Everyone’s Cheering China’s $11 Billion Chip Champion. Three Things the Number Hides.

China's answer to Micron just posted a headline that stopped the market cold: 77.6 billion yuan ($11 billion) in net profit for the first half of 2026, up from a loss a year earlier. But strip the number back and a lot of what made CXMT look invincible turns out to rest on things the company hasn't actually promised — and one it may not fully own.

ChangXin Memory Technologies, China's largest maker of DRAM memory chips, reported first-half revenue of 150.3 billion yuan ($22.4 billion), up 874% year on year. Analysts rushed to compare its 51.6% net margin to Micron's 63%, noting the gap was closing fast. The stock has run more than 500% since its July debut in Shanghai. On the surface, China has a genuine memory-chip champion trading almost neck-and-neck with the Americans.

The surface is worth a second look.

The margin isn't technology. It's a depreciation trick the cycle is paying for.

CXMT's first-half gross margin hit 84.84%. That sounds like a company with a technology edge. It isn't — it's a company catching a price wave at exactly the right moment in its spending cycle.

Here's the mechanism. Making DRAM requires enormous upfront investment in fabs, and that cost shows up every quarter as depreciation, whether or not chips sell. A year ago, that fixed cost was crushing CXMT: based on its own half-year filing, depreciation ran close to three-quarters of revenue, which is why the company was losing money. This year, a global memory shortage sent DRAM prices soaring. Revenue jumped nearly ninefold. That same depreciation bill — up just 28% — suddenly shrank to under a tenth of revenue.

In plain terms: the cost didn't fall, the price rocketed past it. The 84.84% margin is the gap between a fixed cost and a temporary price spike. When DRAM prices turn — and memory is the most brutally cyclical business in tech — the depreciation bill won't turn with them. The same leverage that created the boom will run in reverse.

The Apple story everyone's trading on? Management never confirmed it.

Since August, headlines have linked CXMT to Apple, after the Wall Street Journal reported Apple was testing its chips for iPhones and MacBooks sold in China. Some traders have treated a supply deal as all but done.

At its Sept. 7 earnings call, CXMT management said only that the company explores cooperation with quality global clients "with an open attitude," and that its products can compete with international rivals on performance. That is not a confirmation. No order, no timeline, no verification stage, no customer named. Apple and CXMT have both declined to comment to Reuters. Whatever this $11 billion came from, it did not come from Apple — it came from an already-disclosed customer base and a price surge the company did confirm.

The $11 billion may overstate what shareholders actually get.

This is the part almost no English-language coverage has flagged. Research firm SemiAnalysis, examining CXMT's earlier IPO prospectus, argued the company's consolidated profit overstates what public shareholders will actually receive — by roughly four times. The reason is structural: CXMT controls the voting rights of key fabs through long-term agreements but owns a minority of their economics, letting it consolidate operations it mostly doesn't own. On the FY2025 numbers SemiAnalysis examined, about 74% of net income was attributable to minority interests, not parent shareholders.

That analysis predates this half-year report and reflects one firm's read of the ownership structure, not an official company figure — CXMT's headline profit is calculated by standard accounting rules. But it raises a question the celebration has skipped: whose $11 billion is it?

What's actually worth watching

CXMT is a real company with real momentum — it is the world's fourth-largest DRAM maker, and its arrival knocked 9% off Micron's stock in a single day. None of the above makes it a paper tiger. But it does mean the three pillars holding up the story are shakier than the headline suggests.

Three signals will tell you which way it breaks. First, the direction of DRAM spot prices — the single thing propping up that 84.84% margin, which management itself concedes will swing with the cycle. Second, whether the Apple talks ever produce an actual filing-grade contract, rather than another round of "open attitude." Third, whether CXMT's push into high-bandwidth memory — the AI chips where Samsung and SK Hynix still dominate and CXMT has yet to mass-produce — turns into revenue with a line item, instead of a roadmap. Until then, the $11 billion is real on paper. What it's worth is a different question.