Changxin Technology's listing: it made more than four times the total profit of last year in just three months, but how much of that will belong to the shareholders?
The performance of Changxin Technology is indeed impressive, strong enough to be compared alongside SK Hynix, Micron, and Samsung in the same performance table. Coupled with the DRAM price surge and the market hype around domestic-made storage, it is easily labeled as "China's SK Hynix." This hype has overshadowed many discussions that should have taken place, including its fundamental business strengths and what kind of investment framework is truly suitable for this company.
What has also been overshadowed is the almost unforgiving brutality of the storage industry. As the inventor of commercial memory, Intel was forced to exit the DRAM market in 1985; in 2012, Elpida, which had received Japanese policy-based equity funding and bank loans, still filed for bankruptcy, and today's SK Hynix has also gone through debt reduction, maturity extension, interest rate cuts, and debt-to-equity swaps. These companies did not collapse at the same point in the market cycle, nor did they fall into trouble for the same mistakes, but they all share one common outcome:
Technology, scale, and policy-supported funding—none of these can guarantee a storage manufacturer a permanent seat at the table.
Therefore, at the current juncture, Changxin, which has not yet caught up with SK Hynix, Micron, and Samsung, remains the most vulnerable player at the table. At least in the first quarter of 2026, Changxin's operating profit margin was comparable to that of SK Hynix; both companies benefited from the price increase of traditional DRAM, but SK Hynix already has scaled-up HBM production, high-capacity server products, and stronger customer lock-in, while Changxin still relies heavily on traditional DRAM, with limited direct revenue from AI-related businesses.
Once prices fall, Changxin will face both shrinking profits and the financial pressure brought by continuous technological investment and capacity expansion.
However, compared with those industry predecessors that exited the market, went bankrupt, or were once on the verge of being eliminated, Changxin has a fundamental difference: it is the only domestic enterprise that has achieved large-scale mass production of DRAM, and as such, it undertakes the historical mission of driving the growth of the local upstream semiconductor industry through mass production demand. Success in this industry has always relied on the iterative co-development of customers, equipment, materials, and manufacturing processes. Amid severe external restrictions on China's semiconductor industry, equipment and material manufacturers are in urgent need of a large-scale customer that can place continuous orders and provide mass-production validation opportunities.
Changxin's large-scale production lines provide upstream suppliers with orders, production data, and opportunities for continued R&D; once the upstream supply chain matures, Changxin will have a solid foundation to reduce costs and narrow the technological gap. This interdependence will not eliminate market cycles, but it will determine how quickly Changxin can catch up. Therefore, when evaluating Changxin, the first thing to look at is what kind of profits it is generating, and whether it can survive the next market downturn.
Making profits does not equal being fully integrated into the AI industry
Changxin's growth curve is impressive, and its operational turnaround began before the 2026 DRAM price surge. From 2023 to 2025, the company's revenue increased from 9.087 billion yuan to 61.799 billion yuan, with net profit attributable to shareholders turning around from a loss of 16.34 billion yuan to a profit of 1.875 billion yuan; EBITDA and operating cash flow turned positive in 2024, and operating cash flow further increased to 36.52 billion yuan in 2025.
In the first quarter of 2026, the company generated 50.8 billion yuan in revenue in three months, equivalent to 80% of its full-year 2025 revenue; operating profit reached 35.433 billion yuan, more than four times the full-year 2025 level, with net profit attributable to shareholders hitting 24.762 billion yuan and operating cash flow reaching 42.566 billion yuan, pushing its operating profit margin close to 70%; over the same period, SK Hynix recorded an operating profit margin of approximately 72%.
In its prospectus, Changxin attributes this explosive growth to computing power demand, the DRAM supply shortage caused by major manufacturers adjusting their production capacity, and the improvement in its own production and sales scale and product structure. The first two factors created a price hike dividend shared across the entire industry, while the last factor came from Changxin's own operational and technological progress.
In 2025, DDR sales volume increased by approximately 282%, unit cost decreased by 26.26%, DDR5 shipments started to ramp up, and LPDDR5X entered mass production. The improvements in sales volume, cost, and product generation have enhanced Changxin's ability to withstand the next round of price declines; however, LPDDR and DDR still contributed 98.3% of its main operating revenue that year, mobile devices accounted for 60% of revenue, and revenue from AI computing power servers remained low, meaning the AI dividend indirectly boosted the company's performance mainly through DRAM price increases.
However, large-scale shipments of high-end AI storage products will not automatically trigger a switch in the company's valuation framework. Even for SK Hynix and Micron, the market is still vacillating between cyclical stock valuation frameworks and growth stock valuation frameworks. JPMorgan Chase advocates shifting from price-to-book ratio valuation to price-to-earnings ratio valuation, some analysts propose valuing traditional storage and HBM separately, while other institutions still stick to the price-to-book ratio method.
The root of this controversy lies in the fact that HBM boosts profits through two channels: on one hand, it directly generates AI-related revenue, and on the other hand, it consumes a large amount of wafer production capacity, indirectly tightening the supply of traditional DRAM and becoming one of the key drivers behind DRAM price increases. Micron estimates that under the same manufacturing process and for the same number of bits produced, HBM3E consumes approximately three times the wafer capacity of DDR5, and HBM4 will consume even more.
HBM does not end the storage industry cycle; it reshapes it: the growth profits from AI and the cyclical profits from traditional DRAM price hikes are now inextricably linked.
SK Hynix's 2025 HBM revenue increased by more than 100% year-on-year; in the first quarter of 2026, DRAM shipments were roughly flat quarter-on-quarter, but the average selling price rose by about 65%, which the company attributed mainly to the accelerated price increase of traditional DRAM.
Micron's financial report for the period from February 27 to May 28, 2026 (FY2026Q3) shows that the company's HBM4 revenue has exceeded 1 billion US dollars; in this quarter, DRAM still accounted for 76% of total revenue, shipments only grew by a low single-digit percentage, and the average selling price rose quarter-on-quarter in the "low 60% range". The adjusted gross margin increased from 74.9% to 84.9%, and the management clearly stated that the improvement mainly came from price increases.
The market is currently pricing in the possibility that HBM could help leading manufacturers survive the market cycle, rather than confirming that this outcome is guaranteed. After traditional DRAM prices fall, the first real test will come for what valuation framework should be used to value storage companies.
Even if the market eventually confirms that SK Hynix is eligible for a growth stock valuation framework, Changxin will hardly be able to directly apply that framework before it achieves substantial shipments of high-end storage products. The company has publicly announced its R&D progress in CXL, in-memory computing, near-memory computing, and new DRAM architectures, and there are media reports about its HBM samples being submitted for testing and customized product development; domestically, alternative technical routes such as super-node architecture, shared memory pools, and customized high-bandwidth memory have also emerged. Changxin does not necessarily have to replicate SK Hynix's path, but before customer validation and mass production are achieved, all these initiatives can only be regarded as option value.
Changxin cannot currently adopt a growth stock valuation framework, but this does not mean its valuation premium is completely unfounded. The differences between Changxin and SK Hynix, Micron are not just about whether high-end products have been shipped, but also about whether the industrial and capital conditions can help it survive the next market cycle, and how much of the value created by these conditions will ultimately belong to the shareholders of the listed company.
Scale itself is a core industrial capability
At this stage, Changxin's importance goes far beyond being just a DRAM manufacturer. For local semiconductor equipment and material manufacturers, it is also one of the few large-scale customers that can place continuous orders, provide production data, and offer mass-production validation opportunities.
Individual niche semiconductor markets used to be far smaller than they are today, while the costs of technology R&D and factory construction have continued to rise. The industry has evolved from an integrated model where design, manufacturing, and process development were all done in-house by a single company, to a split structure covering fabless design, foundry manufacturing, equipment, materials, and packaging and testing, with collaborative relationships now spanning different enterprises. The finer the division of labor, the more closely linked each link becomes, because no player can afford the cost of betting on a wrong generation of technology alone.
This mutual interdependence is not a one-way "support" from downstream to upstream. Large-scale customers use long-term orders, R&D funding, and real production lines to reduce the commercialization risks for suppliers; when upstream suppliers make breakthroughs, they in turn help improve the downstream customers' yield rates, reduce costs, and support the development of next-generation products. Industry leaders rely on this cycle to maintain their competitive edge, while latecomers use it to narrow the gap.
ASML's EUV lithography machine is a prime example. In 2012, Intel, TSMC, and Samsung committed to providing 1.38 billion euros in R&D funding to ASML over five years, with Intel also making advance purchase commitments. These customers did not wait for the equipment to mature before placing orders; instead, they shared risks through capital investment, orders, and production line collaboration. After years of joint development by upstream and downstream partners, EUV lithography technology finally entered commercial mass production in 2019.
In the DRAM industry, which features high standardization and strong price cyclicality, this collaboration ultimately translates into improvements in storage density, yield rates, and unit costs. Samsung uses five layers of EUV lithography to mass-produce 14-nanometer DRAM, increasing wafer productivity by about 20% compared to the previous generation; SK Hynix uses EUV technology for its 1a-nanometer DRAM, which is expected to increase chip output per wafer of the same size by 25%. These benefits cannot be entirely attributed to the lithography machine alone, but they will significantly improve the cost position of manufacturers when the next market downturn arrives.
For Changxin, which is still playing catch-up and operating in an environment with restricted access to foreign equipment and technology, this collaborative ecosystem is even more critical. The prospectus shows that the company still uses multi-patterning exposure technology for its third-generation process, and its fifth-generation platform will continue to optimize multi-patterning processes. Changxin cannot make up for all the shortcomings in equipment and materials on its own, but without such a mass-production customer, upstream suppliers would find it extremely difficult to get their products into DRAM production lines for repeated validation and refinement.
Changxin now has the ability to provide a full-scale mass production scenario. Construction of its first phase in Hefei began in 2017, with DDR4 mass production achieved in 2019; the Beijing project was launched in 2020, and the Beijing pilot production line went into operation while the second phase of the Hefei project was completed in 2022. Currently, the company operates three 12-inch DRAM wafer fabs in Hefei and Beijing, and it states that all these facilities will reach full production capacity in 2026.
Evidence of this collaborative ecosystem can be found in the order books of upstream companies. For one of Tuojing Technology's equipment products, after validation and sales at multiple customers including Changxin, its total shipments increased from 41 units to 86 units, with related revenue rising from about 149 million yuan to 321 million yuan. The on-site gas supply contracts between Guanggang Gas and the second phase of Changxin's Hefei project and the Beijing project are expected to remain in effect until 2038 and 2039 respectively.
While the market boom continues, Changxin is still expanding its production scale. According to a Reuters report citing three people familiar with the matter (the company has not commented on the report), Changxin is building two new fabs in Shanghai and Hefei, and is negotiating with other local governments for a third new fab; if all these projects are completed, its monthly production capacity could increase from about 300,000 wafers to more than 600,000 wafers. Amid the tight supply in the storage market, Samsung, SK Hynix, and Micron are also increasing their capital expenditures; but while these industry leaders already have mature HBM products, high-capacity server offerings, and a more optimized cost curve, Changxin still needs to expand its production capacity while continuing to narrow its technological gap.
No one knows how long this market boom will last, but turning planned capacity into actual production requires substantial real capital investment. At the end of 2025, Changxin's total fixed assets and construction in progress amounted to 214.814 billion yuan, accounting for approximately 63.8% of its total assets; from 2023 to 2025, it spent approximately 164.627 billion yuan in cash on acquiring long-term assets. The net proceeds from this IPO, totaling about 57.638 billion yuan, are only sufficient to fund its subsequent investment needs.
Changxin cannot afford this pace of capacity expansion on its own. From 2023 to 2025, its subsidiaries absorbed 77.909 billion yuan in minority shareholder investments, accounting for 81.4% of all cash raised through equity financing. The listed entity controls the Xinqiao and Beijing factories through concerted action arrangements, but it only owns about 30% of their economic equity; at the end of 2025, minority shareholder equity accounted for 63.17% of the consolidated equity.
In another segment that urgently needs technological breakthroughs (wafer manufacturing), SMIC adopted a similar ownership structure and has blazed a relatively smooth capital exit path. At the end of the first quarter of 2026, non-controlling interests accounted for 39.53% of SMIC's consolidated equity. In June 2026, SMIC completed the acquisition of the remaining 49% equity stake in SMIC Northern, issuing 547 million consideration shares that converted external capital from factory shareholders into shareholders of the listed company, transforming the risk-sharing model during the construction period into a profit-sharing model for mature assets.
The case of SMIC Northern shows that minority shareholder equity does not have to remain in subsidiaries permanently. After the factory matures, the listed company can pay consideration to acquire this equity stake, so that more profits can be attributable to the parent company's shareholders; but before Changxin launches a similar consolidation process, its valuation should still be calculated based on the current economic equity structure, and consolidated profits cannot be pre-allocated to the shareholders of the listed company.
Conclusion
On July 27, Changxin Technology went public. Based on its basic offering size, the company raised 57.919 billion yuan, making it the largest A-share IPO since 2022. An offering of this scale, coinciding exactly with the peak of Changxin's performance and the market sentiment for the storage industry, is itself a major event in the capital market; coincidentally, in the same week, leading storage manufacturers and major payers for AI infrastructure will also release new pricing clues one after another.
On July 29, SK Hynix released its second-quarter performance report; on July 30, Samsung announced its full second-quarter financial results. Microsoft and Meta released their financial reports on the 29th, followed by Amazon on the 30th. SK Hynix and Samsung will reveal how much of this market boom has been realized, how much DRAM prices have risen, how well HBM sales are performing, and their next capacity expansion plans; Microsoft, Meta, and Amazon will provide insights into future demand expectations, showing whether AI capital expenditures will continue to grow or start to slow down.
The first week of trading after listing is destined to be lively, with significant price volatility. But looking back a few quarters later, what truly matters are three other things: how much profit Changxin retains, whether its high-end products have achieved shipments, and what percentage of the profits generated by new factories will belong to the listed company.
Disclaimer: This article is for learning and communication purposes only and does not constitute investment advice.
This article is from the WeChat official account "Tangping Index