Micron's print was the real test of the AI-memory selloff, and the China read-through isn't what people think
The whole memory complex got torched this week, KOSPI down about 10% with a circuit breaker on Tuesday, Samsung and SK Hynix both off around 12%, Nasdaq -2.21%. All of it was really a bet on one number: whether AI hardware spending is peaking. Micron, up roughly 325% YTD into the print, reported after the close on June 24 as the cleanest read on that question.
Here's the China angle, and the part I think people get backwards. The tempting take is "if memory cools, Chinese inference chips win", buy Cambricon and Huawei Ascend as the hedge. It doesn't hold. Those accelerators are themselves HBM-dependent, and high-bandwidth memory is the exact bottleneck the Korean selloff is about. So China's AI-compute chain is coupled to the same memory cycle, not insulated from it. A weak Micron read-through isn't a rotation into Chinese chips, it's a warning for them too.
Which is why I stopped trying to play this with single names. The A-share AI-compute and battery names, CATL, Cambricon, Zhongji Innolight, mostly have no US ADR, and the big China ETFs (KWEB especially) are almost all internet platforms with basically no semis. The one wrapper I found that actually holds the A-share semis and hardware side alongside the usual Tencent and Alibaba is CNQQ. Still a small, young fund, so I'm watching more than sizing in.
Curious where people land: is a soft Micron guide bullish or bearish for the Chinese AI-compute names?