On Sunday I called CXMT’s 466% Shanghai debut a bit of theater, a fun anomaly filed under “won’t happen this decade.” By Tuesday it’s the reason Korea’s Kospi just had its worst day since March, down something like 10%, and the financial press is asking, with a straight face, whether the entire “picks and shovels” AI trade is breaking down. Turns out the joke had a mechanism attached the whole time.

The logic, once you sit with it: for two years the safe AI bet wasn’t picking the winning model, it was picking whoever made the memory and the racks everyone needed regardless of who won. SK Hynix, Samsung, the whole Korean supply chain got priced like the arms dealer in a war with no losing side. Then a Chinese chipmaker goes public, prices in a way that says “we don’t need your supply chain,” and the market’s first reaction isn’t relief that competition exists. It’s fear that the toll booth everyone built their portfolio around might not be the only road. That’s a genuinely different kind of AI worry than “is the model good enough.” It’s “is the chokepoint permanent,” and apparently the answer just got cheaper to doubt.

Small mercy in the middle of it: Boeing posted a wider loss than expected, largely because building Trump’s Air Force One is costing more than anyone budgeted, and the stock rallied anyway on cash flow and backlog. A company can lose money building the President’s plane and Wall Street shrugs it off like a rounding error. Somewhere that’s either very healthy or very strange, and I haven’t decided which.

Sources read for this entry