China’s DRAM disruptor? China A-shares’ new “stock king” — ChangXin Technology’s IPO debut saw turnover surpass 1 trillion, how far can the DRAM super cycle go?

On July 27, ChangXin Technology (688825.SH) officially listed on the Shanghai Stock Exchange’s STAR Market. This new stock, priced at 8.66 yuan, opened at 49.50 yuan (+471.6%). Its market cap instantly surged past 3.31 trillion yuan, overtaking Industrial and Commercial Bank of China to become #1 by market cap among A-shares. But this is only the opening act of today’s three-act drama.

1. First act (09:25-11:00): wild celebration. After the open, capital poured in. At 9:49, the trading value surpassed 33.1k yuan, setting a new record for a single day’s trading value for a newly listed A-share. At 10:04, it broke 74.5B yuan, surpassing the previous A-share single-day record set by East Money. Within the first hour of listing, the trading value exceeded 1 trillion yuan—making it the first individual stock in A-share history to see a single-day trading value over 1 trillion. Around 11:00 in the morning, the share price touched the day’s high of 55.03 yuan, with a gain of +535.2%. Market cap was about 3.68 trillion yuan. Globally, it jumped to #24 on the market-cap ranking, exceeding Intel ($46.40 billion) and Tencent (33.9 billion yuan).

2. Second act (11:00-14:00): deep squats. After a sharp rise, selling pressure surged. The share price fell steadily from 55.03 yuan; during the session, the low reached 38.11 yuan—down 30.8% from the intraday high. In the latter half of the lunch break period, panic selling continued to push trading higher. Turnover rate hit 61.5% at 14:05, and trading value surpassed 1.3 trillion yuan. Investors who chased at around 55 yuan logged paper losses of nearly 31% within three hours.

3. Third act (14:00-15:00): convergence. Funds flowed back near the close, and the share price rebounded from the 38-yuan range to around 49 yuan. It ultimately closed at 49.00 yuan, up +465.8%, slightly below the opening price of 49.50 yuan—this is one of today’s most important signals: the first-day close was below the open, meaning late-chasers have already generated unrealized losses.

| Metric | Value | | --- | --- | | Offering price | 8.66 yuan | | Opening price | 49.50 yuan (+471.6%) | | High price | 55.03 yuan (+535.2%) | | Low price | 38.11 yuan (+340.0%) | | Closing price | 49.00 yuan (+465.8%) | | Intraday range | 44.40% | | Full-day trading value | 100B yuan (A-share historical record) | | Turnover rate | about 61.5% | | Closing market cap | 3.28 trillion yuan (#1 among A-shares) | | Profit for winning 1 lot (based on close) | about 20,170 yuan | | Total shares outstanding after issuance | about 36.8k shares |

ChangXin Technology—data from its first trading day

According to the prospectus, this IPO issued 33.9k shares (before exercising the over-allotment option). The total offering market value was about 130B yuan. Before the over-allotment option is exercised, the total fund-raise amount was 141.19B yuan; after full exercise, it was 32.8k yuan. This scale exceeds the 532.30 billion yuan record set by Semiconductor Manufacturing International Corporation in mid-2020, making it the largest IPO in the history of the STAR Market.

Company profile: a decade from zero to the world’s fourth

A leapfrogging R&D strategy

ChangXin Technology was founded in 2016 and is headquartered in Hefei, Anhui. The company adopted a leapfrogging R&D strategy—no low-end products, no outdated process nodes. It directly launched its own-designed and manufactured 8Gb DDR4 chips in 2019, achieving a breakthrough for China’s DRAM industry from zero to one.

According to the prospectus, the company has built a full product portfolio spanning DDR4, LPDDR4X through DDR5 and LPDDR5/5X. The LPDDR5X launched in 2025 reaches a top speed of 10,667 Mbps. The first domestic DDR5 speed reaches 8,000 Mbps, with a per-chip capacity of 24Gb. The technological gap between ChangXin Technology and the world’s leading DRAM memory players is gradually narrowing.

Since its founding in 2016, ChangXin Technology has broken through key core DRAM technologies and successfully realized independent R&D, design, and commercialization leading to mass production—filling the gap in mainland China’s DRAM products in the global market.

Product matrix and customer ecosystem

| Product line | 2025 revenue (hundred million yuan) | Revenue share | Application scenarios | | --- | --- | --- | --- | | LPDDR series | 40.704 | 65.86% | Phones/tablets/mobile devices | | DDR series | 19.531 | 31.60% | Servers/PCs | | Other (wafers/modules/services) | 1.565 | 2.54% | — | | HBM | 0 (in R&D) | — | AI server GPUs |

Terminal customers include Alibaba Cloud, ByteDance, Tencent, Lenovo, Xiaomi, TRANSSION, HONOR, OPPO, vivo, and other mainstream domestic technology companies. Tencent has signed long-term DRAM supply agreements of more than 20 billion yuan. The adoption rate of LPDDR products in domestic Android-brand smartphones has already exceeded 30%. In addition, industry reports say Apple is testing ChangXin Technology’s DRAM chips—while not officially confirmed, it indicates the quality of its products has begun attracting attention from top global customers.

Capacity scale

The company has laid out three 12-inch DRAM wafer fabs across two locations: Hefei and Beijing. By the end of 2025, total DRAM combined monthly production capacity was about 2.8 hundred thousand wafers (12-inch equivalent). Of the funds raised in this IPO, 7.5 billion yuan will be used for wafer manufacturing technology upgrades and renovations, 18 billion yuan for DRAM technology upgrades, and 9 billion yuan for forward-looking technology R&D. In addition, more than 6,000 employees received shares through an employee shareholding plan. Founder Zhu Yiming pledged not to reduce his holdings within 10 years after listing. As of December 31, 2025, cumulative unrecouped losses totaled 36.65 billion yuan, but as profitability is expected to improve sharply in 2026, this historical burden is expected to be digested quickly.

DRAM: the most violent cyclical track in global semiconductors

To understand ChangXin Technology’s valuation logic, you must first understand what DRAM really is—the strongest cyclical track in the global semiconductor industry. Over the past more than half a century, the DRAM industry has fully gone through nine cycles. Each cycle starts with the narrative “this time is different,” only to end with price crashes and players liquidating positions.

| Cycle | Time | Driving force | Price behavior | Players eliminated | | --- | --- | --- | --- | --- | | First | 1993-1998 | Windows PC adoption | Down 51% (1996) +65% (1997) | Japanese makers exit one after another | | Second | 1999-2003 | Internet bubble | Down 95% | Japanese DRAM consolidated into Elpida | | Third | 2004-2009 | Laptops + Vista | Down 87% | Qimonda bankruptcy (Europe exit) | | Fourth | 2010-2015 | Smartphones | Down 67% | Elpida bankruptcy (Japan exit) | | Fifth | 2016-2019 | Cloud computing | Down 79% | Industry solidifies into a triopoly | | Sixth | 2020-2023 | Work-from-home | Samsung/Hynix/Micron collectively incur losses | None (triopoly remains stable) | | Seventh | 2024-present | AI supercycle | DDR5 rises from $20 to $118 | — |

After 2009, the global DRAM industry固ified into a triopoly dominated by Samsung (about 39%), SK hynix (about 29%), and Micron (about 22%), accounting for more than 90% of the global market. The three giants can maintain supply discipline through capacity management—something impossible in the past era of “six-way melee.” This is the biggest difference between this AI supercycle and earlier cycles, and the most core argument for the bulls.

Historically, every DRAM supercycle ends in a crash. 1995, 2000, 2008, 2018—without exception. Among all technology industries, semiconductors are the most brutally cyclical track.

How AI changes the DRAM supply structure

The three giants抢 the meat of HBM, leaving only DDR soup

HBM (high bandwidth memory) is a core component for AI servers. The wafer area consumed in its production is 3 to 4 times that of standard DDR5. In 2026, global DRAM total capital expenditures are projected to rise 65% year over year, reaching nearly $120 billion. However, 60% of新增 advanced wafer capacity will prioritize HBM. The three major original manufacturers have allocated 70% of their新增 production lines toward HBM, yet the capacity gap is still as high as 50% to 60%.

The three giants’ 2026 HBM capacity has already been locked in by top-tier customers such as NVIDIA through long-term agreements of 3 to 5 years. Goldman Sachs estimates that between 2026 and 2030, HBM will occupy about 30% of the three original manufacturers’ available capacity. This creates a rare window for ChangXin: when the leaders focus entirely on HBM, a structural supply shortfall emerges in mid-tier general DRAM.

HBM market size

The HBM market is expanding at an exponential pace. According to Goldman Sachs, the HBM market size will grow from $56 billion in 2026 to $116 billion in 2027, reaching $168 billion in 2028. Bank of America is even more aggressive—HBM sales would grow from $6.46 billion in 2025 to $26.94 billion in 2026 (up +317% year over year), and break $62.7 billion in 2028. Gartner expects HBM revenue to grow at a compound annual growth rate of 60.5%.

How much share can ChangXin take?

Based on Omdia data, ChangXin Technology’s global DRAM market share has further improved from 7.67% in Q4 2025 to about 7.6% in Q1 2026, placing it fourth globally. SemiAnalysis predicts ChangXin Technology’s full-year 2026 revenue could exceed $50 billion, with market share rising from 9% in 2025 to 12% in 2027. But two Korean giants have already announced large-scale capacity expansion plans—ChangXin’s window is not infinite.

Full panorama of financial data: from massive losses to a surge

A three-year leap

| Year/period | Operating revenue (hundred million yuan) | YoY growth | Net profit attributable to shareholders (hundred million yuan) | Net profit after non-recurring items attributable to shareholders (hundred million yuan) | Gross margin | | --- | --- | --- | --- | --- | --- | | 2023 | 9.087 | — | -16.34 | -16.752 | Extremely low | | 2024 | 24.178 | 1.66 | -7.145 | -7.87 | 5.58% | | 2025 | 61.799 | 1.556 | 1.875 | 5.316 | 40.99% | | 2026Q1 | 50.8 | 7.191 | 24.762 | 26.341 | Not disclosed | | 2026H1E | 110-120 | 605.23 | 50-57 | — | — |

From 2022 to 2024, the company was in a sustained loss-making stage. In 2025, it achieved revenue of 66.88B yuan, with net profit attributable to shareholders of 66.88B yuan, turning around to profitability for the year.

Entering 2026, performance shows explosive growth. In the first quarter, the company generated operating revenue of 50.8 billion yuan, up 719.13% year over year; net profit attributable to shareholders was 57.92B yuan, turning from loss to profit. The company expects operating revenue for the first half of 2026 to reach 110 billion to 120 billion yuan, up 612.53% to 677.31%; net profit attributable to shareholders to be 50 billion to 57 billion yuan, up 2,244.03% to 2,544.19%. Based on the midpoint, first-half performance could completely wipe out all cumulative losses since the company’s establishment, with average daily profit close to 150 million yuan.

Where does the profit come from? Three-layer drivers

From massive losses in 2024 to the 500-570 billion yuan profit forecast for the first half of 2026, the profit surge is driven by three factors combined:

First layer: DRAM prices explode upward. According to a research report by Soochow Securities, the spot average price of DDR4 (8Gb) jumped from about $8.1 at the end of 2024 to about $140 by the end of June 2026, cumulative gains of more than 16 times. DDR5 (16Gb) rose rapidly from $20 in November 2025 to $117.5 by the end of June 2026, about 4.9 times. TrendForce data shows that in Q1 2026, contract prices for general-type DRAM rose 93% to 98% quarter over quarter, and again rose 58% to 63% in Q2. Gartner forecasts that global DRAM annual prices will increase 125% in 2026.

Second layer: capacity utilization moves from trough to full production.

Third layer: product mix upgrade. An increased share of DDR5 and LPDDR5X lifts ASP structure upward.

Among the three factors, the first two (price + capacity utilization) are cyclical, while the third (product mix upgrade) is structural—this is the core judgment for analyzing the sustainability of ChangXin’s performance.

Closing valuation analysis

Using the closing price of 49.00 yuan and total shares outstanding of 66.61B shares, the closing total market cap is about 3.28 trillion yuan. The implied valuations are:

• 2025 net profit attributable to shareholders: 280k yuan → extremely high static PE
• 2026 first-half forecast net profit attributable to shareholders: 50-57 billion yuan; annualized about 100-114 billion yuan → about 29-33x PE
• 2028 consensus forecast net profit attributable to shareholders: about 290.6 billion yuan → about 11.3x PE

Comparison: Micron Technology’s expected PE for 2026 is about 12x, SK hynix about 8x. ChangXin’s market share is less than one-third of Micron’s, yet its PE is more than twice. The market premium comes from three narratives: (1) China’s only DRAM IDM (scarcity), (2) certainty of domestic substitution (policy), (3) the AI supercycle (growth). Whether these three narratives are worth a 1.5-2x PE premium is the biggest disagreement in the current market.

Broker target valuation

Multiple brokers have issued target valuation ranges for ChangXin. Northeast Securities sets a 3.2-5.7 trillion yuan target based on three perspectives: relative market-share valuation, profitability decomposition, and market value per unit of capacity. Nomura Securities initiates coverage with a Buy rating and a target price of 116 yuan, corresponding to a market cap of about 7.76 trillion yuan—1.4 times the upper end of Northeast Securities. The 3.28 trillion yuan closing value on day one already falls into the lower bound of Northeast Securities’ range, showing the market’s significant valuation disagreement.

Competitive landscape: three triopoly players and challengers

| Company | Global market share (2025Q4) | 2026E PE | Core advantages | | --- | --- | --- | --- | | Samsung Electronics | 33.96% | ~15x | Full portfolio + HBM + scale | | SK hynix | 34.48% | ~8x | HBM leader (operating profit margin 72%) | | Micron Technology | 23.41% | ~12x | HBM + US domestic manufacturing | | ChangXin Technology | 7.67% | ~30x (~closing valuation) | China’s only IDM + DDR5/LPDDR5X |

The Chinese market is the largest DRAM consumption market globally. In 2025, mainland China’s DRAM market size was about 298 billion yuan, accounting for about 18% of global memory chip total consumption. Currently, China’s DRAM self-sufficiency rate remains low, and high-end HBM almost fully depends on overseas procurement. Even if ChangXin does not grab overseas market share, growth potential remains considerable based on domestic substitution alone.

At the same time, note that Samsung and SK hynix’s large-scale capacity expansion plans are underway. Global DRAM capex in 2026 is expected to grow 65% year over year—supply-demand dynamics may reverse in 2027-2028.

In addition, ChangXin Technology’s 2025 revenue came from traditional DRAM for more than 98%, and currently it has zero revenue in HBM. This means that in the most profit-elastic high-end market, ChangXin cannot yet compete with the three giants; but in the traditional DRAM market, it will face price pressure from the three giants’ expansions.

HBM: the biggest variable determining the valuation ceiling

ChangXin Technology’s current HBM revenue is zero. The company’s technology roadmap is leapfrogging R&D: its current status is that HBM3 has completed sample testing for submission, targeting mass production by the end of 2026. It plans to launch HBM3E in 2027 and start laying out HBM4 in 2028. The technical difficulty lies in 3D stacking and TSV vertical packaging processes—key breakthroughs for the medium-to-long term.

The gap versus the three giants is an order of magnitude: SK hynix’s HBM monthly capacity alone already exceeds 150k wafers. The HBM technical barrier is extremely high—covering TSV silicon vias, stacked packaging, hybrid bonding, and multiple advanced processes. Micron took about 5 years to achieve batch supply of HBM3E. Whether ChangXin can break through within 3 to 5 years is highly uncertain, but the upside is a valuable option.

Recent signals: In early June, NVIDIA CEO Jensen Huang confirmed that Samsung, Hynix, and Micron have all passed HBM4 certification. Morgan Stanley pointed out that from HBM3E to HBM4 to HBM4e, the value of memory per rack increases by more than four times. The track itself is still accelerating in expansion—if ChangXin cannot secure a position in time, the gap to the first tier will widen further.

There may be a “risky turn-around” leap in the technical route: the traditional 6F² architecture faces physical limits; a 4F² architecture could reduce unit area by about 30% and lower EUV lithography costs. ChangXin and the three original manufacturers are all advancing the 4F² approach. Industry analysts remind that the window is at most three years—if domestic manufacturers cannot push yields above 80% during the price-up cycle, once overseas giants complete their expansion, they could be pushed back into low-end markets again.

Risks investors need to watch

• DRAM cycle reversal risk (high) — ChangXin explicitly warns in its prospectus that the DRAM industry has a strong cyclical nature and high volatility; DRAM product prices are at historical highs, and sustained large-scale price increases are not sustainable. If prices peak and then fall, the company’s revenue and profits could face substantial downside risk.
• Supply-side capacity release pressure (high) — Samsung and SK hynix have announced expansion plans totaling 8 million trillion won; the three original manufacturers have simultaneously raised capex. Global DRAM capex in 2026 is expected to grow 65% year over year. Over the next few years,新增 capacity will be released gradually. If demand-side growth does not keep up with the speed of supply-side capacity deployment, the industry supply-demand structure could fundamentally reverse.
• HBM R&D and mass production coming in below expectations (medium-high) — Currently ChangXin Technology’s HBM revenue is zero. HBM3 is still in the sample testing stage, while the three giants have already mass-produced HBM4, leaving a gap on the order of magnitude level. If ChangXin cannot achieve large-scale mass production of HBM within the next 3 to 5 years, it may be locked into the position of a general-purpose DRAM supplier for a long time, missing the biggest profit incremental in the AI storage track.
• Overseas revenue proportion may be understated (medium) — In 2025, overseas revenue was 61.8B yuan, but after excluding Hong Kong transit trade and related transactions, revenue that truly enters overseas end markets is only about 2.79%. Being ranked fourth globally currently relies heavily on support from mainland China’s single market; a truly established globalized revenue structure has not been formed yet.
• Equipment and materials supply constraints (medium) — ASML EUV lithography machines face ongoing restrictions from US export controls, and DUV equipment delivery timelines have stretched to 18 months. Under equipment constraints, domestic DDR5 yields are currently about 65%, while Samsung has exceeded 85%, leaving a clear cost and efficiency gap.
• High turnover and high volatility risk on day one (medium) — The day-one turnover rate reached 61.5%, meaning more than half of the tradable shares completed turnover within the first trading day. Near-term speculation has been fully played out, but it also indicates the market’s pricing disagreement around 49 yuan is extremely large. The 44.4% intraday trading range (calculated as 30.8% based on the opening price and the low) shows that the stock’s near-term volatility risk is at an extremely high level.

Key observation windows for the next 12 months

• July 28-31, 2026: trading continues for the first 5 days after listing — after a 44% intraday range on day one, volatility on the following 4 days may be even more intense. The market needs to go through the first round of sentiment digestion.
• August 2026: official disclosure of the 2026 semiannual report — the final confirmed data of first-half guidance revenue of 110-120 billion yuan and net profit attributable to shareholders of 50-57 billion yuan will be released. This is the first “cushion” for the full-year performance expectations.
• October 2026: disclosure of the 2026 Q3 report — the single-quarter revenue and profit growth rates in Q3 are the most critical window for judging whether DRAM price increases show signs of marginal slowdown.
• November 2026: TrendForce releases its 2027 DRAM market outlook — the annual update to the supply-demand forecasting model will directly affect market judgments on where the 2027-2028 cycle inflection point lies.
• March 2027: release of the 2026 annual report — full-year financial data is finalized, and it will also disclose the first post-listing dividend plan, reflecting the company’s true stance toward shareholder returns.
• July 27, 2027: one year after listing; some lock-up shares are released — founder Zhu Yiming pledged not to reduce holdings for 10 years, but other strategic shareholders and early investors’ lock-up periods end. The actual selling pressure after unlock will be tested by the market.

Summary

ChangXin Technology’s day-one closing valuation of 3.28 trillion yuan is not just a price—it is a pricing of three core propositions:

Signal 1: A-shares’ ability to value hard tech is maturing. A single-day trading value of 129 billion yuan, turnover of 60.99%, and a trading range of 195.38%—these figures show that A-shares can provide sufficient liquidity and pricing depth for mega IPOs. The prior record was East Money’s 90 billion yuan single-day trading value (October 2024). ChangXin pushed it to 129 billion yuan in one shot. This is not just “speculating on the new listing,” but a full-scale battle with real money.

Signal 2: The market has disagreement on the cycle. The move—chasing up to 535% in the morning and then getting smashed down to 340% in the afternoon—perfectly reflects fierce divergence in how the market views DRAM’s cycle position. Bulls see an AI-driven structural regime change; bears see a historical replay of a top-of-cycle IPO. These two narratives will run through the first full cycle after ChangXin’s listing.

Signal 3: The safety margin of 3.28 trillion yuan is thin. Using annualized first-half 2026 profit of about 107 billion yuan (midpoint), the implied PE is about 30x. If DRAM prices peak and then fall in 2027, this valuation will quickly come under pressure. The lower end of broker target ranges (3.2 trillion yuan from Northeast Securities) has already been touched. Future upside depends on three variables: how long DRAM prices stay at high levels, ChangXin’s HBM R&D progress, and the actual rollout pace of capacity expansion by Samsung/SK hynix.

ChangXin Technology FAQ

Is ChangXin Technology’s 3.28 trillion yuan valuation expensive?

By static PE, it’s extremely expensive. But based on annualized first-half 2026 profits, the implied PE is about 29-33x. Using Nomura’s 2028 forecast EPS of 5.8 yuan and a target PE of 20x, the implied target price is 116 yuan. For a DRAM maker still growing at high speed, with market share climbing from 7.6% toward higher targets, the forward valuation has some support. The key question is whether the 2028 profit forecast can be realized. This entirely depends on whether DRAM prices can remain at high levels for the next three years. The 195% intraday range on day one already shows the market’s huge disagreement on this.

How big is the gap between ChangXin and the three giants?

On the product lines, DDR5 and LPDDR5X have entered the international first-tier (10,667 Mbps). The technical generation gap has narrowed to about 1-2 years. In HBM, the gap is enormous: the three giants have already mass-produced HBM4, while ChangXin’s HBM3 is still in the sample testing stage, and HBM revenue is zero. On capacity scale, ChangXin’s monthly capacity is about 280k wafers (12-inch equivalent), while Samsung’s DRAM monthly capacity is 5 to 7 times as large.

When will DRAM prices fall?

TrendForce expects the supply-demand balance point to occur around Q4 2027. Historical experience shows that price inflection points often come earlier than the supply-demand balance point because the market trades expectations in advance. The August 2026 and October earnings reports are key observation windows—if Q3’s single-quarter growth slows, it may mean the cycle is approaching the top. Also watch the actual start and ramp-up progress of Samsung/SK hynix’s large-scale expansions: once equipment starts moving in, supply-side expectations could change fundamentally.

What does the day-one turnover rate of 60.99% indicate?

At two levels. Good side: full turnover means that more than 60% of tradable shares have already turned over, releasing a large portion of near-term speculative selling pressure, so later volatility may converge. Bad side: it indicates the market has an extreme disagreement on the pricing around 49 yuan—some think it’s a value trough, while others think it’s the top of the cycle. For a new stock with turnover above 60%, the subsequent performance depends heavily on whether fundamentals deliver.

Is ChangXin Technology worth following?

ChangXin Technology is one of China’s most representative self-controlled hard-tech companies. It is also the only company in mainland China with full capabilities as a DRAM IDM. Against the backdrop of an AI supercycle, its strategic value goes beyond short-term valuation. But investing requires answering three core questions: Do you believe the DRAM supercycle can continue through 2028? Do you believe ChangXin can achieve HBM mass production before 2028? And how much interim volatility are you willing to tolerate in between those two beliefs? If you cannot answer any of these, tracking rather than holding may be a more prudent choice.