CXMT Fell 6.5% and a $19.6 Billion Profit Forecast Wasn't Enough

I spent Monday morning watching a 6.5% slide that, honestly, I should have seen coming. ChangXin Memory Technologies, China's largest memory chip maker, told the market on Friday it expects net profit of at least 131 billion yuan for the first nine months of 2026, roughly $19.6 billion. The same stretch a year earlier produced a 5.28 billion yuan loss. Investors sold the stock anyway.

That gap between the headline and the tape is the whole story. The question worth asking: what's the market actually pricing here, if not earnings?

The Profit That Didn't Matter

Start with the swing. A 5.28 billion yuan loss flipping into 131 billion yuan of profit is a 136 billion yuan turnaround in twelve months. On paper, that's the kind of print memory makers dream about. This is a brutal cyclical business. DRAM and NAND prices sit flat for quarters and then move violently in a single one, and the entire P&. L of every manufacturer on earth swings with them.

The bullish thesis writes itself. AI servers have soaked up high-bandwidth memory capacity, the big three, Samsung, SK Hynix and Micron, have been steering wafers toward the premium end, and that leaves the commodity market tighter than it's been in years. A domestic champion with 131 billion yuan of profit and the weight of Beijing's self-sufficiency push behind it should be the easiest buy in Asia.

But here's what the guidance doesn't tell you. It doesn't tell you what happens when the cycle turns, and it doesn't tell you where the profit came from. Admittedly, a large chunk of CXMT's number depends on pricing its competitors influence far more than it does. A profit forecast in this business is a snapshot of a price deck, not a statement about competitive advantage.

Why a US Selloff Travels

Friday's rout in US chip names showed up in Asia on Monday with remarkable efficiency. That's not coincidence. Memory is the most globally arbitraged part of the semiconductor supply chain. There's one price for DRAM, more or less, and it gets set by supply and demand that don't care about borders. So when New York decides AI capex expectations are too rich, Seoul, Taipei, Tokyo and Shanghai all reprice the same day.

But there's a second layer that matters more for CXMT specifically. Foreign investors treat Chinese memory as a policy trade as much as a pricing trade. Export controls on advanced lithography tools, restrictions on equipment sales, the constant back and forth over what Chinese fabs can buy, all of it sits on top of the earnings. So you get a company that can post a record profit and still get sold because the tape in the US turned red.

Is that rational? Partly. Skeptics would say CXMT's capacity additions are exactly the thing that ends a memory upcycle early, and their point about the track record of state-backed capacity races has merit. Proponents would say the domestic substitution narrative runs for a decade regardless of one quarter's price action. Both can be true at once.

My Read

Here's my honest take. A profit forecast isn't a moat, and the 6.5% drop is the more informative number of the two. Guidance tells you where prices were. The selloff tells you where investors think they're going.

If a $19.6 billion profit estimate can't hold a stock up for 48 hours, what does that say about confidence in the memory cycle itself? That's the real signal buried under Monday's red candles.

I'm not entirely convinced the market's reaction is wrong, granted, and I'll be watching three things. Q4 DRAM contract pricing, because that's where the guidance either holds or gets walked back. Any fresh US export control headlines, since those hit the equipment side of the story hard. And whether Friday's US chip selloff stabilizes, because if it doesn't, Asia keeps importing the mood.

Time will tell, though. Memory has burned the people who called the top early, and it's burned the ones who called the bottom. All we know for sure is that on Monday the profit was real, and so was the selling.