China AI Bubble: STAR 50 Falls 40% and $480 Billion Vanishes

The China AI bubble has already burst, at least in the stock market. The STAR 50, the tech-heavy index on Shanghai's Nasdaq-style board, fell about 40% from its summer high in less than four months and erased about $480 billion of market value, according to Reuters. US AI stocks were down only single digits from their records at the time, so China now offers Wall Street a live example of what an AI bubble deflating looks like.

This piece charts how far the crash went, why valuations made it possible, how Beijing is responding, and which lessons carry over to the US. The short answer: the Chinese boom was far more extreme on valuation and far more concentrated than the US one, but the mechanics of the unwind are the ones to watch.

-40%

STAR 50 from its summer high

$480B

Market value erased, per Reuters

-8.5%

CSI 300 in October, worst month since 2016

$19B

State-favored ETF inflows in one week

The crash in numbers: hardest hit were the hottest names

The stocks that rose the most fell the most. Optical module makers Zhongji Innolight and Eoptolink, suppliers to AI data centers worldwide, are down roughly 60% from their June and July highs, Reuters reported. Chip designers MetaX and Moore Threads, which jumped 700% and 400% on their market debuts last year, have halved since July. Robot maker Unitree, whose August listing was about 8,000 times oversubscribed by retail investors, has also halved. Memory maker CXMT lost more than a quarter of its value in a month.

The damage spread to the broad market. China's blue-chip CSI 300 fell 8.5% in October, its worst month since early 2016, and had fallen for 12 straight weeks, its longest losing run since the index launched in 2005, according to International Finance. CNN reported that key Chinese indexes are down more than 20% from their June peaks, crossing the threshold for a bear market. On Tuesday, November 3, the STAR 50 slid another 4% while the CSI 300 closed little changed, Business Insider reported.

Valuations made the fall almost inevitable

At its peak the STAR 50 traded at nearly 250 times its trailing earnings. Even after a 40% fall it sits around 100 times, Reuters reported, against roughly 30 for the Nasdaq 100. A market priced at 100 times earnings has to keep growing profits very fast just to stand still, which leaves almost no margin for disappointment.

Concentration was the other ingredient. AI-linked companies made up 42% of the combined market value of mainland A-shares and Hong Kong H-shares, and 57% of trading turnover, according to Goldman Sachs estimates cited by Reuters. When one theme is that large a share of the market and an even larger share of the trading, there are few buyers left when sentiment turns.

No single trigger, many small ones

Reuters found no obvious selling trigger. Instead it pointed to high prices, tax bills and the threat of the US or other countries banning Chinese equipment in data centers. Authorities had also cracked down on leverage and high-frequency trading, and slowed approvals for humanoid robot listings. Business Insider added waning optimism about US AI capex and liquidity pressure as funds sold to meet redemptions. Gary Ng, senior economist at Natixis, told CNN the selloff reflects a repricing of crowded AI trades, weak domestic demand and a lack of stimulus.

The fundamentals did not help. Unitree's third-quarter revenue rose 70%, but net profit fell 17.4% to 112.7 million yuan, about half what was expected, according to International Finance. A Rhodium Group study cited by the same outlet found combined free cash flow at Alibaba, Tencent and Baidu swung from 170 billion yuan in 2025 to minus 16 billion yuan in the first half of 2026, a pattern that mirrors the cash flow squeeze at US hyperscalers.

Beijing is trying to put a floor under the market

State-favored funds poured about $19 billion into exchange-traded funds in a single week in October, the largest such intervention since the tariff turmoil of April 2025, according to Bloomberg data cited by CNN and International Finance. Wu Qing, chairman of the China Securities Regulatory Commission, said he would do everything possible to keep the capital market safe and stable, CNN reported.

On Monday, November 2, authorities went further. Thirty-one securities firms and fund managers applied for more than 100 billion yuan, about $15 billion, in liquidity support through central bank swap operations, and officials lowered financing costs for share buybacks and encouraged big tech companies to buy back stock, Business Insider reported. Analysts at 22V Research wrote that the measures are intended to put a floor under markets, and doubted a broad 2024-style rebound. Lorraine Tan, Morningstar's director of equity research in Asia, told CNN such measures tend to be short-lived without policies that fix the underlying issues.

What it means for US AI stocks and the AI Bubble Index

The AI Bubble Index tracks the US AI trade, not China's, and it read 63.5 at the November 2 close, in the Normal band, against 63.9 a week earlier. The split inside it is telling. Sentiment and attention, at 92.5, is the hottest family, while valuation and concentration reads 59.1. Fragility, at 71.4, sits above Euphoria at 64.1. In plain terms, the US market is talking about a bubble far more than its prices show one, which is close to the opposite of China this summer. The full method is on the methodology page.

Three lessons still carry over. First, the most crowded suppliers fell hardest: China's optical module makers dropped about 60%, and the same kind of company is at the center of this week's $115 billion Marvell Lumentum deal. Second, an unwind does not need a single trigger once valuations are extreme. Third, state or corporate buying can slow a fall but has not, in China's case, reversed it.

China 2026 next to the dot-com bust

A 40% fall in a few months from a valuation near 250 times earnings is closer to the Nasdaq of 2000 than to anything in the US AI trade so far. The Nasdaq Composite lost about 78% from its March 2000 peak over two and a half years, and it did so in waves, with sharp rallies along the way. Our dot-com bubble timeline shows that pattern year by year, and our history of tech bubbles explains why valuation extremes like China's rarely end quietly.