ￥57.638 billion substantially improved liquidity, but did not change the fundamental economics of DRAM: sustained R&D and capacity investment, together with heavy depreciation.
4. The valuation treats peak-cycle earnings as though they were relatively stable. ￥50.35 is 481.4% above the ￥8.66 IPO price and implies a market capitalization of approximately ￥3.37 trillion. Even mechanically annualizing the ￥50-57 billion of estimated first-half attributable net profit produces a P/E of roughly 29.5-33.7x; mechanically annualizing first-half revenue implies a P/S of approximately 14.0-15.3x. The sharper comparison is that SK Hynix's market capitalization was approximately ￥5.94 trillion as of July 13, 2026, according to the listing announcement. CXMT is already worth 56.7% as much, even though its 7.67% share is only about 22% of SK Hynix's 34.48% share. The current price embeds not only domestic substitution, but also a prolonged period of unusually high net margins and continued narrowing of the technology gap.
5. The negative view has clear conditions for reversal. The current valuation could be absorbed through earnings delivery if 2026 full-year attributable net profit reaches at least ￥120 billion, gross margin before inventory write-down reversals remains above 40% after DRAM pricing normalizes, global share exceeds 10%, and CXMT formally discloses that HBM or another next-generation high-bandwidth product has entered volume production for a leading customer. Conversely, if gross margin falls below 30%, server revenue mix stops rising, free cash flow remains negative, or the valuation still exceeds 30x annualized peak earnings around the early-2027 lock-up expirations, the ￥3.37 trillion market capitalization would look more like a scarcity premium amplified by the limited free float than sustainable value.
    

2. Company Overview, Business Mix, and Core Operating Metrics: Mobile Memory Provides the Base While Server DDR5 Reshapes the Profit Pool

Core Operating Metrics and Changes

3. Fundamental Quality: Accounting Earnings Have Surged, but Cash Returns Still Fall Short of Heavy Capital Requirements

4. Industry and Competitive Landscape: CXMT Wins in Domestic Substitution, While SK Hynix Captures AI's Richest Profit Pool

5. Principal Risks

1. A reversal in DRAM pricing and the supply-demand cycle. DDR and LPDDR ASPs rose 61.00% and 24.46%, respectively, in 2025, while 1H26 earnings assume continued rapid price increases. If the international top three expand capacity, conventional DRAM supply recovers after HBM capacity comes online, or cloud-provider capex slows, falling prices will first compress gross margin and then magnify the impact on net profit and valuation through inventory write-downs, capacity utilization, and fixed costs.
2. The gap in HBM and premium AI products. CXMT's disclosed core products are DDR, LPDDR, and modules, while AI-compute server revenue remains relatively low. If HBM continues to capture the incremental profit in server memory and CXMT cannot achieve volume production for leading accelerator and cloud customers, the company may win standard DRAM share while missing the highest-margin profit pool, requiring a reduction in the current AI premium.
3. International trade restrictions and supply-chain instability. The U.S. Department of Defense added subsidiary ChangXin Memory Technologies to a relevant list on June 8, 2026; the company says this does not have a material adverse impact on daily operations. If future restrictions expand to advanced equipment, EDA software, materials, components, or customer cooperation, process iteration and capacity expansion could slow, simultaneously impairing unit costs, yields, and delivery capability.
4. Heavy capital requirements, depreciation, and asset impairment. Fixed assets were ￥183.024 billion, depreciation ￥24.680 billion, and purchases of long-term assets ￥49.739 billion in 2025. If new lines ramp more slowly than expected, the technology roadmap changes, or product prices fall below cost, substantial depreciation, inventory write-downs, and fixed-asset impairments could again cause operating cash flow and accounting earnings to diverge.
5. Customer, distributor, and inventory concentration. The top five customers generated 68.08% of principal-business revenue, distribution represented 85.38% of sales, and year-end 2025 inventory was ￥29.390 billion. If major distributors destock, end customers change suppliers, or payment terms and channel financing shift, shipment data may lag underlying demand while revenue, cash collections, and inventory valuation come under pressure simultaneously.
6. Limited free float, lock-up expirations, and valuation volatility. Unrestricted shares represented only 6.73% of total shares at listing, amplifying price sensitivity through limited supply. Approximately 1.521 billion offline-placement shares become eligible for release six months after listing, equal to 33.8% of the initial float; strategic-placement and pre-IPO shares have 12-36 month lock-ups. If earnings expectations fall before tradable supply increases, the current ￥3.37 trillion market capitalization may be digested through a lower share price rather than earnings growth.
    

6. Monitoring Checklist

• Formal 1H26 report: Revenue should be at least ￥110 billion and attributable net profit at least ￥50 billion. Results below management's estimated range would directly weaken the credibility of the current peak-earnings denominator.
• Gross margin before inventory write-down reversals: Remaining above 35% after pricing normalizes would demonstrate a cost advantage derived from yield, scale, and product mix; falling below 30% would reinforce the view that the cycle has been overcapitalized.
• DDR5 and server revenue mix: An increase in the server-application revenue share from 26.51% to more than 35%, provided it is not driven solely by pricing, would strengthen the product-upgrade thesis. Stagnation or decline would indicate that premium-product penetration is stalling.
• HBM or another next-generation high-bandwidth product: Only formal disclosure of sampling, qualification, and volume-production revenue from leading customers should count as entry into AI's richest profit pool. R&D disclosure or server DDR5 growth alone is insufficient to change the valuation view.
• Capacity, production-to-sales ratio, and inventory: Capacity utilization above 90%, a production-to-sales ratio recovering above 95%, and no material increase in inventory write-downs would show that demand can absorb expansion. Utilization below 85% would be a warning of oversupply.
• Operating cash flow and capital expenditures: Positive full-year simplified FCF would be the first evidence that CXMT can partially self-fund expansion. If FCF remains negative after the IPO, the assessment of capital returns should be reduced.
• Global share and competitor capex: Share above 10% would strengthen CXMT's position as the global No. 4. Investors should simultaneously track Samsung, SK Hynix, and Micron's allocation of capacity between conventional DRAM and HBM; synchronized expansion by all three would weaken the pricing outlook.
• Lock-up expirations and free float: Focus on approximately 1.521 billion offline-placement shares around January 27, 2027, followed by strategic-placement expirations. If the valuation still depends on more than 30x annualized peak earnings as tradable supply increases, share-price volatility risk will rise.
   

7. Sources

• CXMT Initial Public Offering and STAR Market Listing Prospectus (2026-07-09) (https://static.cninfo.com.cn/finalpage/2026-07-09/1225416303.PDF)
• CXMT Initial Public Offering STAR Market Listing Announcement (2026-07-24) (https://static.cninfo.com.cn/finalpage/2026-07-24/1225439534.PDF)
• Shanghai Stock Exchange: Announcement on the STAR Market Listing and Trading of CXMT ￥ Ordinary Shares (2026-07-24) (https://www.sse.com.cn/disclosure/announcement/listing/ipo/c/c_20260724_10826610.shtml)
• Eastmoney: CXMT Real-Time Quote (Snapshot at 11:06 on 2026-08-10) (https://quote.eastmoney.com/kcb/688825.html)
• Eastmoney: CXMT Daily Price History Since Listing (Through 2026-08-10) (https://quote.eastmoney.com/unify/cr/1.688825)