In mid-July 2026, two pieces of news about China's memory-chip industry surfaced one after another on the same day, sketching the portrait of an industry approaching a tipping point: ChangXin Memory Technologies (CXMT) is projected to match Micron in DRAM capacity by year-end, while the company has formally launched a listing on Shanghai's STAR Market, seeking to raise roughly ¥57.9 billion at a price-to-earnings ratio of 308.92. Over the following fortnight the story ran its course: a 471% first-day surge made CXMT the largest company on the A-share market by capitalization, Asian chip stocks plunged in sympathy, a bipartisan group in the US Congress demanded a national-security investigation, and the Commerce Department fortified its own semiconductor lineup with an "equity for subsidies" play — a single Chinese company's IPO became the physical result of a deep restructuring of the global semiconductor pricing system. In early August the story kept advancing: CXMT refused Apple's price-cut demand and began to set the price floor for commodity DRAM, Nvidia halved the HBM allocation on its Rubin Ultra and pulled SK hynix down nearly 20% in a single session, and Apple moved from preliminary talks to formal testing of Chinese memory in iPhone and MacBook lines — market forces were, in effect, endorsing CXMT's expansion.
Catching Up on Capacity: From Second-Tier Bit Player to Micron's Peer
According to research by analysts Zephyr and Citrini, CXMT's DRAM capacity is set to grow by 600,000 to 1.1 million WPM (wafers per month), bringing total capacity to roughly 950,000–1.45 million WPM. That projection includes the 350,000 WPM CXMT is expected to reach by the end of 2026 — and it is precisely this number that almost exactly overlaps with the 375,000 WPM Micron is projected to hit this year.
A Chinese memory maker will catch Micron in DRAM capacity by the end of 2026 — this is not a long-range forecast but a baseline estimate issued by a Western industry research house in mid-2026. A gap in overall capacity remains: as the world's third-largest DRAM producer, Micron's portfolio spans multiple generations, while CXMT's current lineup is still dominated by DDR4 and LPDDR4. But in terms of magnitude, the Chinese firm has moved from an irrelevant second-tier player to a capacity heavyweight that can stand alongside the industry's number three.
Several structural drivers lie behind the capacity explosion. The analysts note in particular that cleanroom construction cycles in China run about 12 months, versus 21 to 24 months in the rest of the world — meaning the same amount of investment translates into roughly twice the expansion speed in China as in the West. That speed advantage is not natural: it rests on sustained capital support from the Chinese government and banking system, a mature engineering and construction supply chain, and a labor pool that can be rapidly mobilized for factory-building at scale.
Supply-Chain Acceptance: From Substitute to Standard Option
A change more telling than the capacity numbers themselves is happening on the supply-chain side. First-tier motherboard makers such as MSI and ASUS are adapting to CXMT's memory chips through BIOS updates, while brand owners including Corsair and Lexar are already using — or planning to use — CXMT DRAM in their products for the Chinese market.
The signal value of this shift is clear: the willingness of top international brands to run compatibility validation on CXMT's products means that Chinese DRAM has passed the supplier-admission standards of Western brand owners on quality, stability, and supply reliability. In the semiconductor supply chain, the distance from "substitutable" to "validated" is often longer and harder to cross than the R&D breakthrough from zero to one.
The BIOS updates from MSI and ASUS are not a technical decision but a market decision. When first-tier motherboard vendors are willing to commit engineering resources to validating Chinese memory chips, it means their customers — OEM system builders and retail consumers — have developed real demand for this supplier's presence.
The Scale of the IPO: Market Signals From ¥57.9 Billion and a 308× P/E
The IPO announcement of ChangXin Technology (CXMT's parent company, ticker 688825.SH) shows an offer price of ¥8.66 per share, an initial offering of 669 million shares, and projected proceeds of roughly ¥57.919 billion — corresponding to a P/E ratio after excluding non-recurring items of 308.92, far above the industry average of 76.32 and the 134.62 of comparable companies.
A set of comparisons helps decode what this pricing implies. One of the most closely watched IPOs in China's semiconductor sector in 2024 — Hua Hong Semiconductor — raised about ¥21.2 billion, and its P/E was already considered high by industry standards. CXMT's raise is 2.7 times Hua Hong's, and its roughly 307× P/E implies that in pricing this company, the capital market is not betting on current profitability but on "gusher-like" growth expectations over the next three to five years.
What does a 308× P/E actually mean? By the simplest valuation model, it implies the market expects CXMT's net profit to reach 300 times its current level within a few years. DRAM is a commodity-grade semiconductor sector with strong cyclicality and violent price swings — industry-wide losses during downturns are not unusual. The risk embedded in an excessive valuation is this: should the capacity-expansion plan face delays, should yield ramp-up fall short, or should the global DRAM market enter a down-cycle in prices, secondary-market investors will face volatility far beyond the industry norm.
The Last Step From Capacity to Market
Read together, the most striking thing about the two stories is not the message each carries on its own, but the tension between them.
On one hand, the rapid expansion of capacity is the foundation of the whole story. Monthly wafer capacity of 950,000 to 1.45 million WPM means CXMT's physical output in DRAM is already close to — or at — the level of the world's top three. The compatibility validation by MSI and ASUS, meanwhile, signals that the downstream supply chain has established confidence in the end products of that output.
On the other hand, the 308× P/E handed out by the IPO market is a financial signal almost decoupled from the physical expansion of capacity — it is betting not on whether CXMT "can do it," but on whether CXMT "can do it and scale the profits."
When an industry's capacity figures and capital-market valuation leap by this magnitude at the same time, it usually means the industry has moved past the hardest "zero-to-one" phase and entered the "one-to-N" phase of scaling. The problem is that the scaling phase has its own traps — the attrition of price wars, the urgency of technology iteration, and rising global trade barriers. For CXMT, the stretch from "catching Micron" to "surpassing Micron" is about more than a capacity number.
A Potential Shift in the Competitive Landscape
Micron will reach roughly 375,000 WPM of capacity by the end of 2026, putting it on the same order of magnitude as CXMT. But widen the lens to the DRAM industry as a whole, and Micron is the world's third-largest DRAM maker, while Samsung and SK hynix command roughly 40% and 30% of the market respectively.
"That component makers are running compatibility validation on their products against Chinese-made memory is, in fact, not surprising" — this phrasing betrays a more fundamental cognitive shift: as Chinese DRAM enters the global mainstream supply system, the world's supply chains are moving from the probing stage of "whether to accept Chinese DRAM" to the adaptive stage of "how to make use of Chinese DRAM."
For the global DRAM industry, the rise of Chinese capacity means that the stable oligopoly held by Samsung, SK hynix, and Micron for the past two decades is being broken by a new entrant unbound by the existing rules of competition. Historically, every capacity expansion in the DRAM industry has come with price crashes and consolidation — but this time, China's structural advantages (construction cycles half as long, state-capital support, and the scale of its downstream market) may change the rules of the game.
The IPO Lands: Investor Fervor and a Leap in Valuation Multiples
According to Reuters, CXMT's IPO was priced at ¥8.66 per share, corresponding to an IPO valuation of ¥579.18 billion. Subscription for the offering was open for just one day, on July 17, with a listing planned in Shanghai on July 27.
Wu Zhou, a fund manager at Shenzhen Deyuan Investment, plans to subscribe to the new shares, betting that CXMT's post-listing valuation "will very likely exceed ¥3 trillion ($443.33 billion) … and may even reach ¥5 trillion." That expectation rests on comparisons with trillion-dollar-class companies like Micron, Samsung, and SK hynix. Tan Zhiqiang, chief investment officer of Hong Kong Central Asset Investment, said that although China lags by two to four years in DRAM and HBM technology, CXMT's IPO valuation "looks very cheap," and he expects shares to "surge several-fold" on the first trading day.
If the over-allotment option is exercised in full, total proceeds will rise to roughly ¥66.6 billion — a figure that exceeds the ¥57.9 billion preliminary estimate in the earlier prospectus. Market demand has outrun the underwriters' initial pricing range.
A Leap in the Financials
The latest financials disclosed in the prospectus explain the valuation logic more viscerally. CXMT's first-quarter revenue reached ¥50.8 billion, up 700% year-on-year; net profit was ¥25 billion, reversing a ¥1.6 billion loss in the same period last year. Fund manager Wu Zhou projects that CXMT's profit this year will reach ¥100 billion.
The logic behind a 300× P/E rests on one premise: current profits are merely a starting point. When an explosion of demand from driverless cars, robotics, and AI PCs meets the wave of China's semiconductor self-sufficiency drive, the growth curve could make today's valuation look "cheap" — but that depends on global DRAM demand avoiding a cyclical downturn, and on China's domestic market being able to absorb CXMT's incremental capacity.
Overlay With China's Semiconductor Self-Sufficiency Strategy
CXMT's IPO is the largest ever by a Chinese chipmaker, reflecting Beijing's latest effort to pour capital into a strategic industry it deems vital in its competition with Washington. ChangXin is the world's fourth-largest DRAM manufacturer and China's only mass-producer of DRAM at scale.
One instructive comparison: SMIC, China's largest contract chipmaker, listed in Shanghai in July 2020, and its stock price nearly halved from the offer price just over two months after the debut. Whether CXMT's sky-high valuation will undergo a similar "post-listing reversion to value" depends on the real-world speed at which its capacity expansion materializes — the most fragile link in the current valuation narrative.
The Hefei Multiplier — A Patient-Capital Case Study Behind the CXMT IPO
Reuters' deep-dive analysis of July 23 approached CXMT's story from a different set of numbers: not capacity or market share, but the city behind the chip giant and its investment logic.
CXMT was founded in 2016 by an investment vehicle affiliated with the Hefei Economic and Technological Development Zone, with RMB 10 million (about $1.5 million) in seed money. That starting sum looks almost comically small today — for a semiconductor manufacturer, RMB 10 million would not even buy a single second-hand lithography machine. But it was, in fact, CXMT's starting point.
Nearly a decade later, investors linked to the Hefei municipal government hold 36.8% of CXMT, making them its largest shareholder group. Priced at the IPO valuation, that stake is worth roughly RMB 213 billion ($31.5 billion). A simple comparison: that figure is more than twice Hefei's total fiscal revenue for all of 2025, and equivalent to 15% of its economic output.
Initial investment in 2016: RMB 10 million; value of the stake at the 2026 IPO: RMB 213 billion (roughly 21,300 times the initial investment); Hefei's shareholding: 36.8%; coverage of Hefei's fiscal revenue: more than twofold; share of Hefei's economic output: 15%.
Cornell University professor Eswar Prasad offered a sober assessment: "This IPO fuses multiple government objectives, including a vigorous push for artificial intelligence and advanced technology… and the development of capital markets. AI and advanced technology can of course raise productivity and output growth, but they are unlikely to lift household income and employment growth on a broad scale." Christopher Beddor, deputy China research director at Gavekal Dragonomics, gave a more concise peer assessment from the local-government angle: "This is a major deal for Hefei. The city is known for its aggressive industrial policy, and CXMT is its single biggest bet. The local government has supported the company through years of losses, and now stands to reap a substantial return."
Founder Zhu Yiming graduated from Tsinghua University, studied and worked in the United States, and later returned home to build China's memory-chip industry. Priced at the IPO, his shares are worth nearly RMB 14 billion, and he has committed part of them to employee incentives.
The macroeconomist's question. Standard Bank economist Jeremy Stevens's remark supplied a footnote to this "success narrative": "This is exactly the heart of the current macroeconomic puzzle — the economy has a growth engine, but that engine is driving only export and corporate-profit growth, not the wage growth needed to sustain domestic demand." CXMT's IPO generates enormous paper wealth, but most of that wealth will flow back into the next generation of strategic industries (through reinvestment by Hefei's state capital) rather than being distributed to households — a fundamental structural difference from the way American venture capital transmits to consumption.
Under the Microscope on IPO Pricing — The ¥8.66 Offer Price and a Valuation Ladder From ¥1 Trillion to ¥4.25 Trillion
The pricing-results announcement ChangXin Technology disclosed on the evening of July 21, along with research notes published afterward by several brokerages, supplied finer-grained numerical coordinates to the IPO narrative already in place.
The anchoring logic of the offer price. ChangXin's offer price of ¥8.66 per share corresponds to a market capitalization on the order of roughly ¥400 billion on a fully issued share base — an unremarkable absolute number on its face, but set against its 2025 full-year net profit attributable to the parent of ¥1.875 billion and a first-quarter 2026 net profit of ¥24.762 billion, the leap in growth shifts the core variable underpinning the valuation from "past profits" to "future profit streams."
Guojin Securities drew a more precise capacity line in its research note: ChangXin's monthly capacity at the end of 2025 was roughly 280,000–290,000 12-inch wafers. The significance of this figure is that it simultaneously catches up with and partly exceeds the capacity-ramp curve the industry had expected — in an environment where equipment-import controls have not yet been fully eased, this pace of capacity growth is itself an independent technical signal.
Four scenarios and one contested curve. Guotou Securities laid out four valuation scenarios in its research note: conservative ¥1 trillion, neutral ¥1.5 trillion, optimistic ¥2.3 trillion, and hyper-optimistic ¥4.25 trillion. Huaxi Securities, meanwhile, offered a more holistic range: a stable market value of ¥2 trillion to ¥3 trillion under neutral expectations, and up to ¥4 trillion even in an optimistic scenario. The two sources agree on direction and magnitude; their divergence lies mainly in the definition of a "stable market value" — Guotou's "conservative" case already far exceeds the lower bound of Huaxi's "neutral" case, which itself shows that sell-side valuation expectations for ChangXin are nowhere near converged.
Against the ¥8.66 offer price and a ¥1 trillion–¥4.25 trillion market-value range, the first-day gain lands in a 70%–600% band, with a profit of roughly ¥3,000 to ¥26,000 per lot (500 shares). This calculation presupposes that the first-day gain is driven entirely by the valuation multiple — in actual trading, turnover rate, market sentiment, and the lock-up periods on institutional placements all affect first-day pricing.
Signals in the subscription data. Online investors gave up 6.5862 million shares and offline investors gave up 31,567 shares, all of which were underwritten by the joint lead underwriters. That proportion is not high for a large A-share IPO, but combined with ChangXin's history of cumulative losses exceeding ¥30 billion over three consecutive years (roughly ¥31.8 billion in total losses across 2022–2024), retail investors' hesitation points to one thing: in ChangXin's narrative, the certainty of "technological breakthrough" and "market share" far exceeds the certainty of "earnings sustainability." Whether the underwritten portion flows out quickly after listing will be a window into the market's true confidence in ChangXin.
The first-half earnings pre-announcement. ChangXin projects first-half 2026 revenue of ¥110 billion to ¥120 billion (up 612%–677% year-on-year) and net profit attributable to the parent of ¥50 billion to ¥57 billion (up 2,244%–2,544%). Extrapolated to the full year, profit is on track to wipe out all cumulative losses since the company's founding — meaning ChangXin has gone from "a company that lost money for a decade" to "a company that earned back a decade of losses in a single year." The turning point itself is fact, but the slope of that turning point — from an annual loss of ¥8 billion to a half-year profit of ¥50 billion — reflects both the double effect of a DRAM price recovery and capacity release, and a base effect in capacity utilization. Whether this slope can be sustained over the coming quarters depends on the direction in which the global economic cycle pulls DRAM demand.
The Number Narrative of Listing Day — A 471% Surge and the New ¥3.3 Trillion King of A-Shares
At 9:30 a.m. on July 27, ChangXin Technology (688825.SH) officially began trading on the Shanghai Stock Exchange's STAR Market. It opened at ¥49.50, up 471.59% from the ¥8.66 offer price, corresponding to a total market value of ¥3.31 trillion — surpassing ICBC, Moutai, and CATL to claim the top spot on the A-share market by capitalization.
Verification in the trading data. Of the four valuation scenarios brokerages had laid out earlier, the conservative ¥1 trillion and neutral ¥1.5 trillion were blown past at the open; the optimistic ¥2.3 trillion was also exceeded; and the hyper-optimistic ¥4.25 trillion had not been touched but was already closing in. The 471% first-day gain shows that the market not only accepted ChangXin's valuation narrative but chose the optimistic end of the scenarios — echoing CXMT's first-quarter revenue of ¥50.8 billion (up 719% year-on-year) and net profit of ¥24.762 billion, both leaps in growth.
By the midday close, ChangXin Technology's turnover had broken ¥120 billion, surpassing the ¥90.038 billion record set by East Money in October 2024 to become the new single-day turnover champion among A-share individual stocks. The turnover rate was 58.06%, indicating that roughly six-tenths of the tradable shares changed hands on the morning of listing day — a figure that is itself on the high side for a large STAR Market IPO, but given that there were no price limits on the first day and only 6.73% of total shares were tradable, congestion in the trading channel and institutions' demand to build positions jointly pushed up trading density.
Cross-verification of the capacity data. The first-half 2026 earnings pre-announcement ChangXin Technology disclosed in its listing announcement corroborates the capacity projections in earlier industry research notes: half-year revenue of ¥110 billion to ¥120 billion (up 612%–677% year-on-year), and net profit of ¥50 billion to ¥57 billion. The core significance of these numbers lies not in the absolute volume but in the fact that they mean ChangXin essentially wiped out its three consecutive years of cumulative losses in a single half-year — the turning point at which capacity utilization leapt from the loss line into the profit zone has been confirmed.
ChangXin's monthly capacity is projected to reach 350,000 WPM by the end of 2026, closing in on Micron (375,000 WPM). But whether the US White House will place CXMT on the Entity List remains the biggest variable in its capacity ramp — ChangXin has already been added to the US Department of Defense's "1260H" list, and members of the US House of Representatives have jointly demanded that the Commerce Department bar American firms from buying its chips and add it to the Entity List. After the euphoria of listing day, the sanctions risk has not gone away.
The employee's perspective. On the eve of the listing, one ChangXin Technology employee said something worth recording to a media camera: "Stay level-headed. Whatever people say online, it doesn't really have much to do with us ordinary employees — we care more about whether our future salaries and benefits will change." This is a parallel reality beyond the capital-market carnival data: for the employees inside, the IPO is not a wealth-creation myth but a question mark over the certainty of future pay.
Signals on the board. During the opening session, the share price fluctuated around ¥49.50, without the rapid pullback pattern of "opening at the peak" seen in some recent STAR Market listings — a sign that sustained buying power is supporting the market. But a structural issue deserves attention: calculated on the post-issue total of 66.881 billion shares, only 6.73% of shares (about 4.5 billion) were tradable on the first day, and the unlocking pressure from a large volume of restricted shares will be released gradually in subsequent years. The high turnover on listing day is, in a sense, a form of "restricted-flow damming" — trading volume is concentrated on an extremely small proportion of tradable chips, and does not equate to a market-wide consensus on value.
The Day After the IPO — Asian Chip Stocks Plunge and CXMT's Market Shockwave
On July 28, the day after CXMT's listing, Asian semiconductor stocks suffered a collective rout. Reuters' analysis attributed it to three forces pressing at once.
Korean stocks lead the fall — the inertia of eight circuit breakers. Samsung Electronics' shares closed down 13.4%, their biggest single-day drop in nearly two decades. SK hynix closed down 14.7%. Together the two stocks make up nearly half of the KOSPI index's weighting, and the KOSPI itself closed down 10.8%, triggering circuit breakers for the eighth time this year. The chain reaction in Japan and Taiwan was just as severe: Kioxia Holdings fell 18.3%, and MediaTek fell nearly 10%.
Three layers of pressure stacked together. Reuters' analysis pinned the day's rout on three narratives unfolding simultaneously.
The first layer: a reshaping of the narrative around China's technological catch-up. Reports emerged that Chinese firms are developing domestic deep-ultraviolet (DUV) lithography equipment — the news itself is thin on detail (which firm, what performance, when it can be mass-produced), but at a moment when the investment logic for semiconductor stocks was already soft, its emotional impact far outweighed its practical effect. Kiwoom Securities analyst Han Ji-young said bluntly that although the details had not been disclosed, the news was enough to "reignite market concerns about Chinese memory-chip makers accelerating their capacity expansion."
The second layer: spreading doubt about AI-infrastructure financing. The Wall Street Journal reported that Nvidia may provide roughly $250 billion in financial support for OpenAI's data-center projects, sending Nvidia's shares down nearly 5%. The market began to scrutinize in earnest that Nvidia's role is changing — from chip supplier to capital intermediary — and that "circular financing," should AI profit growth fall short of expectations, carries amplified risk across the entire funding chain.
The third layer: low-cost AI models challenging the "infinite compute demand" narrative. The growing popularity of low-cost Chinese open-source AI models such as Kimi K3 has introduced a question the market had previously overlooked: if future AI workloads demand less in the way of computing resources than imagined, then demand for advanced AI chips and HBM will also fall accordingly. Once that "infinite compute demand" narrative is loosened, the entire valuation logic of semiconductors needs to be rewritten.
CXMT's role. Among these three forces, CXMT is not the only variable, but its presence has changed the behavior of every participant. Reuters' report noted in particular that Apple has been lobbying the Trump administration to allow the use of China-made chips in some of its products — market forces pushing back against the logic of the Entity List, and also the "on-the-field" effect that CXMT's listing brings: when a Chinese competitor truly enters the global market, every pricing, capacity, and procurement decision has to take it into account.
Cameron Systermans, head of Asian multi-asset at Mercer Investments, commented: "This news poses a threat to Asian chip makers that is more long-term in nature. CXMT is indeed a rising competitor, but in commodity DRAM — though in HBM it may still lag its Korean rivals by years."
The day's rout was not a direct result of CXMT's listing, but it revealed a market perception taking shape: CXMT is not a story but a real variable now cutting into the global DRAM pricing system. When its capacity approaches Micron's, its market value tops the A-share market, and its chips are evaluated by Apple as a potential procurement target, the market's price-discovery mechanism is merely catching up with changes that have already happened.
A Ticking Time Bomb — The US Congress's National-Security Investigation Offensive Against CXMT
Less than 48 hours after CXMT's listing, Washington's reaction went from a question to a ticking time bomb.
According to The Washington Post and the New York Post, a bipartisan group of at least six members of Congress plans to send a formal letter to the Trump administration in the coming days, demanding an official national-security investigation into ChangXin Memory (CXMT). The congressional leaders of the House and Senate known as the "Gang of Eight," meanwhile, have called for an emergency classified briefing on CXMT's ties to the Chinese government.
A revision emanating from Washington is underway: in the eyes of the Chinese market, CXMT's IPO was a successful "monetization of patient capital"; in Washington's eyes, it is "the result of the Chinese Communist Party's involvement."
The Rarity of the Bipartisan Consensus
In a US Congress highly fractured in the Trump era, skepticism toward CXMT crosses nearly every political divide. One senior US official told The Washington Post that "people genuinely suspect that the Chinese Communist Party intervened in this so-called IPO," and just six weeks earlier the Pentagon had added CXMT to its 1260H list of military companies. Another senior official was more blunt: CXMT has innumerable links to "China's military-industrial complex," and its rapid market expansion is "a ticking time bomb for national security."
Forty-two hours earlier, CXMT's 471% first-day surge in Shanghai had A-share investors celebrating. Forty-two hours later, the same number had bipartisan lawmakers in Washington sitting down at the same table. This is no coincidence — it is two sides of one signal.
An Imported "Time Bomb" or an Exported "Self-Fulfilling Prophecy"?
The core argument driving the investigation is not what CXMT has done, but what it could be used to do: CXMT's DRAM chips are embedded in consumer electronics, electric vehicles, and data centers — if Beijing has the ability to make those chips "shut down" or "report backdoor data" in a moment of crisis, then the supply chains of the US military and critical infrastructure are exposed to uncontrollable risk.
This argument has a logical gap: to date, there is no public evidence that CXMT's chips contain backdoors or supply-chain manipulation by design. Washington's worry rests on an assumption about "latent capability" — if a competitor's chip market share reaches a high enough level, it acquires the ability to convert commercial advantage into strategic leverage. This is not an investigation of what has already happened but a prevention of what might happen. And "preventive sanctions" are themselves a kind of self-fulfilling prophecy — once an investigation is launched, CXMT's American customers (including the procurement plan Apple is lobbying for) will automatically seek alternatives, thereby confirming the charge of "supply-chain risk."
Apple is lobbying the Trump administration to allow it to buy chips from CXMT (driven by market forces), while Congress demands an investigation of the same company (driven by security forces). In CXMT, the logics of market and security are colliding — Apple wants the chips, Congress wants control. Neither demand is right or wrong, but they are hard to satisfy simultaneously within the policy framework of a single administration.
A Fast Lane Into the NDAA
The proposal to push for a formal investigation is reportedly likely to be folded into the National Defense Authorization Act (NDAA) — the legislation that must pass every year, which determines the Pentagon's budget and defense strategy and is Washington's principal legislative weapon against China's rise. This means CXMT's investigation may not stop at the stage of internal deliberation within the executive branch but will gain institutional standing at the level of congressional legislation — once written into the NDAA, the investigation becomes a legal obligation rather than a temporary action.
This stands in contrast to CXMT's current legal status: added to the 1260H list on June 8 (which affects only Defense Department procurement decisions), but not yet placed on the Entity List (which restricts all US technology exports). An escalation in Congress's demand for an investigation could mean an institutional leap from a "warning label" to a "restrictive sanction" — a path that, once started, is harder to reverse than export controls.
CXMT's Position on Washington's Chessboard
The tags at the end of Lingshi Xiantan's post say it all: #CXMT IPO Alarms US Congress#. Not "China's response," not "diplomatic protest" — the reaction of the US Congress is itself the news. In the US–China tech contest of the past few years, few issues have prompted bipartisan lawmakers to co-sign across party lines within days to demand a security investigation; CXMT's IPO has become the newest member of that list.
The heart of the matter may not lie in CXMT itself, but in the stage at which the whole narrative now sits: when one Chinese company's IPO can trigger a drop in major US indexes, a collective circuit-breaking of the Asian chip sector, an urgent bipartisan congressional demand for a security investigation, and behind-the-scenes lobbying by Apple and Tesla to buy its products — that density of signals has already exceeded any single company's commercial story. It points to the deep restructuring the global semiconductor pricing system is undergoing. CXMT is not the cause of the restructuring; it is its physical result — and all of Washington's reactions are merely catching up with that result.
The Post-Listing Roller Coaster — CXMT's AI Dividend and Washington's Equity-for-Subsidies
On July 31, CXMT's story entered a new chapter. Two reports pieced together happen to be two sides of the same coin: one is the market's feverish embrace of this Chinese DRAM maker, the other is Washington reinforcing its own semiconductor lineup with a new play — "equity in exchange for subsidies."
A Decade of Burning $5 Billion, Earned Back in a Single Quarter
Financial Times data traces CXMT's turnaround curve: cumulative losses of roughly $5 billion over the past decade, yet thanks to the explosive demand for DRAM from AI data centers, its revenue grew 700% between 2025 and 2026, reaching $7.5 billion in the first quarter of 2026 alone. Listed on the Shanghai Stock Exchange on July 27 at an offer price of ¥8.66, shares closed the first day soaring to ¥49, with market value at one point reaching $80 billion — a price-to-earnings ratio exceeding 1,600, yet Nomura still set a target price of ¥116, implying roughly 135% upside.
Counterpoint data shows CXMT's global memory-chip market share grew from 3% in 2025 to 8%. That still lags the DRAM big three, but analysts believe that as long as CXMT can capture "at least one-sixth of the DRAM market," it has a chance of joining the top ranks.
Reuters notes that US export controls restrict CXMT's access to the equipment needed to produce HBM (high-bandwidth memory) chips. But Counterpoint argues that this restriction may actually become an advantage — unable to rely on ready-made high-performance equipment, CXMT is forced onto a more innovative path in R&D. As its research vice president put it: "Ironically, restricting CXMT's development may actually help it overtake incumbents that delay innovation to protect the revenue from their existing equipment." This is structurally identical to the export-controls-as-pressuring logic on the "China AI's Sputnik moment" page.
Apple's FOMO and Congress's Countermove
CXMT's rise has trapped Apple in "fear of missing out": Bloomberg reported that Apple is reportedly in talks with Treasury Secretary Bessent, hoping to buy memory chips from CXMT and YMTC to ease the pressure of rising memory prices — Apple had just raised iPad and MacBook prices because of tight memory-chip supply. But Congress's reaction is the opposite: on July 16, House Select Committee on China chairman Moolenaar and Representative Whitesides wrote to Commerce Secretary Lutnick, recommending that CXMT be added to the Entity List and that an executive order bar US firms from buying DRAM from CXMT and YMTC; bipartisan senators also co-signed a letter to the outgoing Tim Cook, urging Apple to steer clear of Chinese memory makers.
This is of a piece with the "Gang of Eight's" earlier demand for a national-security investigation — except now there is a concrete target: Apple's procurement intentions have given the "security risk" narrative a living case study.
Washington's Other Hand: Equity for Subsidies
The same week, the US Commerce Department quietly announced $874 million in federal funding to seven semiconductor companies in exchange for a minority, non-controlling equity stake in each — including GlobalFoundries (up to $300 million), AI storage-and-logic technology firm Kepler ($245 million), and Multibeam ($140 million). This is another "quasi-nationalization" move after Intel's 10% stake (in exchange for roughly $11 billion in subsidies); last December it also bought a $150 million stake in laser-technology startup xLight.
"As a condition of receiving funds to enhance returns for American taxpayers, the department will take a minority, non-controlling equity stake in each company."
Lingshi Xiantan's annotation to this news was just three words: "Nationalization…" — Senator Todd Young (a drafter of the CHIPS Act) once said the law's original intent was never for the federal government to hold company shares; today it has become routine practice. Placing CXMT's market frenzy and America's equity-for-subsidies in the same week: on one side the market is pricing in China's capacity leap, on the other the government is directly taking equity stakes in America's capacity rebuild — both paths are being walked, just by different routes.
The Reversal of Bargaining Power — CXMT Refuses Apple's Price-Cut Demand (Incremental Addition 2026-08-06)
A week later, the story turned again: CXMT refused Apple's request for a price cut in their negotiations. According to South Korean outlet Digital Daily, Apple had hoped to source chips from CXMT and had reportedly lobbied the US side for permission to procure those products for its overseas-sold devices — but CXMT quoted prices at parity with — or even higher than — Samsung and SK hynix.
The root of the bargaining-power reversal lies in demand-side support at home. US sanctions against China prompted Huawei, Xiaomi and others to lock in CXMT DRAM capacity ahead of time — the same kind of high-priced take-or-pay contracts Korean firms use. With signed domestic orders in hand, CXMT can cite existing agreements to refuse Apple's pressure for cuts, and the higher prices break the device makers' long-standing procurement playbook of using cheap Chinese components as leverage against other suppliers.
The same round of bargaining shifted the position of the Korean suppliers too. Samsung and SK hynix no longer need to promise low-priced commodity DRAM to Chinese customers and can concentrate on the high-priced HBM needed for AI infrastructure. According to industry sources, CXMT is effectively setting the price floor for the commodity DRAM market.
Tying this thread back to the earlier sections of the page: the earlier passage described Apple's FOMO (the July 16 congressional letters urging CXMT be added to the Entity List and senators pressing Cook to keep Apple away from Chinese memory makers) — that was Washington's fear. Today's entry is the market's mirror image: Apple wants to buy, CXMT is in no rush to sell cheap. The "Chinese makers can only compete on price" assumption embedded in the sanction narrative simply does not survive this negotiation.
The earlier sections recorded Washington's fear of Apple's procurement intent and its countermeasures; this section is the sequel — CXMT uses locked-in domestic orders to back its bargaining position, refuses Apple's pressure, and begins to set the price floor for commodity DRAM.
Huang's Earthquake for Hynix — HBM Cut in Half and CXMT's Unintended Window (Incremental Addition 2026-08-09)
On August 9, a configuration adjustment on Nvidia's next flagship GPU, Rubin Ultra, triggered a directed quake in the memory market. According to multiple sources, the HBM allocation on Rubin Ultra dropped from the originally planned 12-Hi stack of roughly 384 GB to 8-Hi 192 GB — a cut of nearly 50% — while the GPU chiplets shrank from four to two and power consumption was trimmed accordingly. Nvidia is still evaluating combinations such as HBM4e 8-Hi and HBM4 12-Hi, with the final specification to be locked down after verification in the second half of 2026. SK hynix, the most HBM-dependent of the three, saw its share price plunge 19% in a single day.
The mechanism is direct: HBM is the highest-margin category in memory chips, and SK hynix — with around 70% share — is the biggest beneficiary of the current super-cycle. Sell-side estimates put Rubin Ultra at roughly 20% of global HBM demand in 2027; halving the VRAM is equivalent to roughly a 10% reduction in global HBM volume — and the cut lands precisely on the most expensive top-end demand, the segment the Korean vendors had been counting on riding to easy HBM-pricing profits.
The firm that walked away unscathed was CXMT. It focuses on DDR5 and server memory and has not yet mass-produced HBM; instead it has been handed an expansion window: its global DRAM share rose from 3% to 8% over the past year, putting it in fourth place globally; Apple has taken the initiative and is in talks to procure CXMT's LPDDR5X, with CXMT's position clear — its quotes will not be lower than Samsung's or Hynix's. The same day, a reposted message from Lingshi Xiantan added another data point to the same trend: supply tightness is intensifying, and Apple is now testing Chinese-made memory chips.
Placing the two messages side by side, CXMT's position is even more nuanced than the earlier sections suggested: the HBM cut damages the Korean vendors' profit expectations, while CXMT — anchored in commodity DRAM — happens to escape the down-leg of the HBM super-cycle, and its bargaining power keeps appreciating in a tight-supply environment. Apple's path from "in talks" to "testing" is the extension of the same logic: when supply is tight, alternative sources become more valuable.
The earlier "Reversal of Bargaining Power" section recorded CXMT refusing Apple's cut request and pricing at parity with Samsung/Hynix; this section is the stress test of that same bargaining power — Nvidia cuts HBM demand, CXMT's core market (DDR5/server memory) is unaffected, and Apple's procurement intent moves from "in talks" to "testing." The storm validates, rather than weakens, CXMT's bargaining position.
What Nvidia cut was the most expensive top-end HBM demand, and the firms hit hardest are the Korean vendors who had bet their future on rising HBM prices; CXMT's moat is that it "did not board the train" — its focus on commodity DRAM means it is not whipsawed by the down-leg of the HBM super-cycle, while its bargaining power keeps appreciating in tight-supply conditions. For the first time the memory market has produced a divergence where the firm that does not chase the hottest track is also the most stable.
Apple Moves to Formal Testing — From Rumored Talks to Product-Line Validation (Incremental Addition 2026-08-10)
An August 9 Wall Street Journal report moved Apple and CXMT's relationship from "in talks" to "testing": Apple is testing CXMT memory chips in iPhone and MacBook product lines to ease the memory shortage triggered by the AI boom; Apple has held initial discussions with CXMT on parts supply, with the goal of using those chips in some of the devices it sells in China. Earlier, Reuters reported CXMT was considering building a second memory-chip fab in Beijing to lift capacity; laptop makers HP and Acer have already begun using CXMT chips in devices sold outside the United States.
This report connects two earlier threads of the page to one point: in "The Reversal of Bargaining Power," Apple wanted to buy and CXMT was not selling cheap; in "Huang's Earthquake for Hynix," Apple moved from talks to testing. Now the testing scope is explicitly iPhone and MacBook — the two flagship product lines — and the stated target is "some of the devices sold in China." That qualifier is itself revealing — the way geopolitical risk is partially defused by market forces is to let Chinese chips enter devices in the Chinese market, rather than to challenge the supply chain in the American market.
Apple moving from "lobbying for procurement permission" to "formal testing" shows that the market pressure of the memory shortage has already outweighed the political-security narrative of supply-chain safety. HP and Acer's follow-on indicates this is not an Apple-only exception but the PC industry's collective choice under the AI memory crunch — CXMT, as the world's fourth-largest DRAM maker, is shifting from "sanctioned party" to "supplier option," a transition that cross-confirms the judgment of the earlier "Reversal of Bargaining Power" section: market forces are now endorsing CXMT's expansion.