This article's data is primarily sourced from the formal prospectus signed by CXMT on July 22, 2026. The 2025 annual data has been audited, and the Q1 2026 data has been reviewed. The H1 2026 data is estimated by the company, unaudited or unreviewed, and does not constitute a profit forecast or performance commitment. This article is for industrial research and analysis only and does not constitute investment advice.
Compared with the previous draft, the latest prospectus of CXMT has undergone a change significant enough to alter the research conclusion:
The company is no longer an enterprise asking "when it can become profitable," but rather one asking "how long the current ultra-high profitability can be sustained."
In 2025, CXMT achieved an operating revenue of RMB 61.799 billion, a net profit attributable to parent company shareholders of RMB 1.875 billion, and a net profit attributable to parent company shareholders after deducting non-recurring gains and losses of RMB 5.316 billion, officially ending its continuous loss-making status.
Entering 2026, profit elasticity was further released. In Q1 2026, the company's revenue reached RMB 50.800 billion, with a net profit attributable to parent company shareholders of RMB 24.762 billion; the company estimates that the operating revenue for H1 2026 will be between RMB 110 billion and RMB 120 billion, with a net profit attributable to parent company shareholders between RMB 50 billion and RMB 57 billion.
This set of figures is extremely impressive, but also highly cyclical.
It indicates that CXMT has completed the initial commercial closed loop of capacity, products, and customers, and also demonstrates that after DRAM price increases are amplified by the operating leverage of high fixed costs, astonishing profits can be generated in a short period.
Therefore, studying CXMT today can no longer be confined to grand narratives such as "domestic substitution," "the world's fourth," or "benefiting from AI storage."
The real questions that need to be answered are:
How much of the 2026 profits come from long-term competitiveness, and how much from the DRAM price cycle?
Can CXMT leverage this boom window to transform phased profits into cost leadership, technological upgrades, and sustainable free cash flow?
I. IPO Pricing Completed: The Market is Pricing in Cycle-Crossing Expectations, Not "Manufacturing Valuation"
The issue price for this CXMT offering is RMB 8.66 per share, with an initial issuance of approximately 6.688 billion shares, resulting in a total share capital of approximately 66.881 billion shares post-issuance.
Calculated without exercising the over-allotment option, the company's market capitalization post-issuance is approximately RMB 579.2 billion, with total funds raised of approximately RMB 57.919 billion, and net funds raised of approximately RMB 57.638 billion.
The issuance P/E ratio is 308.92 times, and the issuance P/B ratio is 5.06 times.
Such a high static P/E ratio does not mean the market is truly valuing DRAM manufacturers at 300 times profits.
The reason is that the issuance P/E ratio uses the lower net profit attributable to parent company shareholders before and after deducting non-recurring items in 2025, while CXMT was still in the capacity ramp-up and cycle recovery phase for the vast majority of 2025, with profits concentrated in the fourth quarter. In Q4 2025, main business revenue reached RMB 29.551 billion, accounting for 48.23% of the full-year main business revenue.
In Q1 2026, the company's net profit attributable to parent company shareholders has already reached RMB 24.762 billion, far exceeding the full-year 2025 figure.
If the profits of Q1 or H1 2026 are mechanically annualized, CXMT's "forward P/E ratio" would appear very low. However, this valuation method is equally dangerous for strongly cyclical industries:
The higher the profits at the peak of the cycle, the lower the mechanically annualized P/E ratio; and when the P/E ratio is at its lowest, it is often precisely when supply expansion is about to accelerate.
Therefore, CXMT's current valuation is essentially not pricing in 2025 profits, nor can it simply price in H1 2026 profits; rather, it is pricing in three long-term assumptions:
CXMT has entered the cost range of mainstream global DRAM manufacturers;
Memory demand driven by AI can maintain the industry prosperity center in the long term;
The company can transform current profitability into the next round of process and product competitiveness.
If any one of these three assumptions is falsified, the valuation system may change significantly.
II. The Biggest Change: The Company Has Crossed the Break-Even Point, But Has Not Yet Proven the Profit Center
Let's first look at the latest financial data.
| Unit: RMB 100 million | 2023 | 2024 | 2025 | Q1 2026 |
|---|---|---|---|---|
| Operating Revenue | 90.87 | 241.78 | 617.99 | 508.00 |
| Net Profit | -192.25 | -90.51 | 71.44 | 330.12 |
| Net Profit Attributable to Parent | -163.40 | -71.45 | 18.75 | 247.62 |
| Net Profit Attributable to Parent (Excl. Non-recurring) | -167.52 | -78.70 | 53.16 | 263.41 |
| EBITDA | -47.62 | 98.45 | 377.77 | 437.33 |
| Operating Cash Flow | -72.72 | 68.97 | 365.20 | 425.66 |
The Q1 2026 data has been reviewed by Deloitte; the company estimates H1 2026 revenue of RMB 110 billion to RMB 120 billion, and net profit attributable to parent company shareholders of RMB 50 billion to RMB 57 billion.
From the numbers, CXMT's profitability appears to have undergone a sudden leap.
However, from an industrial logic perspective, this did not happen suddenly; rather, three factors crossed the critical point simultaneously:
First, capacity scale and utilization rates have increased, and fixed costs are beginning to be fully amortized.
In DRAM manufacturing costs, manufacturing overheads such as depreciation, power, and equipment maintenance account for an extremely high proportion. Once production crosses the break-even point, for every additional unit of product, a significant portion of the new revenue will be directly converted into profit.
Second, DDR5 and LPDDR5X are rapidly ramping up in volume, leading to product structure upgrades.
In 2025, the average selling price of the DDR series increased by 61.00% year-on-year, and sales volume grew by 282.22% in terms of capacity; the average selling price of the LPDDR series increased by 24.46%, and sales volume grew by 65.18%. Meanwhile, the unit cost of DDR decreased by 26.26%, and the unit cost of LPDDR decreased by 22.85%.
Third, and the most important factor, DRAM prices rose rapidly in the second half of 2025 and Q1 2026.
The company explicitly pointed out in the prospectus that since the second half of 2025, factors such as the growth in AI demand and capacity allocation by major global manufacturers have led to a shortage of DRAM supply and a rapid increase in product prices, which is an important reason for the company's profit growth.
This means CXMT has crossed the break-even point, but the current profit level cannot directly represent the long-term profit center.
A more accurate judgment is:
The inflection point in profitability has been confirmed, but the sustainability of profitability has not yet been confirmed.
III. Gross Margin Rises from Negative to 41%, Operating Leverage Fully Activated
In 2023, CXMT's comprehensive gross margin was -1.93%; in 2024, it increased to 5.58%; and in 2025, it further reached 40.99%.
Excluding the impact of the write-off of inventory decline provisions, the comprehensive gross margin in 2025 was 37.81%, and the gross margin of main business was 37.80%.
Among them, the gross margin of the DDR series reached 41.89%, and the gross margin of the LPDDR series was 37.25%.
This set of data indicates that CXMT's improvement is not solely derived from the reversal or write-off of previously accrued inventory decline provisions.
Even excluding the relevant accounting impacts, the company's main products have already formed a real positive gross margin.
However, in H1 2025, the company's main business gross margin excluding the impact of inventory write-offs was only 4.63%, but it leaped to 37.80% for the full year. In other words, the vast majority of the profit improvement for the full year occurred in the second half of 2025.
This precisely illustrates how strong the operating leverage in the DRAM industry is:
Production line depreciation is basically fixed;
R&D expenses are relatively rigid in the short term;
The unit cost of products decreases as the scale expands;
For every one percentage point increase in the selling price, the contribution to profit may be far higher than the revenue growth rate.
This mechanism can create profit explosions when prices rise, and it will also operate in reverse when prices fall.
In 2023, the gross margin of CXMT's main DRAM products was once -118.69%; in 2025, it increased to 38.75%.
The gross margin swing of over 150 percentage points between the two years is sufficient to illustrate:
CXMT is not only a technology growth company but also a manufacturing enterprise with extremely strong commodity cyclical attributes.
Investors cannot analyze it solely using a growth stock framework, nor can they analyze it solely using the static P/E ratio of traditional manufacturing.
IV. Global Share Rises to 7.67%: CXMT is No Longer a Marginal Participant
The latest prospectus discloses that, calculated based on DRAM sales revenue in Q4 2025, CXMT's global market share has risen to 7.67%.
During the same period, the market shares of Samsung Electronics, SK Hynix, and Micron Technology were 33.96%, 34.48%, and 23.41%, respectively, with the top three still accounting for over 90% combined.
Compared with the previous market share of about 3.97% in Q2 2025, CXMT's share in Q4 showed a significant increase.
The significance of this change is not just about ranking.
DRAM is a typical economies of scale industry. An increase in market share will form a series of positive feedback loops:
Larger wafer output can amortize R&D and depreciation;
Higher procurement scale can improve bargaining power for materials and equipment;
More customer certifications can reduce single-market volatility;
A more complete product portfolio can improve production line scheduling efficiency;
Larger revenue scale can in turn support next-generation technology R&D.
In other words, CXMT is crossing an important scale threshold.
Previously, the market discussed "whether China can mass-produce DRAM"; now, the discussion is already "whether CXMT can become the world's fourth stable, large-scale supplier."
However, 7.67% still cannot be compared with the top three international players.
The top three not only possess larger capacity and customer bases but also master high-profit products such as HBM, GDDR, and Advanced Packaging. CXMT's current growth is mainly concentrated in the DDR and LPDDR fields.
Therefore, the exact meaning of being the world's fourth is:
CXMT has entered the global mainstream DRAM competitive circle but has not yet entered the core position of the global high-end storage profit pool.
V. Technology Assessment: Catch-up in Mainstream Products Succeeded, Product Boundaries Still Need Breakthroughs
CXMT has currently completed the mass production of its fourth-generation process technology platform, while the fifth-generation process platform is still in the R&D stage.
In terms of products, the company's LPDDR5, DDR5, and LPDDR5X achieved mass production in 2023, 2024, and 2025, respectively.
As of 2025, DDR5 has been rapidly ramping up in markets such as servers and personal computers, and LPDDR5/5X has entered the supply chains of brands such as Xiaomi and Transsion. The company's DDR5 modules cover forms such as RDIMM, MRDIMM, CUDIMM, and CSODIMM, with the highest disclosed capacity reaching 128GB and speeds reaching 8000Mbps and above.
CXMT's technological progress can be divided into three levels.
Level 1: Products Have Been "Made"
DDR5, LPDDR5, and LPDDR5X have all achieved commercial mass production, indicating that the company has resolved a series of issues in product design, process integration, yield ramp-up, packaging and testing verification, and customer certification.
This is no longer an experimental technological breakthrough, but an industrialization capability.
Level 2: Products Begin to "Make Money at Scale"
The significant turnaround of gross margins for DDR and LPDDR products in 2025 means that new products are not only achieving shipments but also beginning to form economies of scale and commercial profits.
Level 3: Whether It Can Sustain and Not Fall Behind
The fifth-generation process platform is still in the R&D stage, and products such as LPDDR6 also belong to forward-looking R&D directions.
DRAM technology competition is not a one-time exam. Even if a certain generation of products achieves mass production, the competition for the next generation of process, density, power consumption, capacity, and yield will begin immediately.
Therefore, CXMT's technology has crossed the stage of "whether it can be done" and entered the stage of "whether it can sustain the catch-up and maintain economic viability in every generation."
What truly determines its long-term competitiveness is not the public product names, but several indicators that the prospectus cannot fully display:
Single die size;
Number of effective dies per wafer;
Stable mass production yield rate;
Product power consumption and reliability;
Product certification time;
Unit bit cost at the same capacity.
These are the deepest moats in the DRAM industry.
VI. AI Logic Holds, But "Server DRAM" and "AI Core Storage" Must Be Distinguished
In 2025, CXMT's DRAM revenue in the server field reached RMB 15.970 billion, accounting for 26.51% of DRAM product revenue.
In 2023, the company's server revenue was only RMB 395 million; it grew more than 40 times in two years.
Mobile devices remain the largest application field, with 2025 revenue of RMB 36.381 billion, accounting for 60.40%; personal computer revenue was RMB 2.330 billion, accounting for 3.87%.
The growth in server business has obviously benefited from data center construction, DDR5 upgrades, and AI infrastructure investment.
However, the prospectus also explicitly discloses:
Currently, the proportion of the company's DRAM product revenue applied to AI computing servers remains relatively low.
This statement is crucial.
Revenue in the server field does not equal AI server revenue, and DDR5 demand does not equal HBM demand.
In AI servers, storage can be roughly divided into two value tiers:
One is the HBM next to GPUs or AI accelerators, responsible for providing extremely high bandwidth, which is the most scarce and highest-profit storage product in this round of the AI industry chain.
The other is server system memory, mainly adopting products such as DDR5 RDIMM and MRDIMM, responsible for CPU-side data processing, model loading, and system operation.
CXMT is currently mainly participating in the second type of market.
Its products will benefit from the increase in the number of AI servers and the increase in memory capacity per machine, but it has not yet directly entered the highest value-added segment of HBM.
From a product comparison perspective, Samsung, SK Hynix, and Micron have all deployed DDR, LPDDR, GDDR, and HBM. CXMT's currently disclosed main products are still concentrated in DDR and LPDDR, with a relatively smaller product series.
Therefore, an accurate statement regarding CXMT's AI attributes should be:
CXMT is a beneficiary of DDR5 system memory in AI infrastructure, not a core HBM supplier at this stage.
This does not mean its AI value is unimportant.
As the number of server CPU cores increases, model parameter sizes expand, and inference tasks grow, system memory capacity will also rise. The DDR5 and MRDIMM markets may expand in the long term.
However, in terms of value capture, HBM possesses higher bandwidth, higher technological barriers, stronger customer binding, and higher profit margins. If CXMT is absent from HBM and GDDR in the long term, it may gain sales volume growth driven by AI demand but will find it difficult to obtain the highest profit premium in the AI storage industry.
VII. Capital Closed Loop Significantly Improved, But 2025 Free Cash Flow Remains Negative
In 2025, CXMT's operating cash flow reached RMB 36.520 billion, a significant increase from RMB 6.897 billion in 2024.
However, in the same year, the company paid RMB 49.739 billion in cash for the acquisition and construction of fixed assets, intangible assets, and other long-term assets.
Simply calculating by subtracting capital expenditure from operating cash flow, the free cash flow in 2025 was still approximately RMB -13.2 billion.
This is already a significant improvement over 2024.
In 2024, operating cash flow was only RMB 6.897 billion, while capital expenditure was as high as RMB 71.230 billion, resulting in a funding gap of over RMB 64 billion.
In other words, CXMT is rapidly approaching a capital closed loop:
In 2023, operating cash flow was still negative;
In 2024, operating cash flow turned positive but was far from covering capital expenditure;
In 2025, operating cash flow was already able to cover about 70% of capital expenditure;
In Q1 2026, single-quarter operating cash flow reached RMB 42.566 billion.
If the cash flow improvement in 2026 can be sustained, the company will for the first time possess the ability to support large-scale R&D and technological upgrades with internal operating cash flow.
However, there are still two issues here.
First, 2026 Cash Flow is Highly Dependent on Price Increases
The company explicitly attributes the cash flow growth in Q1 2026 to the rapid increase in DRAM prices and the increase in sales collections.
If prices fall back, cash flow may also decline rapidly.
Second, DRAM Capital Expenditure is Not a One-Time Expense
As of the end of 2025, the book value of the company's fixed assets reached RMB 183.024 billion, of which the book value of machinery and equipment was RMB 169.287 billion; fixed asset depreciation in 2025 reached RMB 24.680 billion.
Even if capacity is no longer significantly expanded, the company still needs to continuously invest in equipment renewal, process transformation, and product migration.
For DRAM enterprises, depreciation cannot be simply regarded as a "non-cash expense" because historical equipment ultimately needs to be replaced by new equipment.
Therefore, to judge whether CXMT's capital closed loop is truly formed, one cannot only look at EBITDA and operating cash flow, but also needs to see:
After crossing a complete cycle, whether operating cash flow can continuously cover maintenance capital expenditure, technological upgrade capital expenditure, and R&D expenses.
The current answer is: it is approaching, but has not yet been validated by a complete cycle.
VIII. Behind the RMB 57.6 Billion Fundraising: Going Public is Not Just Financing, But Also a Balance Sheet Restructuring
The net funds raised by CXMT in this offering are approximately RMB 57.638 billion, while the three proposed fundraising projects plan to use a total of RMB 29.5 billion.
Among them:
| Project | Total Investment | Proposed Use of Raised Funds |
|---|---|---|
| Technical Upgrade and Transformation of Memory Wafer Manufacturing Mass Production Line | RMB 7.5 billion | RMB 7.5 billion |
| DRAM Memory Technology Upgrade | RMB 18.0 billion | RMB 13.0 billion |
| DRAM Forward-looking Technology R&D | RMB 9.0 billion | RMB 9.0 billion |
| Total | RMB 34.5 billion | RMB 29.5 billion |
The fundraising projects mainly involve process transformation and technological upgrades for existing production lines, rather than simply building more homogenized capacity.
The actual net raised funds are about RMB 28.1 billion higher than the scale of funds planned to be invested in the fundraising projects.
This additional capital will significantly enhance the company's balance sheet and anti-cyclical capability, but its ultimate usage efficiency will become an important observation indicator post-listing.
For CXMT, the more sufficient the capital, the more capable it is to:
Maintain R&D investment during industry downturns;
Procure key equipment and materials in advance;
Advance the fifth-generation process platform;
Support customer certification and product iteration;
Avoid being forced to shrink when competitors expand capacity.
However, sufficient capital does not automatically equal excellent capital returns.
If funds continue to be invested in capacity with insufficient returns, or if supply is expanded during the high DRAM price phase, the classic capital cycle of the memory industry may be replayed:
Price increase — Profit improvement — Capital expenditure increase — Capacity release — Oversupply — Price drop.
Therefore, what is most worth paying attention to post-listing is not how much cash CXMT has, but how much cross-cycle profit each additional yuan of capital can ultimately create.
IX. Channel Data Shows Fluctuations: Customer Breakthroughs Are Real, But Distribution Reliance Has Not Disappeared
CXMT has entered the supply chains of top customers such as Alibaba Cloud, ByteDance, Tencent, Lenovo, Xiaomi, Transsion, Honor, OPPO, and vivo.
This means the company's product reliability, performance, and supply capability have been validated by a group of important customers.
However, from the perspective of the sales model, in 2025, the company's distribution revenue still reached RMB 52.316 billion, accounting for 85.38% of main business revenue; direct sales revenue was RMB 8.959 billion, accounting for 14.62%.
Notably, the proportion of direct sales revenue increased to 26.79% in H1 2025, but fell back to 14.62% for the full year.
This indicates that amidst the rapid revenue growth in the second half of the year, a large amount of the increment was still completed through distribution channels.
The distribution model has obvious advantages for rapid expansion:
Fast customer coverage;
Higher efficiency in collections and inventory turnover;
Quick entry into the global electronic component distribution system;
Conducive to digesting rapidly increasing capacity.
However, an excessively high proportion of distribution will also bring uncertainties in research:
Relatively low transparency of terminal demand;
Channel inventory may amplify industry volatility;
Weaker direct binding between the original manufacturer and terminal customers;
It is difficult to judge final consumption and repeat purchases solely based on shipments.
Therefore, the breakthrough in CXMT's customer list is worthy of affirmation, but judging customer quality cannot only look at "entering the supply chain," but also needs to look at:
Direct sales proportion;
Repeat purchases by top customers;
Server platform certification cycle;
Joint R&D of new generation products;
Whether orders have long-term characteristics.
In the DRAM industry, a truly high-quality customer relationship is not a one-time purchase, but the customer's willingness to continue adopting it on the next-generation platform.
X. Between Net Profit Attributable to Parent and Consolidated Profit, Minority Interests Still Need Attention
In 2025, CXMT's consolidated net profit was RMB 7.144 billion, but the net profit attributable to shareholders of the parent company was only RMB 1.875 billion.
As of the end of 2025, the company's owners' equity was approximately RMB 154.100 billion, of which the equity attributable to shareholders of the parent company was approximately RMB 56.754 billion. Calculated accordingly, minority interests were approximately RMB 97.3 billion.
This structure comes from the capital arrangements of important subsidiaries such as ChangXin Xinqiao and ChangXin Jidian.
CXMT directly or through subsidiaries holds 30.68% equity in ChangXin Xinqiao and controls 73.01% of the voting rights through an acting-in-concert agreement; it directly holds 31.72% equity in ChangXin Jidian and controls 75.32% of the voting rights through an acting-in-concert agreement.
This means:
CXMT can consolidate the revenue, assets, and profits of these subsidiaries, but the parent company's shareholders do not own all the economic interests.
Therefore, investors cannot value the company solely based on consolidated revenue, EBITDA, or net profit, but must also pay attention to:
Net profit attributable to the parent company;
Profit distribution to minority shareholders of subsidiaries;
Whether subsequent subsidiary financing continues to dilute economic interests;
The stability of the acting-in-concert agreement;
Whether subsidiary cash can be smoothly remitted to the listed company.
In Q1 2026, the proportion of net profit attributable to the parent company to consolidated net profit has increased to about 75%, a significant improvement over 2025, but it still needs to be observed whether this structure is stable.
XI. Geopolitical Risks Have Shifted from Abstract Risks to Concrete Matters
The new prospectus discloses that on June 8, 2026, the U.S. Department of Defense released a new round of the "Chinese Military Companies List," and CXMT was included in this list.
The company believes that this matter will not have a material adverse impact on daily production and operation activities and going-concern capabilities.
From current disclosures, this list is not equivalent to comprehensive export controls or the Entity List, and cannot be simply deduced as an immediate interruption of equipment, materials, or customers.
However, it indicates that the geopolitical risks faced by CXMT have moved from potential narratives to the actual list management stage.
What truly needs to be tracked in the future is:
Whether further export controls emerge;
Whether maintenance of advanced equipment and components is restricted;
Whether the supply of EDA, materials, and spare parts changes;
Whether procurement policies of overseas customers tighten;
Whether upstream suppliers implement additional compliance reviews.
CXMT's promotion of supply chain diversification and localization is a long-term strategy, but DRAM production involves the coordination of a large amount of high-precision equipment, materials, and processes; complete substitution cannot be achieved in the short term.
For CXMT, supply chain security is not an extra bonus, but a fundamental condition determining continuous production and technological iteration.
XII. How to Understand the "Windfall Profits" in H1 2026
The company estimates that net profit for H1 2026 will be between RMB 66 billion and RMB 75 billion, and net profit attributable to the parent company will be between RMB 50 billion and RMB 57 billion.
This profit scale even approaches or exceeds the normal annual profits of many large global semiconductor enterprises.
Is it sustainable?
The answer is most likely: it cannot be simply extrapolated.
CXMT itself also explicitly pointed out in the risk warnings:
Current DRAM prices are at a high level;
Continuous and substantial price increases are not sustainable;
Samsung, SK Hynix, and Micron are increasing capital expenditure;
New capacity will be gradually released in the coming years;
If AI demand falls short of expectations and supply is concentrated, the industry may enter a downward cycle again;
There is a risk of unsustainability regarding the significant performance growth in H1 2026.
The profit explosion in 2026 is more like a "super operating leverage" formed by the superposition of four factors:
Capacity has been built;
Yield rate and unit costs have improved;
High-value products such as DDR5 are ramping up in volume;
Market prices are rising rapidly.
The first three belong to corporate capabilities, and the fourth belongs to the industry cycle.
The key to CXMT's long-term value is whether the first three can still support reasonable profits and free cash flow after the fourth disappears.
XIII. Three Scenario Deductions
Optimistic Scenario: From Global Fourth to Stable Oligopoly
AI training and inference demand continues to grow, and server DDR5 capacity continues to increase; mobile AI drives the upgrade of LPDDR5X and LPDDR6; CXMT's fifth-generation process platform achieves smooth mass production, and unit bit costs continue to decline.
The company's market share further increases from about 7% to double digits, and achieves clear breakthroughs in GDDR, HBM, or other high-bandwidth products.
Even if DRAM prices fall back from their peaks, the company can still maintain a high capacity utilization rate and positive free cash flow.
In this scenario, CXMT will transform from a domestic substitution enterprise into one of the global DRAM oligopolies, and the valuation logic will shift from policy and scarcity premiums to global competitiveness premiums.
Neutral Scenario: Prominent Strategic Value, But Profits Remain Strongly Cyclical
DRAM prices fall back after 2026, and CXMT's profits are significantly lower than the H1 peak, but relying on product upgrades, cost reductions, and market share increases, it can still maintain cross-cycle profitability.
The company establishes a stable position in the DDR and LPDDR fields, but HBM and high-end products remain dominated by the international big three.
Operating cash flow can cover most capital expenditures, but external financing is still needed in years of large-scale technological upgrades.
In this scenario, CXMT is a high-quality cyclical manufacturing enterprise with strategic scarcity, but should not be valued as an AI platform company.
Pessimistic Scenario: High Prices Stimulate Supply, Capital Cycle Reverses Again
New capacity from international manufacturers is released in a concentrated manner, AI capital expenditure growth slows down, and consumer electronics demand is suppressed by high memory prices.
DRAM prices fall rapidly, while CXMT's new depreciation, R&D, and equipment investments remain rigid.
If the progress of the fifth-generation technology falls short of expectations and high-end products are absent, while encountering further supply chain restrictions, the company's profits may drop significantly from the peak, or even return to losses.
In this scenario, the high profits in 2026 will be proven to be a cyclical peak, not a long-term profit center.
XIV. Eight Most Worth Tracking Indicators in the Future
To judge whether CXMT has truly moved from industrial success to investment success, the following indicators should be continuously observed.
First, DRAM average selling price.
This is the most sensitive variable for short-term profits and the core for judging the cycle position.
Second, unit bit cost.
Only if cost reductions can offset the long-term price decline in the industry can a true manufacturing barrier be formed.
Third, mass production progress of the fifth-generation process platform.
Completion of R&D does not equal economic mass production; yield rate and cost are the ultimate standards.
Fourth, AI computing server revenue, not general server revenue.
Currently, server revenue is growing rapidly, but the proportion of AI server revenue is still low and needs to be verified separately.
Fifth, progress in HBM, GDDR, and high-bandwidth products.
This is the key falsification indicator of whether the company can enter the high-profit pool of AI storage.
Sixth, the coverage ratio of operating cash flow to capital expenditure.
Whether CXMT can generate its own cash ultimately depends on free cash flow, not EBITDA.
Seventh, the proportion of net profit attributable to the parent company to consolidated profit.
This determines how much of the consolidated business growth truly belongs to the shareholders of the listed company.
Eighth, synchronized growth of market share and direct sales customers.
Relying solely on distributors to expand volume is less valuable than directly entering the next-generation platforms of top customers.
Conclusion: CXMT Has Proven It "Can Make, Can Sell, and Can Profit"; The Next Step is to Prove It "Can Earn Sustainably"
CXMT is one of the extremely few enterprises in China's semiconductor industry that has completed the following complete chain:
Independent technology development — Wafer fab construction — Scale mass production — Mainstream customer introduction — Global market competition — Profit realization.
This industrial achievement itself is unquestionable.
The latest prospectus further proves that CXMT's business model has passed the most dangerous initial stage. It is no longer just a strategic project relying on capital investment and long-term losses, but a large-scale DRAM manufacturer capable of generating tens of billions of yuan in operating cash flow and profits during industry boom periods.
But for investors, the new questions are actually harder.
In the past, the judgment was: Can CXMT become profitable?
Now, the judgment is:
How much of the current profits belong to the cycle;
How much profit will remain after industry prices return;
Whether it can enter high-value products such as HBM;
Whether it can use current cash flow to complete the next-generation technological upgrade;
Whether it can maintain positive free cash flow in the next downward cycle.
Therefore, the most accurate conclusion about CXMT is not simply being bullish, nor being bearish because of the cycle peak.
CXMT has completed the leap from "whether domestic DRAM can be made" to "whether the world's fourth can make money." What truly determines its long-term value next is whether it can transform the cyclical profits of 2026 into the next round of technological, cost, and product advantages.
If this transformation is completed, 2026 may not be the profit peak, but the starting point for the formation of global competitiveness.
If it is not completed, the high profits seen today may just be another classic cyclical peak in the memory industry.