CXMT DRAM Is Now Inside HP, Asus and Acer Laptops (August 2026): A 716% Revenue Surge, a Pentagon Listing — and Chinese Memory That Costs the Same as Samsung's

Three of the world's largest PC brands have quietly started shipping memory from a company on the Pentagon's list of Chinese military firms. Nikkei Asia reported on 4 August 2026 that HP, Asus and Acer had begun using limited quantities of DRAM from China's ChangXin Memory Technologies in selected budget notebooks sold outside the United States, having completed qualification of the parts around the middle of the year. It happened eight days after CXMT closed its Shanghai trading debut up roughly 466%, becoming the most valuable listed company in China overnight. And the detail that dismantles the obvious explanation: the Chinese chips are reportedly priced on par with Samsung's rather than at a discount, with supply hard to secure beyond the current quarter. Nobody switched to save money. They switched because conventional DRAM contract prices rose 93–98% in a single quarter and there was nothing else to buy.

This is what a physical shortage looks like when it reaches the end of the chain. For two years the memory story has been told from the supply side — record profits at Samsung and SK hynix, HBM sold out, capacity racing toward AI accelerators. This week it was told from the buyer's side, and that is where the interesting information is. A qualification decision is expensive, slow and reputationally loaded; PC makers do not run one on a whim. That three of them completed it, and a fourth is evaluating, is a harder data point about scarcity than any price index, because it is a revealed preference rather than a quoted number.

What actually happened: three brands, budget notebooks, and no discount

The sequence compressed into ten days. On 27 July 2026, CXMT listed on the Shanghai Stock Exchange's STAR Market. On 3 August, Reuters reported the company was in early-stage discussions over a second Beijing fab. On 4 August, Nikkei Asia reported that HP, Asus and Acer had moved from evaluating CXMT parts to shipping them.

The specifics matter more than the headline. The volumes are limited and the model count is limited. The machines are budget notebooks, and they are sold outside the United States. Qualification — the engineering process by which an OEM validates that a component meets its reliability, compatibility and warranty standards — was completed around the middle of 2026, meaning the commercial decision followed months of technical work that began while the shortage was still building. This was not a panic buy in August. It was a hedge placed in the spring that has now been exercised.

Two details cut against the intuitive reading. First, price. The reporting is explicit that CXMT's DRAM is priced on par with leading suppliers such as Samsung rather than at a discount. The familiar template for Chinese entry into a hardware market — undercut on price, take share, move upmarket later — does not describe what is happening here, because in a shortage the marginal supplier does not need to discount. It needs only to exist. Second, availability. Supplies were reported as difficult to secure beyond the current quarter, because CXMT allocates a significant portion of its capacity to Chinese customers such as Huawei, alongside domestic smartphone brands and cloud operators. The new supplier is itself supply-constrained.

That combination — same price, uncertain availability, meaningful reputational friction — tells you what the alternative looked like. An OEM accepts all three only when the counterfactual is not shipping the product.

Source: TrendForce 1Q26 DRAM revenue report and TrendForce 3Q26 price outlook.

Read that table as a deceleration and you will misread it. Three consecutive quarters of increases compound to something close to a quadrupling of the input cost. The third quarter's 13–18% is smaller than the first quarter's 93–98% only in percentage terms; applied to a base that has already tripled, it is a larger absolute increase than the entire 2024 price range for the same part.

The mechanism: how AI capex became a laptop-parts shortage

The chain from a data-centre construction budget to a notebook bill of materials is shorter than it looks, and it runs entirely through fab allocation.

The critical property is that this is a capacity story, not a demand story. Nobody stopped wanting laptop memory. Samsung, SK hynix and Micron — who between them took roughly 90% of first-quarter DRAM revenue, at 38.5%, 28.8% and 22.4% respectively — reallocated wafer starts toward the products with the highest margin per wafer, which is exactly what a rational manufacturer does. Industry estimates put AI data centres at around 70% of 2026 memory production. The residual is what the consumer electronics industry gets to fight over.

Two consequences follow. First, price increases in this regime do not clear the market in the usual way: the buyers being priced out are consumers whose alternative is an older laptop, while the buyers absorbing the increases are hyperscalers capitalising the cost over years of depreciation. TrendForce's own language for the third quarter is that consumer customers are reaching their affordability limit — a supply chain does not say that about a market that is functioning. Second, a persistent gap between what a part is worth and what the incumbents will supply creates room for a supplier that would otherwise never clear the qualification bar. Scarcity is the most effective market-entry subsidy there is, and no policy designed it.

The 466% debut: what CXMT's Shanghai listing repriced

The market side of this story arrived first. CXMT priced 6.688 billion shares at 8.66 yuan on the STAR Market, raising about 57.92 billion yuan — Asia's largest IPO of 2026 and the largest semiconductor listing the exchange's growth board has hosted. The shares opened at 49.50 yuan on 27 July and closed the session up roughly 466%, lifting the Hefei-based company to a market capitalisation near 3.3 trillion yuan and making it the most valuable listed company in China. Around 9.4 million retail accounts applied; the retail tranche was oversubscribed roughly 212 times. (Coverage: CNBC.)

The operating figures underneath are genuinely strong and genuinely narrow. Counterpoint Research put CXMT's revenue up 716% year-on-year in the second quarter of 2026, the fastest growth of any DRAM supplier, taking it to about 7% of the global market and fourth place worldwide. The same data put Samsung at 39% and showed SK hynix's share falling to 26% from 39% a year earlier despite its own revenue rising 214% — a reminder that in a price cycle this violent, share can move sharply while every participant grows.

What the debut valued, then, is a single-digit share of a commodity market at the most favourable price environment that market has ever produced, on a capacity base of roughly 300,000 wafer starts per month across three fabs, heading toward about 350,000 by end-2026. Expansion plans could eventually push the combined figure past 600,000, and Reuters reported on 3 August that discussions over a second Beijing fab in the Yizhuang district were at an early stage. Those are real numbers, but they are fab-timetable numbers — they arrive over years, while the price environment being capitalised today is a quarterly phenomenon.

One bottleneck no amount of capital resolves quickly: access to advanced chipmaking equipment, which export controls specifically target and on which CXMT's reported ambitions for 2030 depend. That risk cuts both ways — the same controls that cap CXMT's ceiling are why its domestic customers have an incentive to buy from it regardless of price.

The channel to what you actually trade

Be honest about where this lands, because the temptation to route every story into a currency is exactly what produces bad analysis.

The direct exposure is equity, and it is inside indices most readers already trade. The memory names — Micron in the US500 and NAS100, Samsung and SK hynix in Korea — face a competitor gaining qualified design wins at incumbent prices, in the segment they deliberately de-prioritised. The PC OEMs face an input cost that has roughly tripled in eighteen months against products with thin margins and price-sensitive buyers. Those are opposite positions in the same shortage, and the AMD and memory-cycle threads we covered in how Korea's leveraged ETFs amplified the memory selloff and in the Apple and Amazon quarter where one memory shortage explained both results are the same mechanism seen from the supplier and the consumer side respectively.

The currency channel exists but is small, and the arithmetic is the point. US CPI captures this through core goods: the Bureau of Labor Statistics relative-importance tables put computers, peripherals and smart home assistant devices at roughly 0.3% of the basket as of December 2025, with the broader information technology hardware and services grouping near 1.7% (methodology: BLS). Even a dramatic percentage move in a 0.3% weight moves the headline index by hundredths of a point. What is informative is the sign change: a durable-goods category that spent most of the last decade in measured deflation, contributing a small persistent drag to core inflation, flipping to sustained increases. Remove a reliable disinflationary contributor and the core goods component stops helping, in the same bucket where tariff pass-through already sits. For the US dollar, that reaches the rate factor — one of the five the meter scores — through the inflation trend, not through the price of a laptop. Anyone claiming a DRAM contract print will move a major pair is selling a channel that does not carry that much water.

What would change the picture

Four observable things, none of which requires a forecast.

Contract prices flattening rather than decelerating. The 3Q26 estimate of 13–18% is still a large increase. The regime change is a quarter printing near zero, which would indicate the allocation squeeze had stopped tightening.

AI order deferrals. Because the shortage is a reallocation, only slowing demand at the destination frees capacity at the source. Hyperscaler capex guidance is the leading indicator here, not memory pricing itself.

Volume, not presence, at the OEMs. CXMT parts appearing in limited quantities in budget notebooks in non-US markets is a shortage signal. The same parts appearing in mainstream models across regions would be a market-structure signal — a different and much larger story. Watch which one the next round of reporting describes.

The June 2027 product-level 1260H rule. As written, it reaches contracts for goods incorporating items from listed entities. Any vendor with federal exposure needs a supply-chain answer well before that date, and that answer will constrain how far CXMT content spreads in the US market regardless of price.

The unifying observation is that a shortage created by one boom collided with a policy architecture aimed at something else entirely — defence procurement and advanced tooling, not notebook DRAM — and at the margin the shortage was the stronger force. That is not a verdict on the policy or a prediction about share. It is a rule worth carrying into other stories: when supply is genuinely scarce, qualification standards, procurement preferences and country-of-origin instincts all become negotiable, in that order.

For more on how we read supply-driven price signals into the five factors, see the methodology behind the meter.

Educational macro context only — not investment advice.