According to Reuters, the Chinese STAR50 index, which includes AI and semiconductor leaders, fell 40% in less than four months from its July peak, wiping out approximately $480 billion in market capitalization. The broad-market CSI300 index retreated by approximately 20% from its June high, while individual stocks, from robotics company Unitree to optics manufacturers Zhongji Innolight and Eoptolink Technology, lost between half and 60% of their value. Even after such a crash, the STAR50 trades at approximately 100 times annual earnings, whereas the Nasdaq 100 stands at about 30. This is a repricing of expectations around AI, not a shutdown of technologies.

image
image

What Happened

As reported in a Reuters article dated October 26, 2026, the crash began after the July peak and, in less than four months, stripped Chinese tech stocks of approximately $480 billion in market capitalization, sending the STAR50 down 40%. The broad-market CSI300 index retreated by approximately 20% from its June high. Robotics company Unitree (ticker 688836.SS) went public in August with an IPO that was oversubscribed 8,000 times, but its shares have halved in value since their debut. Chipmakers MetaX Integrated Circuits and Moore Threads Technology lost about half their value from July levels. Optical component manufacturers Zhongji Innolight and Eoptolink Technology fell approximately 60% from their June and July peaks. At its peak, the STAR50 was valued at approximately 250 times annual earnings; after the drop, the multiple remained around 100, which is roughly three times higher than the Nasdaq 100.

Context

The backdrop to the event is the Chinese market's extremely concentrated bet on AI. According to Goldman Sachs, AI stocks account for 42% of the market capitalization of Chinese A-shares and H-shares and 57% of trading volume, meaning a repricing in one sector dragged down the entire market. The summer rally was based on the calculation that new tech companies would be funded through the exchange: the public IPO window became part of Beijing's industrial strategy. Investors priced in sustained explosive growth in AI demand, and the Reuters publication explicitly cites inflated valuations and doubts about this demand as reasons for the decline. For scale calibration: the US Nasdaq 100 is down approximately 6% from its record, meaning the correction in China is currently proceeding significantly deeper than the correction in the US.

Why This Matters for the Industry

The blow hit the entire Chinese AI supply chain, from robots and chips to optical components for data centers, and primarily affected access to capital: cheaper stocks reduce the ability to raise funds for accelerator development and production expansion. IPO plans for tech companies, on which Beijing relied for exchange funding, are at risk. A separate layer of risk is geopolitical: threats of US bans on the supply of Chinese equipment to data centers have been voiced, which directly concerns optics manufacturers Zhongji Innolight and Eoptolink Technology. For product builders, the procurement context is changing now: long exclusive contracts with a single vendor appear riskier, and buyers increasingly demand measurable ROI instead of demonstrations. Scenario-wise, in the coming months, a slowdown in capital expenditure on Chinese AI infrastructure and a shift in priorities from scaling to efficiency are likely; the flip side of the crash is the cheapening of hardware, talent, and assets purchased by ecosystem participants.

Why This Matters for Users

Operationally, the crash has not yet affected anything: models, clusters, and APIs are operating as usual, because a drop in stock prices itself does not shut down infrastructure. Practical consequences relate to plans. Researchers and engineers planning to train models on Chinese accelerators from Moore Threads and MetaX should account for the possible slowdown in the development of these platforms and their ecosystems due to funding compression. Teams dependent on Chinese vendors should reasonably review contracts: suppliers' financial stability, service level agreements, and exit terms. For the reader-investor, the event provides a clear example that even companies with strong state support, bought at frenzied prices, can lose half their value in a matter of months, and this is a useful benchmark for assessing risks in investments in the Chinese tech sector.

What Is Still Unknown / Limitations

Comparing multiples of "approximately 100 times annual earnings for STAR50 versus approximately 30 for Nasdaq 100" is incorrect without normalization for index composition: Nasdaq 100 is predominantly profitable platforms, while STAR50 is semiconductors and capital-intensive production, so a direct comparison does not prove the superiority of one market over the other in substance. The Reuters headline "In China, the AI bubble has already burst" is a journalistic interpretation, not a measurement result: stock price is not a valid benchmark for model progress, and "the bubble burst" here means "valuations burst," not "models got worse." Scenarios on a 6-month and 2-year horizon, including a slowdown in capex, the fate of the IPO window, and the survivability of compute startups, are scenarios, not facts: their outcome depends on whether AI demand is confirmed by actual revenue and whether the US moves from threats to active restrictions.

Sources

Author

Look at AI, editorial team