CXMT IPO Debut: China’s Memory Champion Gained 466% on a 6.7% Float. Here’s What That Actually Prices In

Last updated: July 27, 2026, 6:15 p.m. ET

ChangXin Memory Technologies closed its first day of trading on Shanghai’s STAR Market at 49.00 yuan, up 466% from an offer price of 8.66 yuan. That single session moved China’s largest DRAM manufacturer from a company nobody could buy to the most valuable business listed on a mainland Chinese exchange, at roughly 3.3 trillion yuan — about $487 billion at Monday’s exchange rate of approximately 6.77 yuan to the dollar.

The number that explains the number is 6.73%. That is the share of CXMT’s enlarged capital that was actually free to trade on Monday. Everything else — the founder’s stake, the state investment vehicles, the strategic allocations to Alibaba, Tencent, Xiaomi, Meituan, ZTE, Chery and NIO — sits behind lockups running from twelve to thirty-six months. A little over four and a half billion shares changed hands against a valuation built on roughly sixty-seven billion.

That does not make the debut fake. CXMT’s operating turnaround is real, documented in its prospectus, and larger than almost anything in the recent history of the semiconductor industry. First-quarter revenue rose 719% year over year. A company that lost money for nine consecutive years now expects to book more profit in six months than most listed Chinese firms earn in a decade. But the gap between what the business has proven and what Monday’s close assumes is wide enough to matter, and the mechanics of a 6.73% float are the reason it opened so fast.

Key Takeaways

• Main development: CXMT (Shanghai Stock Exchange STAR Market, ticker 688825) began trading on Monday, July 27, 2026, closing at 49.00 yuan against an IPO price of 8.66 yuan — a first-day gain of approximately 466%, with an intraday high of 49.50 yuan, or roughly 471.6% above the offer.
• Key figure: The IPO raised at least 57.9 billion yuan (approximately $8.6 billion), rising to as much as 66.6 billion yuan if the over-allotment option is exercised. It is the largest listing in STAR Market history, surpassing SMIC’s 53.2 billion yuan offering in 2020, and mainland China’s largest since Agricultural Bank of China in 2010.
• Market response: The close valued CXMT at approximately 3.3 trillion yuan (about $487 billion), overtaking Industrial and Commercial Bank of China at roughly 2.6 trillion yuan to become the most valuable company listed on a mainland exchange, and exceeding the market capitalizations of Intel and Cisco.
• Why it matters: CXMT is the world’s fourth-largest DRAM producer and the centerpiece of Beijing’s effort to build a domestic memory supply chain. Its listing arrives during the sharpest memory shortage in fifteen years and three weeks after SK Hynix raised $26.5 billion on Nasdaq — the largest foreign IPO in U.S. history.
• What comes next: STAR Market rules suspend daily price limits for the first five trading sessions; a 20% daily band applies from the sixth. Separately, the Pentagon reaffirmed CXMT’s designation as a Chinese military company on June 8, 2026, and U.S. procurement restrictions tied to that list took effect June 30.

The arithmetic of a 6.73% float

Start with what the offering itself said the company was worth. At 8.66 yuan per share, CXMT’s IPO valued the business at approximately 583 billion yuan — call it $86 billion. That price was set by a bookbuild among institutional investors who had access to the prospectus, the audited financials and the management presentations. It was not a token valuation. It made CXMT one of the larger technology listings anywhere in the world this year before a single share traded.

By 9:45 a.m. Shanghai time the market had multiplied that figure by roughly five and a half.

Two structural features made the move possible. The first is the STAR Market’s trading regime. The Shanghai Stock Exchange’s Science and Technology Innovation Board, launched in 2019 as a domestic answer to Nasdaq, imposes no daily price limit during a newly listed stock’s first five sessions, against the 44% first-day cap that applies on China’s main boards. Intraday circuit breakers exist — a 10-minute halt if the price moves 30% from the open, another at 60% — but they slow a repricing rather than stop it. A stock that wants to travel 466% in a day is permitted to.

The second is supply. Of roughly 6.7 billion new shares issued, a substantial portion went to strategic investors on lockups. Alibaba alone was reported to have committed 7.6 billion yuan. Xiaomi, Tencent, Meituan, ZTE, Chery and NIO were each allotted 18.24 million shares for 158 million yuan apiece, with lockup periods running twelve to thirty-six months depending on investor category; NIO’s subsidiary accepted an eighteen-month lock. What survived that process as genuinely tradable stock was 6.73% of the enlarged capital.

Run the numbers and the picture sharpens. A 3.3 trillion yuan valuation on approximately 67 billion shares implies a free float worth roughly 222 billion yuan — about $33 billion — supporting a headline market capitalization of $487 billion. Fourteen dollars of valuation for every dollar of stock that can actually be bought or sold. These figures are calculated from the reported market capitalization, offer size and disclosed float percentage rather than taken directly from a filing, and the share count is approximate.

Retail demand did the rest. Subscription exceeded 200 times the retail tranche, leaving a winning allocation rate around 0.47%. Fewer than one in two hundred retail applications was filled. Every investor who wanted CXMT and did not get it in the allocation had exactly one way to obtain shares on Monday: bid for a slice of that 6.73%.

This is not a Chinese peculiarity, and it is not evidence of manipulation. Low-float debuts produce violent first-day moves in every market that permits them. What it does mean is that Monday’s closing price is a poor guide to what a broad, liquid market would pay for the whole company. It is the clearing price for a small quantity of stock among buyers with unusually concentrated demand — a fact that will begin to change only as lockups roll off, starting twelve months from now.

Theodore Shou, chief investment officer at Skybound Capital, made a version of this point in a television interview on the day of the debut, describing the listing as driven in the short term by sentiment and momentum, and saying he expected it to fall gradually back into line with the broader market over time. He was more positive on the company itself, saying he had no doubt CXMT would grow into a global leader in the sector. But he drew a sharp line between the business and the moment: the margins and net profitability visible today, he argued, are not sustainable and need to be normalized across a cycle. His view was that the industry sits at the peak of a demand-supply imbalance, not necessarily the peak of the share price — and that for average individual investors there was no need to rush in at current valuations.

What the financials actually show

The operating story underneath the float mechanics is genuinely extraordinary, and it deserves to be read carefully rather than admired from a distance.

CXMT reported first-quarter 2026 revenue of 50.8 billion yuan, approximately $7.51 billion, a 719% increase from the same quarter of 2025. That comparison implies a year-earlier quarter of roughly 6.2 billion yuan. For the first half of 2026, the company guided to revenue of 110 billion to 120 billion yuan — meaning six months of 2026 will produce close to double the 61.8 billion yuan CXMT recorded across the whole of 2025.

Profit moved further and faster. The prospectus guides to first-half 2026 net profit of 50 billion to 57 billion yuan, against a net loss of 2.3 billion yuan in the first half of 2025. Both figures are company guidance disclosed in the listing document, not audited half-year results, and should be read as such.

Take the midpoint of both ranges and the implied net margin for the first half of 2026 is roughly 46%. For context, CXMT’s full-year 2025 net margin was about 3% — 1.88 billion yuan of profit on 61.8 billion yuan of revenue, the company’s first profitable year after nearly a decade of losses. Before that: a loss of 8.33 billion yuan in 2022, 16.34 billion yuan in 2023, and 7.14 billion yuan in 2024. Accumulated unrecovered losses still stood at 36.65 billion yuan at the end of 2025.

A 3% net margin becoming a 46% net margin in eighteen months is not an operational achievement. No manufacturer improves that much through execution. It is what happens to a commodity producer with high fixed costs when the price of the commodity roughly triples and the incremental revenue drops almost intact to the bottom line. The same mechanism, running the other way, is what produced the 16.34 billion yuan loss in 2023.

This is the single most important thing to understand about valuing CXMT, and it is where Monday’s price becomes difficult to defend on fundamentals.

Annualize the guided first-half profit naively and you get 100 billion to 114 billion yuan. At the IPO price of 8.66 yuan, the company was valued at roughly five and a half times that figure. At Monday’s close, it is valued at roughly thirty times.

Neither multiple should be taken at face value, because annualizing peak-cycle earnings for a memory manufacturer is precisely the wrong way to value one. A five-times multiple looks cheap only if peak margins persist; a thirty-times multiple looks reasonable only if today’s earnings are a floor rather than a ceiling. The honest framing is that the market moved CXMT from a price that assumed sharp normalization to a price that assumes very little — in a single session, on 6.73% of the shares, with no new information about the business between the bookbuild and the open.

One more disclosure deserves attention. Accumulated unrecovered losses of 36.65 billion yuan at the end of 2025 are not merely historical color. Under Chinese company law, distributable profits are calculated after prior losses are made good, which shapes when and whether CXMT can begin returning capital to shareholders. An investor buying at thirty times annualized peak earnings is buying a company that must first work through nearly 37 billion yuan of accumulated deficit before the question of dividends becomes live.

How CXMT got here

The company was founded in Hefei, capital of Anhui province, in 2016. Its technical foundation came from an unusual source: patents and engineering talent from Qimonda, the German memory manufacturer spun out of Infineon that entered insolvency in 2009. Buying the intellectual estate of a failed European DRAM maker gave CXMT a starting point that would otherwise have taken years and a great deal of litigation to assemble.

The figure at the center of it is Zhu Yiming, founder of GigaDevice Semiconductor — a Chinese designer best known for NOR flash memory used to store code in electronics — who is described in company filings as central to CXMT’s creation and development and who later became its chairman. Forbes estimated his fortune at $15.9 billion following Monday’s debut. He has also drawn scrutiny: reporting on the offering noted that he sold roughly 2.5 billion yuan of GigaDevice shares while pledging CXMT stock worth more than 20 billion yuan toward employee incentives.

What made the company viable was not the Qimonda patents but the patience of its backers. Hefei’s municipal government and Anhui provincial vehicles absorbed nine years of losses while a local supply chain was assembled around the fabs. The prospectus puts state-owned shareholders at 36.29% before the IPO, including Anhui Investment Group and the National Integrated Circuit Industry Investment Fund Phase II — the vehicle universally known in the industry as the Big Fund.

That figure is worth stating precisely, because a broadcast account of the debut described state investors as holding “roughly half” of the company. The prospectus number is 36.29%. Separate counts that aggregate all Hefei-linked entities land near 36.8%. Both are substantial; neither is half. The distinction matters for anyone assessing how much of CXMT’s future capital allocation will answer to industrial policy rather than to minority shareholders, and it is the kind of number worth taking from the filing rather than from a live television segment.

Government support was not only financial. Chinese authorities have restricted state-owned enterprises from purchasing foreign memory, according to two people who spoke to Reuters, and have asked CXMT and its flash-memory counterpart YMTC to prioritize domestic customers. A guaranteed home market with structurally suppressed foreign competition is a material asset, and it does not appear on the balance sheet.

The shortage that changed everything

CXMT’s transformation is inseparable from what has happened to memory pricing since 2025, and that story starts with a capacity decision made by its competitors.

High-bandwidth memory — the stacked, ultra-fast DRAM that feeds AI accelerators — carries far better margins than conventional DDR5. It also consumes roughly three to four times the wafer capacity per gigabyte. As Samsung, SK Hynix and Micron redirected fabs toward HBM to serve AI data center demand, they withdrew capacity from the conventional DRAM that goes into phones, laptops, cars and general-purpose servers. Micron exited the consumer memory market entirely. SK Hynix has said its HBM, DRAM and NAND capacity is essentially sold out for 2026.

The result is a supply-demand gap that industry researchers put at 4.9% for DRAM, 4.2% for NAND and 5.1% for HBM in 2026 — the widest since 2011. Prices responded accordingly. Contract pricing for DDR5 has more than doubled. Samsung raised 32GB DDR5 module pricing to $239 from $149 in a single step, a 60% increase, and its DRAM average selling price rose roughly 90% quarter over quarter in the first quarter of 2026, with a further 50% to 60% increase in the second.

Suppliers have restructured how they sell. SK Hynix has reportedly abandoned long-term contract price caps so that supply prices fully reflect spot-market shortages. Micron has kept ceiling and floor mechanisms but reset the cap to the highest market level of the second quarter of 2026. Both moves transfer pricing risk to buyers, and both are only possible in a market where customers have nowhere else to go.

This is the environment that produced CXMT’s 719% revenue growth. It is not a story about CXMT winning share from Samsung through superior technology. It is a story about a global commodity shortage lifting every producer’s realized price simultaneously, with CXMT lifted hardest because it entered the cycle with the lowest base.

Which raises the question that determines everything about the valuation: how long does it last? Several research firms expect the shortage to persist at least through the end of 2027, and SK Hynix has warned it could extend past 2030. Those are forecasts, not commitments, and memory forecasts have an unusually poor track record precisely because the industry’s response to high prices is to build capacity, and capacity takes about two years to arrive. CXMT itself is building two new plants, in Shanghai and Hefei, with talks underway for a third.

From price taker to price setter

The most consequential reporting on CXMT published in the run-up to the listing was not about the IPO at all. On July 24, three days before the debut, Reuters correspondent Fanny Potkin published an investigation based on interviews with more than a dozen executives, engineers, suppliers and U.S. officials, along with a review of fifty Chinese government policy documents and company disclosures.

Its central finding is that CXMT has stopped behaving like a discount alternative.

For months, according to two people familiar with the matter, CXMT had been raising prices on Huawei — one of China’s largest technology companies and, until recently, precisely the sort of customer a state-backed national champion might be expected to accommodate. Huawei demanded relief. CXMT held firm.

The dispute reached the factory floor in June. Engineers from SiCarrier, a chipmaking-equipment vendor with deep strategic ties to Huawei, had been working in the cleanrooms of CXMT’s core research and development zone in Hefei, helping with equipment maintenance. Without warning, CXMT ordered them to pack their tools and leave immediately. SiCarrier executives concluded the confrontation was a product of the power struggle between the two companies. The firms still do business, but the engineers have not been allowed back into the R&D zone. CXMT, Huawei and SiCarrier did not respond to Reuters’ questions about the incident.

On pricing, four people told Reuters that Chinese memory makers are now picking clients and dictating terms. In recent weeks CXMT has charged more than Samsung’s roughly $1,240-per-unit price for comparable 64-gigabyte DDR5 server memory modules, according to two of those sources, who declined to disclose the precise CXMT price. Samsung did not respond to questions; SK Hynix declined to comment.

The commercial consequences are visible in the contracts. CXMT signed a five-year agreement with ByteDance this month worth more than $7 billion, according to three people familiar with the arrangement — a deal Reuters reported for the first time. It followed a June agreement with Tencent worth over $3 billion. ByteDance and Tencent did not respond to requests for comment.

There is a counterweight in the same reporting. Several Chinese electronics and technology firms complained to China’s Ministry of Industry and Information Technology this year about price increases by CXMT and YMTC, blaming the hikes for delaying product launches, according to two people; the sources declined to name the complaining companies. The ministry, which did not respond to Reuters’ questions, said in April that it would crack down on memory-chip hoarding aimed at driving up prices.

That last detail is easy to skip past and shouldn’t be. CXMT’s pricing power at home exists partly because Beijing restricted the alternatives. The same authority that created the protected market can decide the protection has become expensive for the rest of Chinese industry. A national champion that raises prices on Huawei and delays domestic product launches is testing the limits of the arrangement that made it.

The read-across: incumbent memory stocks fell

The most useful evidence about what CXMT’s debut means did not come from CXMT’s own share price. It came from what happened to everyone else’s.

By late Monday morning in New York — around 11:07 a.m. Eastern, according to 24/7 Wall St.’s market report — Micron Technology was down approximately 5% at $871. SK Hynix’s newly listed ADRs were off roughly 6% at $145, giving back an earlier gain. SanDisk fell about 12% to $1,270 and Western Digital about 7% to $483. The Roundhill Memory ETF, in which Samsung, SK Hynix and Micron together account for roughly 72% of net assets, dropped around 4% to $51. All of this occurred on a day when the broader Nasdaq 100 was mixed and the S&P 500 was modestly higher. Prices are intraday and subject to change; readers should check current quotes.

A coordinated decline across both DRAM and NAND names on an otherwise unremarkable market day is a sector signal rather than a single-stock story. It was also not the first. Memory stocks sold off on July 15 and 16 as the scale of CXMT’s offering became clear, with Micron falling and, by one account, briefly surrendering a $1 trillion market capitalization.

What makes the reaction interesting is that it contradicts the simple bullish reading of CXMT’s debut. If the memory shortage were purely a demand story, a new entrant’s listing would be neutral or positive for the sector — confirmation that the market is large enough to support a fourth major producer. The selling suggests investors read it differently: as a signal that Chinese supply will eventually arrive, and that the record margins currently supporting incumbent valuations have a competitive expiry date.

Those incumbent margins are worth stating, because they establish what is at risk. Micron reported fiscal third-quarter 2026 revenue of $41.46 billion, up 345.7% year over year, with non-GAAP earnings per share of $25.11, and guided fourth-quarter revenue to approximately $50 billion. SanDisk posted fiscal third-quarter revenue of $5.95 billion with a 78.4% gross margin and non-GAAP EPS of $23.41; chief executive David Goeckeler described it as a fundamental inflection point for the business. Note that these are non-GAAP figures as reported by the companies, and that Micron and SanDisk use different fiscal calendars from each other and from the calendar-year data cited elsewhere in this article.

Positioning amplified the move. Heading into Monday, SanDisk was up approximately 505% year to date, Micron approximately 223%, and Western Digital approximately 202%. A sector that has multiplied several times over in seven months does not need much bad news to produce a 5% down day, and some of Monday’s selling has the character of profit-taking rather than fundamental reassessment.

One additional detail from the same report deserves attention because it connects directly to the Washington fight: Apple is reportedly testing CXMT’s DRAM chips. If accurate, that would represent Chinese memory reaching a tier-one Western customer considerably sooner than the incumbents’ bulls have assumed, and it explains why Apple has argued against an Entity List designation while Micron has lobbied for one. Both companies’ positions follow straightforwardly from their commercial interests.

The counterargument is being made too. Research desks at Morgan Stanley and Mizuho have characterized the recent weakness in memory names as a buying opportunity rather than the beginning of a downturn — attributed views, not consensus, and worth weighing against the analysts’ own coverage incentives. South Korea unveiled a $950 billion AI initiative package on Saturday, July 25, involving Samsung, SK Group and U.S. technology partners. And the near-term supply arithmetic has not changed: CXMT’s new fabs do not come online before 2027, and its access to advanced tooling remains restricted, so nothing about Monday’s listing eases the current shortage.

The immediate test arrives quickly. SK Hynix reports second-quarter results on July 28 after the U.S. close, and whatever management says about 2027 DRAM supply will either reinforce or undercut the competitive-threat narrative that moved these stocks on Monday. Micron’s next report is scheduled for September 28.

The gap that money cannot close quickly

CXMT is the world’s fourth-largest DRAM producer. It is also, by the assessment of the people who watch it most closely, several years behind the three companies ahead of it in the technologies that matter most for AI.

The prospectus puts CXMT’s 2025 global DRAM market share at about 7.7%. TrendForce ranked it fourth by revenue in the first quarter of 2026 with 7.6%, against a combined 89.7% for Samsung, SK Hynix and Micron. Counterpoint Research, measuring by shipments, put CXMT at roughly 8% of the global market in 2025 — Samsung 36%, SK Hynix 29%, Micron about 24% — rising to approximately 9% in the first quarter of 2026, with a forecast of about 11% by 2028.

Counterpoint’s estimate of what CXMT actually needs is the more interesting number: at least 15% global share to be competitive over the long term. On the firm’s own forecast, CXMT does not reach that threshold by 2028.

Why high-bandwidth memory decides this

For readers who do not follow semiconductors closely, the distinction between conventional DRAM and HBM is where the entire competitive question lives, and it is worth setting out plainly.

Conventional DRAM sits beside a processor on a circuit board and communicates with it across a relatively narrow channel. That arrangement is adequate when the processor’s job is running applications. It is inadequate when the job is training or serving a large AI model, because the calculation is no longer limited by how fast the chip can compute but by how fast data can be delivered to it. An accelerator waiting on memory is an expensive idle asset.

High-bandwidth memory solves this by stacking DRAM dies vertically and connecting them to the processor through a very wide interface, placed on the same package. The result is an order-of-magnitude increase in data delivered per second. It is also considerably harder to manufacture: the dies must be thinned, stacked, and bonded with through-silicon vias, and yield losses compound across the stack. Producing a gigabyte of HBM consumes roughly three to four times the wafer capacity of a gigabyte of standard DDR5.

Two consequences follow, and both matter for CXMT.

The first is that HBM is where the industry’s profit has migrated. Nvidia-class accelerators require it, AI data center buildouts require those accelerators, and the memory makers who can supply qualified HBM at volume capture pricing that conventional DRAM has never supported. The second is that HBM’s wafer intensity is what created the shortage in everything else. When Samsung, SK Hynix and Micron moved capacity into HBM, each gigabyte of AI memory they produced removed three to four gigabytes of conventional DRAM from the market. CXMT’s windfall is a direct consequence of its competitors chasing a product it cannot yet build competitively.

That is a comfortable position while the shortage lasts and an uncomfortable one afterward. It means CXMT’s current earnings are a byproduct of a market it does not participate in, generated by a supply gap its rivals opened and can close. And it means the company’s long-term position depends on closing a gap in the one product where the manufacturing difficulty is highest and the tooling constraint bites hardest.

The technology gap concentrates in high-bandwidth memory. CXMT has produced its own HBM, but five sources told Reuters it remains two generations — several years — behind its rivals. Since HBM is the memory that Nvidia-class AI accelerators require, and since U.S. restrictions bar China from importing advanced HBM, this is simultaneously CXMT’s largest commercial opportunity and its most stubborn technical constraint. Kyle Chan, a fellow at the Brookings Institution who studies China’s technology policy, described CXMT as China’s best shot at developing its own cutting-edge HBM to power Chinese AI models, while noting the company faces supply chain bottlenecks in scaling manufacturing capacity.

The bottleneck has a name: lithography. Both CXMT and YMTC depend on deep ultraviolet lithography machines from ASML of the Netherlands. Their Korean and American competitors make DRAM using ASML’s more advanced extreme ultraviolet systems, which China has been unable to obtain since the Dutch government began withholding export licences in 2019. The Dutch government has faced sustained U.S. pressure to restrict DUV sales as well.

“If more restrictions are imposed on lithography equipment, that would be the biggest challenge for Chinese memory makers,” said Ray Wang, an analyst covering memory and AI supply chains at SemiAnalysis. “China remains quite behind in that part of the equipment supply chain compared to other tool segments.” ASML declined to comment on the potential impact of future export regulations. MS Hwang, a research director at Counterpoint specializing in memory semiconductors, put it more bluntly: trade restrictions on tools remain the key challenge for CXMT.

The asymmetry with YMTC is instructive. YMTC has been on the U.S. Entity List since 2022 and has responded by replacing roughly half its equipment with domestic machinery and developing techniques to stack memory layers using less advanced tools, according to two people who spoke to Reuters. Having already absorbed the shock, YMTC is more insulated from further restrictions than CXMT — which has never been listed, and has consequently never been forced to build the workarounds.

CXMT’s answer is scale. The company is building fabs in Shanghai and Hefei and is in talks with local authorities about a third, according to three sources. Together those projects would more than double capacity to more than 600,000 wafers per month; on plan, one person said, CXMT’s capacity would overtake Micron by 2030. It aims to enter the U.S. market in the long term, three sources said, but domestic demand currently consumes its output. New fabs coming online as soon as 2027 could allow it to serve both.

The prospectus says IPO proceeds will fund exactly this: production capacity, manufacturing technology upgrades, and research and development.

Washington was already moving

The regulatory story around CXMT is more advanced, and considerably stranger, than the debut coverage suggested.

The Pentagon added CXMT to its Section 1260H list of Chinese military companies operating in the United States in January 2025. In February 2026 the Department of Defense published an updated list that appeared to remove both CXMT and YMTC — then deleted it almost immediately, without explanation. On June 8, 2026, the Department published a new version adding 65 entities, including Alibaba, Baidu, BYD, BOE, NIO and Unitree. The principal substantive difference between the withdrawn February list and the June list, according to a client alert from the law firm WilmerHale, is that CXMT and YMTC remain designated.

Both companies deny the military-civil fusion charge underlying the designation. China’s Ministry of Foreign Affairs responded to the June update by calling the 1260H list an example of the United States overstretching national security to target Chinese businesses, and said China would take necessary measures. As of WilmerHale’s alert, no specific retaliatory action had been taken.

What has changed is that 1260H designation now carries legal weight rather than reputational weight alone. Section 805 of the FY2024 National Defense Authorization Act bars the Pentagon from entering into, renewing or extending contracts with listed entities — an Entity Ban that took effect June 30, 2026, with a broader Product Ban following June 30, 2027. Section 8531 of the FY2026 NDAA requires the President to report to Congress within two years on whether any 1260H entity qualifies for Treasury’s Non-SDN Chinese Military-Industrial Complex Companies list, a designation that would prohibit U.S. persons from buying or selling the company’s publicly traded securities.

That last provision is the one with direct relevance to CXMT’s newly listed stock, and it is a genuine tail risk rather than a certainty. No such determination has been made.

Meanwhile, a separate fight is running over the Entity List — the Commerce Department instrument that restricts access to U.S.-origin technology. Reuters reported last month that CXMT was approved by a U.S. interagency committee for addition to the blacklist last year, but that officials have held off. The lobbying on both sides is unusually visible. Apple has argued it needs Chinese memory and has sought assurances that CXMT will not be listed, according to two people familiar with the discussions. Micron, the Chinese firms’ principal Western competitor, has pushed U.S. lawmakers to enact further restrictions on both CXMT and YMTC, including curbs on their access to chipmaking equipment. Apple, Micron, the White House and the Departments of Commerce, Defense and State did not respond to Reuters’ questions.

Congress is applying pressure of its own. House Select Committee on China chair John Moolenaar and House Foreign Affairs Committee chair Brian Mast have called on the administration to issue an executive order prohibiting U.S. persons and U.S.-incorporated entities from procuring DRAM, HBM or other memory components from YMTC, CXMT or any 1260H-designated entity, characterizing sourcing from either company as a grave mistake.

The Trump administration is divided on whether to act, according to four people familiar with the discussions. That division is the operative fact. A company whose regulatory fate depends on an unresolved argument inside a single government is a company whose valuation carries a discount that Monday’s close did not obviously apply.

The other twin star

CXMT does not travel alone. Within China’s semiconductor industry, it and Yangtze Memory Technologies Corporation are known as the “twin stars” of memory — CXMT in DRAM, YMTC in NAND flash. Understanding the pair matters for reading what Monday means, because the two companies represent a controlled experiment in what U.S. restrictions actually do.

YMTC was added to the Commerce Department’s Entity List in 2022, restricting its access to U.S.-origin suppliers, software and tools. CXMT has never been listed. Four years later, the two companies occupy noticeably different positions.

YMTC responded to restriction by localizing. It has replaced roughly half its equipment with domestic machinery and developed techniques to stack memory layers using less advanced tools, according to two people who spoke to Reuters. The result is a company that is now more insulated from further U.S. action than its unrestricted sibling — an outcome that runs directly counter to the intent of the designation, and one worth weighing against the argument that adding CXMT to the Entity List would meaningfully slow it.

YMTC has also gone on the offensive commercially. In June it entered the South Korean market with a consumer memory storage brand, exploiting a segment that Samsung, SK Hynix and Micron have vacated as they shifted toward more advanced products. Selling consumer storage in Samsung’s home market is a deliberate gesture as much as a commercial one.

And YMTC is preparing its own listing. Some executives are internally pushing for a valuation target of 1 trillion yuan — approximately $148 billion — according to two people familiar with the discussions. That figure was reported before CXMT’s debut. Whether it now looks conservative or merely reflects the same conditions that produced Monday’s repricing is a question the market will answer.

The pairing also clarifies what the Big Fund actually is. Both companies are backed by the National Integrated Circuit Industry Investment Fund, alongside provincial support — Anhui for CXMT, Hubei for YMTC. This is not venture capital with a strategic tilt. Chinese officials describe the companies in corporate filings and policy documents as strategic infrastructure central to technological self-reliance, and the funding behaves accordingly: patient across nine years of losses, indifferent to the return timelines that would govern a private investor, and paired with demand-side intervention in the form of restrictions on state-owned enterprises buying foreign memory.

For an investor pricing CXMT at thirty times annualized peak earnings, that architecture cuts both ways. It is the reason the company survived to reach this point and the reason it is unlikely to fail in a downturn. It is also the reason its capital allocation will not always optimize for minority shareholders, and the reason a fourth producer may persist in a market that has historically supported three.

Timeline

• 2009: Qimonda, the German DRAM manufacturer spun out of Infineon, enters insolvency. Its patents and engineering talent later form part of CXMT’s technical foundation.
• 2016: CXMT founded in Hefei, Anhui province.
• 2019: The Dutch government begins withholding export licences for ASML’s extreme ultraviolet lithography systems to China.
• 2022: YMTC added to the U.S. Entity List. CXMT reports a net loss of 8.33 billion yuan.
• January 2025: The Pentagon adds CXMT to the Section 1260H list of Chinese military companies.
• FY2025: CXMT records its first annual profit — 1.88 billion yuan on revenue of 61.8 billion yuan — after nearly a decade of losses.
• February 2026: The Pentagon publishes an updated 1260H list appearing to remove CXMT and YMTC, then deletes it without explanation.
• April 2026: China’s Ministry of Industry and Information Technology says it will crack down on memory-chip hoarding aimed at driving up prices.
• Q1 2026: CXMT revenue reaches 50.8 billion yuan, up 719% year over year.
• June 2026: CXMT expels SiCarrier engineers from its Hefei R&D cleanrooms amid a pricing dispute with Huawei, per Reuters. CXMT signs a supply agreement with Tencent worth over $3 billion.
• June 8, 2026: The Pentagon publishes a 1260H update adding 65 entities; CXMT and YMTC remain designated.
• June 30, 2026: Section 805 Entity Ban on Pentagon contracting with 1260H entities takes effect.
• July 10, 2026: SK Hynix raises $26.5 billion in a Nasdaq ADR listing, the largest foreign IPO in U.S. history.
• July 22, 2026: CXMT’s IPO prospectus is filed with the Shanghai Stock Exchange.
• July 24, 2026: Reuters publishes its investigation into CXMT and YMTC’s pricing power, revealing the ByteDance agreement.
• July 27, 2026: CXMT debuts on the STAR Market, closing at 49.00 yuan, up approximately 466%.

Reading the day-one commentary

Broadcast coverage of the debut converged on three claims. Two hold up well. The third does not, and the errors in the fourth are worth correcting because they have propagated.

The claim that scarcity drove the move is correct and is the most important thing said about the listing. A 6.73% float, more than 200 times retail oversubscription and a 0.47% allocation rate produce exactly this outcome. The related observation that investors sold existing technology holdings to fund participation is plausible and consistent with how large domestic offerings behave, though it is difficult to verify directly from public data.

The claim that CXMT has gained real commercial leverage is also correct, and Reuters’ reporting supports it more strongly than the on-air summary conveyed. Charging above Samsung on comparable DDR5 server modules, holding firm against Huawei’s demand for relief, and signing more than $10 billion of five-year supply agreements with ByteDance and Tencent is a different company from the budget alternative of three years ago.

Three factual points need correction. State investors were described as holding “roughly half” of CXMT; the prospectus says state-owned shareholders held 36.29% before the IPO. A market capitalization of “more than $480 billion” was cited during the opening minutes; the figure derived from the closing price is approximately $487 billion, and any intraday market capitalization on a first-day low-float listing should be treated as an estimate rather than a measurement. And a claim that the stock was up “more than 530%” during the session is not corroborated by the exchange-linked reporting, which puts the intraday peak at 49.50 yuan, or 471.6%. In fairness to the broadcast, it was not alone: at least one U.S. market report described the debut as “more than 500%” with a market capitalization near $540 billion. Live coverage of a first-day listing with no price limit and a 6.73% float will produce competing numbers, which is itself an argument for waiting on the close.

One further slip: the segment referred to whether “SMIC” could use the IPO proceeds to narrow the technology gap. SMIC is a separate company, China’s largest contract chipmaker, listed on the STAR Market since 2020. The proceeds in question are CXMT’s.

The most valuable commentary of the day came from Theodore Shou of Skybound Capital, and it was more pointed than the framing around it. Shou’s argument was structural: the current margins reflect a demand-supply imbalance at its peak, not a permanent repricing of the business, and normalization across a cycle is the appropriate lens. He expected survivors of the eventual downturn to emerge stronger — a reasonable historical read of memory consolidation — while cautioning individual investors against rushing in at current valuations. That is a considerably more careful position than “China’s biggest listed firm,” and it is the one the evidence supports.

The case for CXMT

The strongest argument for the company does not depend on Monday’s price being correct.

CXMT occupies a position that is genuinely difficult to replicate. It is the only pure-play listed memory manufacturer in the world’s largest electronics market, operating in an industry where the barriers to entry are measured in tens of billions of dollars and years of process learning. Chinese state-owned enterprises are restricted from buying foreign memory. Authorities have asked CXMT to prioritize domestic customers. That is a protected demand base of a scale no Western competitor enjoys.

The technology trajectory is real, if incomplete. TrendForce notes that CXMT’s transition to its G4 process has significantly improved product quality and output over the prior generation. The company has produced its own HBM — behind the leaders, but produced. It has shifted its mix away from mobile DRAM, weak amid soft smartphone shipments, toward server DRAM, which is expected to become its largest segment. These are the decisions of a management team reading the market correctly.

The capacity plan, if executed, changes the company’s position materially. More than doubling to over 600,000 wafers per month, with new fabs online as soon as 2027, would take CXMT past Micron in capacity by 2030 on one source’s estimate. Combined with the 15% share threshold Counterpoint identifies as the requirement for long-term competitiveness, the capacity build is the mechanism by which CXMT could plausibly get there — later than the 2028 forecast implies, but not never.

The IPO itself strengthens the balance sheet at an unusually favorable moment. Raising 57.9 billion yuan, potentially 66.6 billion, at the peak of a pricing cycle, to fund capacity that arrives after the peak, is textbook cyclical capital management. Memory history rewards companies that build through downturns with cash raised during upturns; it punishes those that do the reverse.

And the shortage may run longer than skeptics assume. If the supply gap persists through 2027, as several research firms expect, CXMT will generate extraordinary cash flow for another eighteen months — enough to fund a substantial portion of the capacity build from operations rather than from the IPO proceeds.

The case against

The skeptical reading starts with the same facts and weights them differently.

Every element of CXMT’s current profitability is cyclical. The 46% implied net margin exists because of a supply-demand gap the company did not create and cannot control, opened by competitors reallocating capacity to HBM. When that capacity returns — and CXMT’s own 600,000-wafer expansion is part of what will bring supply back — the same operating leverage that produced these margins will work in reverse. The company’s own record shows what that looks like: a 16.34 billion yuan loss in 2023.

Thirty times annualized peak-cycle earnings is a demanding multiple for a commodity manufacturer that has been profitable for eighteen months and carries 36.65 billion yuan of accumulated losses. It is more demanding still for one whose access to the critical production input — advanced lithography — is controlled by a foreign government responding to pressure from another foreign government.

The technology gap is not closing on the timeline the valuation implies. Two generations behind in HBM, per five Reuters sources, in the product category that captures most of the value in AI memory. Counterpoint forecasts 11% share by 2028 against a 15% competitive threshold. CXMT can buy fabs with IPO proceeds; it cannot buy EUV lithography at any price, and the workarounds YMTC developed took years of forced adaptation that CXMT has not yet undergone.

The float that produced Monday’s gain is also a liability. Lockups begin expiring in twelve months. Between now and then, price discovery operates on 6.73% of the shares. When strategic investors and state vehicles become able to sell into a market that has repriced their holdings by a factor of five and a half, the supply picture changes fundamentally — and the natural buyers for that supply are not obvious.

Regulatory risk is live rather than theoretical. CXMT sits on the 1260H list. Pentagon procurement restrictions took effect June 30. An interagency committee has already approved Entity List addition, held back only by an unresolved policy argument. Two House committee chairs are pressing for an executive order banning U.S. purchases outright. Section 8531 creates a pathway — remote, but real — to a designation that would bar U.S. persons from holding the stock.

And the domestic political economy is less stable than it appears. CXMT raised prices on Huawei and expelled engineers from a Huawei-linked vendor. Chinese electronics firms have complained to the industry ministry that memory price increases are delaying their product launches. The ministry has said it will crack down on hoarding intended to drive up prices. The protected market that underwrites CXMT’s pricing power exists at Beijing’s discretion, and Beijing has other constituents.

What the DRAM industry’s own history says

Theodore Shou’s remark that survivors of the cycle emerge stronger is not a platitude. It is a compressed description of what actually happened to this industry, and the record is more brutal than the phrase suggests.

At its peak in the late 1980s, the DRAM market supported 23 suppliers. By 1995, the top ten accounted for roughly 80% of share. Today three companies hold close to 90%. The consolidation was not achieved through mergers of equals or orderly exits. It was achieved through bankruptcy.

The pattern is consistent enough to be predictive. DRAM is a commodity with enormous fixed costs, long capacity lead times and near-zero product differentiation. When prices are high, every producer builds. Capacity arrives two years later, typically in a cohort, and prices collapse. The producers with the weakest balance sheets and the highest cost structures cannot fund the next process node, fall behind, and are acquired or liquidated. The survivors buy their assets cheaply and emerge with more share than they started with.

Qimonda is the case CXMT knows best, because it inherited the wreckage. Spun out of Infineon in 2006, Qimonda was among the largest DRAM makers in the world when the 2008 financial crisis collapsed memory pricing. It filed for insolvency in January 2009 and was liquidated. Its patents and engineers dispersed — a portion of them, eventually, to Hefei.

Elpida followed. Japan’s last remaining DRAM manufacturer, itself the product of an earlier consolidation of NEC and Hitachi’s memory operations, Elpida was squeezed between falling prices and a strong yen that made its cost base untenable. It filed for bankruptcy protection in February 2012 — at the time the largest bankruptcy in Japanese manufacturing history — and was acquired by Micron, whose completion of the deal in 2013 established the three-player structure that has governed the industry since.

Two observations follow, and they cut in opposite directions for CXMT.

The first is that CXMT is unusually well-insulated against the mechanism that killed Qimonda and Elpida. Both failed because they could not fund the next node when prices collapsed and their access to capital closed. CXMT has state shareholders holding 36.29% who absorbed nine consecutive years of losses without demanding an exit, a protected domestic market, and now 57.9 billion yuan of fresh equity raised at the top of the cycle. Whatever else is true, this is not a company that gets liquidated in the next downturn. Shou’s point about survivors applies to CXMT more comfortably than to almost any producer in the industry’s history.

The second is less comfortable. The reason the surviving three earn extraordinary returns at the top of the cycle is that the industry consolidated to three. A fourth well-capitalized producer that cannot be killed by a downturn — because its losses are underwritten by a provincial government pursuing a strategic objective rather than a return on capital — changes the structure that generates those returns. This is precisely what Monday’s selling in Micron, SK Hynix and SanDisk was pricing.

It also complicates the bull case for CXMT itself. If the historical pattern holds and the next downturn eliminates a competitor, the survivors’ share rises. If CXMT’s presence means nobody is eliminated, the industry ends up with four producers sharing a market that historically supported profitable operations for three. Everyone’s normalized margin is lower, including CXMT’s.

Which of those futures arrives depends less on CXMT’s execution than on how far Beijing is willing to fund losses in pursuit of self-sufficiency, and for how long. That question does not have a market-based answer.

What comparable listings suggest

The STAR Market has produced first-day moves like this before. All 25 companies in the board’s opening batch in July 2019 at least doubled on debut, with the strongest rising as much as 520%. QuantumCTek, an information security firm, recorded a 924% first-day gain in 2020 — still the largest for any Chinese IPO. A 466% debut is remarkable in absolute terms and unremarkable by the standards of a market that suspends price limits for five sessions.

The more apt comparison is SMIC’s own STAR Market listing in July 2020. China’s largest contract chipmaker raised 53.2 billion yuan — the record CXMT has now broken — amid comparable enthusiasm about semiconductor self-sufficiency and comparable state backing. Investors who bought SMIC’s A-shares in the first days of trading experienced a substantially different outcome from those who bought at the offer price. The parallel is imperfect: SMIC listed into a foundry market it did not dominate, while CXMT lists into a memory shortage that has handed it exceptional pricing power. But the structural setup — strategic national champion, constrained float, policy tailwind, technology gap versus global leaders — is close enough to be worth holding in mind.

The third comparison is the one the market has largely ignored. On July 10, seventeen days before CXMT’s debut, SK Hynix raised $26.5 billion through a Nasdaq ADR listing — 177.9 million ADRs at $149, the largest U.S. share sale ever by a foreign company, surpassing Alibaba’s $25 billion in 2014. Orders covered seven times the shares on offer. The ADRs closed their first session at $168.01, above the $149 pricing but below the $170 open.

Two memory manufacturers, three weeks apart, raising a combined $35 billion at the top of the same cycle. SK Hynix will spend its proceeds expanding Korean manufacturing and buying EUV scanners — the machines CXMT cannot obtain. That contrast is the entire investment case for and against CXMT compressed into a single procurement line item.

Material risks

• Cycle reversal. Memory is the most cyclical major segment in semiconductors. CXMT’s margins depend on a supply gap that its own capacity expansion will help close.
• Equipment access. Dependence on ASML DUV systems, with EUV unavailable and further DUV restrictions under active discussion in The Hague and Washington.
• Entity List addition. Approved by a U.S. interagency committee, not yet implemented. Would restrict access to U.S.-origin suppliers, software and tools.
• Securities-level sanctions. Section 8531 of the FY2026 NDAA creates a pathway to NS-CMIC designation, which would bar U.S. persons from trading the stock. No determination has been made.
• Lockup expiry. Free float of 6.73% expands materially as strategic and state holdings unlock from twelve months onward.
• Domestic policy reversal. Pricing power rests substantially on restrictions barring Chinese state-owned enterprises from buying foreign memory, and on guidance to prioritize domestic customers. Both are policy choices.
• Customer concentration. Large multi-year agreements with ByteDance and Tencent concentrate revenue in a small number of domestic buyers whose own capital spending is cyclical.
• Governance and disclosure. State-linked shareholders held 36.29% pre-IPO, and Chinese authorities treat the company as strategic infrastructure. Minority shareholder interests and industrial policy objectives will not always align.
• Accumulated deficit. 36.65 billion yuan of unrecovered losses at end-2025 constrains distributable profits.
• Execution risk on capacity. Two fabs under construction and a third in negotiation, targeting more than 600,000 wafers per month, using a domestic equipment supply chain that is less mature than the international alternative.

What to watch next

Confirmed and scheduled. STAR Market price limits of 20% daily apply from CXMT’s sixth trading session. The over-allotment option, if exercised, would lift proceeds toward 66.6 billion yuan. CXMT’s first-half 2026 results will convert the 50–57 billion yuan profit guidance into audited figures. The Section 805 Product Ban takes effect June 30, 2027. Strategic investor lockups begin releasing twelve months from listing.

Unresolved and consequential. Whether the Commerce Department implements the Entity List addition already approved at interagency level. Whether the administration issues the executive order Moolenaar and Mast have requested. Whether the Dutch government further restricts DUV lithography exports. Whether China’s industry ministry acts on domestic complaints about memory pricing. And whether YMTC’s own IPO proceeds — with some executives internally pushing for a 1 trillion yuan valuation target, according to two people who spoke to Reuters — arrives into a market still willing to pay these multiples.

The datapoint that matters most. DRAM contract pricing for the fourth quarter of 2026 and the first quarter of 2027. CXMT’s entire earnings profile is a function of that number. Everything else is commentary.

Frequently asked questions

What is CXMT?

ChangXin Memory Technologies is China’s largest producer of DRAM — dynamic random-access memory, the short-term memory used in smartphones, PCs, servers and AI systems. Founded in Hefei in 2016, it is the world’s fourth-largest DRAM manufacturer, with roughly 7.7% of the global market in 2025 according to its IPO prospectus.

How much did CXMT rise on its first day of trading?

CXMT closed at 49.00 yuan on July 27, 2026, against an IPO price of 8.66 yuan — a gain of approximately 466%. The intraday high was 49.50 yuan, roughly 471.6% above the offer price.

Why did CXMT stock rise so much on debut?

Three factors combined. Only 6.73% of the enlarged share capital was freely tradable at listing, with strategic and state holdings locked up for twelve to thirty-six months. Retail subscription exceeded 200 times the available tranche, producing an allocation rate near 0.47%. And the STAR Market imposes no daily price limit during a stock’s first five sessions. Scarce supply met concentrated demand in a venue that permits unlimited movement.

Is CXMT now China’s most valuable listed company?

Yes, measured by market capitalization on a mainland Chinese exchange. Monday’s close valued CXMT at approximately 3.3 trillion yuan, ahead of Industrial and Commercial Bank of China at roughly 2.6 trillion yuan. It also exceeds the market capitalizations of Intel and Cisco, though it remains below Samsung Electronics.

How much did the CXMT IPO raise?

At least 57.9 billion yuan, approximately $8.6 billion, rising to as much as 66.6 billion yuan if the over-allotment option is exercised. It is the largest listing in STAR Market history, surpassing SMIC’s 53.2 billion yuan offering in 2020, and mainland China’s largest since Agricultural Bank of China’s 2010 offering.

Is CXMT profitable?

Yes, recently. CXMT reported its first annual profit in 2025 — 1.88 billion yuan on revenue of 61.8 billion yuan — after losses of 8.33 billion yuan in 2022, 16.34 billion yuan in 2023 and 7.14 billion yuan in 2024. The company has guided to first-half 2026 net profit of 50 billion to 57 billion yuan. That guidance has not yet been confirmed by audited results. Accumulated unrecovered losses stood at 36.65 billion yuan at the end of 2025.

Who owns CXMT?

The prospectus states that state-owned shareholders held 36.29% before the IPO, including Anhui Investment Group and the National Integrated Circuit Industry Investment Fund Phase II, known as the Big Fund. Zhu Yiming, founder of GigaDevice Semiconductor, is chairman and is described in filings as central to the company’s creation. Strategic investors in the offering included Alibaba, Tencent, Xiaomi, Meituan, ZTE, Chery and NIO.

How far behind is CXMT technologically?

In conventional DRAM, close enough to compete on price and in some cases charge above Samsung. In high-bandwidth memory — the format AI accelerators require — five sources told Reuters that CXMT remains two generations, or several years, behind Samsung, SK Hynix and Micron. The binding constraint is lithography: China has been unable to obtain ASML’s extreme ultraviolet systems since 2019.

Is CXMT subject to U.S. restrictions?

It is designated on the Pentagon’s Section 1260H list of Chinese military companies, a designation reaffirmed on June 8, 2026, and one the company denies. Pentagon procurement restrictions tied to that list took effect June 30, 2026. CXMT is not currently on the Commerce Department’s Entity List, though Reuters reported that a U.S. interagency committee approved its addition last year and officials have held off. Two House committee chairs have called for an executive order barring U.S. purchases of its memory.

What happens to CXMT’s stock price limits now?

The STAR Market suspends daily price limits for a newly listed stock’s first five trading sessions, applying a 20% daily band from the sixth. Intraday circuit breakers of 10 minutes apply at 30% and 60% moves from the opening price during the initial five-day window.

How does this compare to SK Hynix’s Nasdaq listing?

SK Hynix raised $26.5 billion through American depositary receipts on July 10, 2026 — the largest foreign IPO in U.S. history, priced at $149 per ADR with orders covering seven times the shares on offer. The ADRs closed their first session at $168.01. SK Hynix said it would use proceeds to expand Korean manufacturing and purchase extreme ultraviolet lithography scanners, the equipment CXMT cannot obtain.

What did CXMT’s debut do to Micron and other memory stocks?

Memory and storage names fell on Monday, July 27, while the broader market was mixed. Around 11:07 a.m. Eastern, Micron was down approximately 5% at $871, SK Hynix ADRs off roughly 6% at $145, SanDisk down about 12% at $1,270 and Western Digital about 7% at $483, with the Roundhill Memory ETF off around 4%. Those are intraday figures. The decline followed a similar sell-off on July 15 and 16 when the size of CXMT’s offering became clear, and it landed on top of very large year-to-date gains across the group.

Will CXMT ease the global memory shortage?

Not in the near term. CXMT’s two new fabs in Shanghai and Hefei are not expected to come online before 2027, its output is currently absorbed by domestic demand, and Chinese authorities have asked it to prioritize domestic customers. Its access to advanced chipmaking tools also remains restricted. Kyle Chan of the Brookings Institution identified whether CXMT could help with the broader shortage as one of the open questions raised by the listing.

What is the biggest risk to CXMT’s earnings?

The memory cycle turning. CXMT’s implied first-half 2026 net margin of roughly 46% reflects the widest DRAM supply-demand gap since 2011, produced by competitors shifting capacity toward high-bandwidth memory. As that capacity returns — including through CXMT’s own expansion to more than 600,000 wafers per month — the operating leverage that generated these margins reverses.

Final assessment

CXMT’s debut is being read as a verdict on China’s semiconductor ambitions. It is better understood as two separate events that happened to occur on the same morning.

The first is a genuine industrial achievement. A company founded a decade ago on the salvage of a bankrupt German manufacturer is now the world’s fourth-largest DRAM producer, charging more than Samsung for comparable server modules, signing ten-figure multi-year contracts with China’s largest technology firms, and — as of June — powerful enough to expel a Huawei-affiliated vendor’s engineers from its own cleanrooms over a pricing dispute. That did not happen because of the IPO and does not depend on the share price. It is the most substantive evidence to date that Chinese industrial policy can produce a competitive manufacturer in a capital-intensive sector previously closed to it.

The second is a price discovery event on 6.73% of a company’s shares, in a venue with no daily price limit, after an offering that filled fewer than one in two hundred retail applications. It moved the valuation from roughly five and a half times annualized peak-cycle earnings to roughly thirty times, without any new information about the business emerging between the bookbuild and the open. Both multiples are unreliable, because annualizing peak earnings for a memory manufacturer is a category error. But the direction of travel says something: the market repriced from assuming substantial normalization to assuming very little.

What the evidence supports is roughly Theodore Shou’s position. The business is likely to become a global leader in DRAM over a long enough horizon, given its capacity plans, its protected domestic market and the state financing behind it. The current margins will not survive the cycle. Those two statements are compatible, and holding both is the discipline the situation requires.

What remains uncertain is concentrated in two places, and neither is the share price. The first is lithography: CXMT can spend 66 billion yuan on fabs, but the gap in high-bandwidth memory closes only through equipment it cannot currently buy, and YMTC’s forced adaptation to Entity List restrictions suggests the workarounds take years. The second is Washington, where an interagency committee has already approved a restriction that officials have not implemented, and where Apple and Micron are arguing opposite sides of a question the administration has not resolved.

For readers watching from the United States, the number that will matter most over the next six months is not CXMT’s share price. It is DRAM contract pricing for the fourth quarter of 2026 and the first quarter of 2027, and whether the Commerce Department acts on the Entity List determination sitting in its inbox. Those two variables will decide considerably more about this company than Monday’s session did.

This article is provided for general informational purposes and does not constitute financial, investment, tax, or legal advice.

Sources

• CXMT IPO prospectus, filed with the Shanghai Stock Exchange, July 22, 2026
• Reuters, “China’s memory chip makers ride AI boom to new power — and U.S. scrutiny,” by Fanny Potkin, July 24, 2026
• Reuters, “Explainer: What is CXMT and how did it become China’s DRAM champion?” by Eduardo Baptista, July 15, 2026
• TrendForce, “CXMT’s 471% STAR Debut Makes It China’s Top Listed Firm,” July 27, 2026
• Fortune and the Associated Press, “A Chinese chip maker’s shares surged 466% in their first day of trading,” by Chan Ho-Him, July 27, 2026
• CNBC, “Chipmaker CXMT’s 466% market debut surge makes it the most valuable China-listed company,” July 27, 2026
• U.S. Department of Defense, Section 1260H list of Chinese military companies, June 8, 2026
• WilmerHale client alert, “Pentagon Adds 65 New Entities to the 1260H List of Chinese Military Companies,” June 12, 2026
• Shanghai Stock Exchange, trading mechanism and STAR Market rules
• TechCrunch, “SK Hynix raises $26.5B in the biggest foreign IPO in US history,” July 10, 2026
• The Korea Herald, “SK hynix sets foreign IPO record with $26.5b Nasdaq offering”
• Forbes, “CXMT Chairman Zhu Yiming Amasses $15.9 Billion Fortune,” by Yue Wang, July 27, 2026
• CnEVPost, “Nio appears on strategic investor list for memory giant CXMT’s blockbuster IPO,” July 15, 2026
• Nikkei Asia, “China’s CXMT jumps 471% on debut, riding AI memory boom to record IPO”
• 24/7 Wall St., “SanDisk Sinks 12%, Micron Drops 5%, SK Hynix Falls 8% as China’s CXMT IPO Rattles Memory Stocks,” by David Moadel, July 27, 2026
• Computerworld, “DRAM maker Elpida files for bankruptcy,” February 2012
• McKinsey & Company, “Memory: Are challenges ahead?” on DRAM industry consolidation and market structure

Affiliate disclosure: Businessfinance.news may earn compensation from qualifying actions completed through selected links on this website, at no additional cost to the reader. Affiliate relationships do not influence our editorial reporting, analysis, or conclusions.