What CXMT Actually Is, and Why Beijing Built It

CXMT is the only company in China that mass-produces DRAM. That sentence is the entire strategic case.

DRAM is the working memory in every phone, PC, server and car. Until recently three foreign firms controlled essentially all of it. For a country that imports more semiconductors than oil, that was a single point of failure, and CXMT was the answer to it.

The company was founded in Hefei in June 2016, built on patents and engineering talent acquired from the bankrupt German memory maker Qimonda, and funded patiently by Hefei municipal investors and the national semiconductor fund. It now runs three 12-inch fabs across Hefei and Beijing and employs roughly 19,300 people.

From Qimonda's Patents to 7.6% of the World's DRAM

The trajectory is measured in wafers, not in press releases.

Omdia's tally puts global DRAM revenue share in the first quarter of 2026 at 38.6% for Samsung, 28.8% for SK Hynix, 22.4% for Micron and 7.6% for CXMT, with CXMT up from 4.7% the previous quarter.

Here is the detail most coverage misses. CXMT's monthly wafer output is roughly 90% of Micron's, yet its revenue share is about a third of Micron's. Capacity is not value. Mix is. Micron converts its wafers into HBM and high-end server DRAM at several times the price per bit that CXMT gets for commodity DDR5. Closing a capacity gap and closing a value gap are different projects.

The Numbers Behind the Fastest Turnaround in Memory History

CXMT lost 16.3 billion yuan in 2023 and 7.1 billion in 2024. Then the cycle turned.

Revenue grew 155.6% in 2025 to 61.8 billion yuan for the company's first annual profit. First-quarter 2026 revenue reached 50.8 billion yuan, up 719% year on year, with 24.76 billion yuan attributable to shareholders. Six guided months of 2026 are set to earn roughly 28 times what the company earned in all of 2025.

Two things caused this, and only one of them is CXMT's doing. The company did stabilise its G4 node and shift mix toward DDR5 and LPDDR5X. But the larger cause was that Samsung, SK Hynix and Micron redirected their best capacity toward high-bandwidth memory for AI accelerators and partly vacated commodity DRAM. CXMT walked into the space they left.

Why Did a 466% Debut Happen?

CXMT raised 57.92 billion yuan by selling 6.688 billion shares at 8.66 yuan, the largest semiconductor listing in STAR Market history and Asia's biggest IPO of 2026. Those shares represented 10% of enlarged capital, implying a 579 billion yuan valuation at listing.

Then the tape took over. The stock opened at 49.50 yuan and closed the first session at a 3.31 trillion yuan market capitalisation, passing Industrial and Commercial Bank of China as the most valuable company listed on the mainland. It added another 12.66% the next day and touched 55.03.

The shockwave went global. Samsung and SK Hynix fell double digits, Micron and SanDisk sold off hard, ASML dropped, and the KOSPI triggered a circuit breaker. Markets did not treat this as a listing. They treated it as a supply announcement.

The Float Is the Whole Story

CXMT's July 24 listing announcement showed that less than 7% of its enlarged share capital was initially unrestricted, while institutional demand ran to roughly 570 times the final offline allocation. Only 6.73% of the expanded share capital was tradable, so investors competed for a limited number of shares and amplified the rise.

Run that through the arithmetic. At 52.87 yuan across roughly 66.9 billion shares, the market capitalisation is about 3,536 billion yuan, or 522 billion dollars. Multiply by 6.73% and the tradable stock is worth 238 billion yuan, about 35 billion dollars.

So 35 billion dollars of freely trading shares is setting the price of a 522 billion dollar company. The other 93% has never been tested against a real bid. Anyone quoting CXMT's market cap as a measure of what the market thinks CXMT is worth is quoting a number that 93% of the owners have not voted on.

That cuts both ways, which is exactly why it matters. Thin floats overshoot upward on the way in and downward when supply arrives.

Company Health: Cash Is No Longer the Constraint

Operating cash flow reached 36.5 billion yuan in 2025, before the price surge. Add 57.9 billion yuan of IPO proceeds, potentially 66.6 billion with the overallotment, and a first half guided to 50 to 57 billion yuan of profit, and the funding question that dogged CXMT for nine years is settled for now.

What the money buys is the open question. Of the 29.5 billion yuan in planned net proceeds, 20.5 billion funds wafer production lines and DRAM technology upgrades and 9 billion supports forward-looking DRAM research, with no dedicated HBM project disclosed in the prospectus.

That allocation is telling. CXMT is doubling down on commodity DRAM capacity rather than chasing the high-margin AI memory where the incumbents make their money. It is a rational choice given tool access. It also caps the ceiling.

For scale, the incumbents are outspending CXMT by roughly fourteen times. Samsung's 2026 investment plan exceeds 110 trillion won, Micron's fiscal 2026 capital expenditure is set to pass 25 billion dollars, and SK Hynix raised 26.5 billion dollars in a Nasdaq share sale days before CXMT's subscription window opened.

What Is CXMT Actually Worth?

Start with the multiples. Half-year guidance of 50 to 57 billion yuan annualises to roughly 107 billion yuan, or about 1.60 yuan per share across 66.9 billion shares.

Now the comparison that matters.

Divide each producer's market value by its share of global DRAM revenue. SK Hynix costs 29.0 billion dollars per point of share. Micron costs 41.4 billion. CXMT costs 68.7 billion, which is 1.7 times Micron and 2.4 times SK Hynix, for the producer with the worst cost position, no meaningful HBM revenue and restricted access to advanced tools.

There is a second problem with 33x. These are peak-cycle earnings. Memory has never rewarded peak earnings with a high multiple, for the good reason that the multiple is the market's way of saying the earnings are temporary. Micron's price-to-earnings multiple has historically bottomed between roughly 3.5 and 8 during past supercycles. Micron itself currently trades at a low double-digit multiple of this year's earnings despite better margins, better mix and take-or-pay contracts running to 2030.

Price Targets and the Math Behind Them

Three scenarios, each anchored to an explicit earnings figure and an explicit multiple, on a twelve to eighteen month horizon.

Bear reasoning: 13x is still a premium to Micron's historical trough multiple, and 1.10 yuan assumes a normalisation rather than a collapse. This case lands close to Morningstar's 14.90 yuan fair value and the 16.10 yuan low on the street, which is a useful cross-check, since three independent methods converging near the same number is more informative than any one of them.

Base reasoning: 21x on maintained earnings gives CXMT roughly double Micron's current multiple purely for its strategic position and domestic monopoly. That is a generous premium, not a stingy one.

Bull reasoning: 35x on 2028 earnings requires HBM revenue that does not exist yet and tool access CXMT does not currently have. It sits below Nomura's 116 yuan target.

Probability-weighting at 30% bear, 45% base and 25% bull gives 42.2 yuan, about 20% below spot. Note that this sits below the 61.70 yuan consensus. The difference is not a disagreement about CXMT's growth. It is a disagreement about paying 33x for peak earnings in the most cyclical business in technology.

Is CXMT Expected to Grow?

Yes, and quickly. That was never the question. Three things decide whether shareholders capture it.

• Cycle duration. TrendForce sees DRAM revenue rising 303% in 2026 to 619 billion dollars and expanding to 903 billion in 2027. If that holds, CXMT grows into part of its multiple.

• Mix migration. Commodity DDR5 is where the volume is. HBM is where the margin is, and CXMT is roughly three years behind with no disclosed project.

• Tool access. Without advanced lithography, cost per bit stays roughly 30% above Samsung and SK Hynix, which caps margins through the next down-cycle.

Key Risks

• Lock-up expiries in 2027 release supply into a float that is currently 6.73%.

• A federal procurement ban on Chinese memory takes full effect in the United States in December 2027.

• CXMT appears on the Pentagon's list of firms with alleged military ties, and export controls on advanced tools remain in force.

• The memory cycle itself. Every prior peak was followed by a capacity glut.

• Cost per bit roughly 30% above the leaders, with no HBM revenue to offset it.

What to Watch Next

• First-half results against the 110 to 120 billion yuan revenue guidance, which is the first reported number since listing.

• Any confirmed HBM customer qualification, which would change the ceiling rather than the cycle.

• DRAM contract pricing through the fourth quarter, the single largest earnings variable.

• Stock Connect eligibility, which would open the shares to offshore capital and change the float dynamic.

• Apple's reported testing of CXMT parts, a genuine credibility marker if it converts to volume.

How to Trade a Stock Most Investors Cannot Buy

CXMT sits on the Shanghai STAR Market under 688825, quoted in yuan. It was not included on the Northbound Stock Connect eligibility list updated on July 24, 2026, so access from outside the mainland depends on a brokerage or institutional route supporting A-shares.

That is the practical problem. The most consequential listing in semiconductors this year is one that most global investors have no way to express a view on, in either direction.

You can do it on ApeX Omni with the CXMT-USDT RWA perpetual, with up to 50x leverage, long or short, no KYC and full self-custody. No A-share account, no quota, no settlement window. The two-sided interest is already there: on-chain short positioning in CXMT has been building since the debut.

Thin float, peak-cycle earnings, a 100 yuan gap between the highest and lowest published target. Position size accordingly, because leverage magnifies losses as fast as gains.

Trade CXMT

This article is for educational and informational purposes only and does not constitute financial, investment, or legal advice. Traditional finance (TradFi) markets and financial products involve risk, and past performance does not guarantee future results. Market conditions, valuations, and other financial information referenced are accurate to the dates and sources cited but may change without notice. Always conduct your own research and consult a licensed financial professional before making any investment decisions. Figures are current as of July 31, 2026, with conversions at approximately 6.77 yuan per dollar. Do your own research.

We appreciate your continued support.