Key Notes
- China’s STAR50 has fallen 40% in less than four months, erasing about $480 billion in market value.
- The selloff has spread across AI stocks and broader Chinese benchmarks, raising questions about valuations and profitability.
- Wall Street faces its own concentration and leverage risks, although the Chinese and American figures are not directly comparable.
China’s STAR50 index has fallen 40% in less than four months, erasing an estimated $480 billion in market value as investors retreat from the country’s artificial intelligence and semiconductor trade.
Reuters reported the drawdown in an October 26 analysis examining the unwinding of China’s technology rally. The scale of the reversal is renewing scrutiny of expensive AI-linked shares worldwide, including the companies that dominate US stock benchmarks.
The damage extends beyond technology. Bloomberg’s October 25 analysis put Chinese equities’ decline from their May peak at $3.5 trillion. That broader estimate covers a different universe and period from the STAR50 figure; the two losses should not be added together.
China’s AI Trade Reverses
The STAR50 represents leading companies on Shanghai’s technology-focused STAR Market. The exchange’s index overview distinguishes it from dedicated AI and chip indexes within the wider STAR family.
According to Reuters, Unitree shares have halved since their August listing, while optical-component suppliers Zhongji Innolight (SZSE: 300308) and Eoptolink Technology (SZSE: 300502) are roughly 60% below their June and July highs.
Hong Kong’s listed model developers have also suffered. Bloomberg reported on October 23 that MiniMax and Zhipu had collectively lost $150 billion from their respective peaks. Its account pointed to increasingly aggressive model pricing and prospective rival listings that could compete for investor capital.
The broader CSI 300 reached the conventional bear-market threshold, 20% below its June high, on October 22, as we previously reported. That benchmark tracks large mainland companies across industries, making its retreat a wider signal than weakness in a handful of AI names.
High Valuations Leave Little Room for Disappointment
Reuters put the STAR50’s trailing price-to-earnings ratio near 100, down from almost 250 at its peak. Traders cited expensive shares and the threat of restrictions on Chinese suppliers serving overseas data centers among the pressures on prices.
Such multiples depend heavily on future earnings growth. A company can increase sales and still deliver poor shareholder returns if its profits fall short of the expectations embedded in its starting valuation. Falling share prices therefore need not mean the underlying technology has stopped improving.
China’s problems also reach outside AI. Bloomberg’s broader market analysis found that 2026 earnings forecasts for MSCI China constituents had fallen more than 4% since late March, while projections for the Hang Seng Tech Index had dropped more than 13%.
Weak consumption and price competition can squeeze margins even in industries with growing output. That combination makes the selloff harder to explain solely through enthusiasm for one technology.
Wall Street Faces Its Own Concentration Test
Goldman Sachs figures cited by Reuters placed AI-linked stocks at 42% of market capitalization and 57% of turnover across mainland A-shares and Hong Kong H-shares.
In the United States, an October 8 Seeking Alpha report described a JPMorgan basket of 41 AI-linked companies representing 49% of the S&P 500’s market value.
Those percentages measure different markets using different company selections and dates. They show substantial thematic exposure, but they do not establish a common threshold at which a bubble must burst. Market-capitalization shares also say nothing directly about how much revenue those companies earn from AI.
The portfolio implication is nevertheless significant: investors holding a broad benchmark can have considerable exposure to the same spending cycle through chipmakers, cloud platforms, equipment suppliers and utilities.
US Leverage Remains Elevated
FINRA’s latest margin statistics show customer securities-account debit balances of $1.480 trillion in September, up from $1.454 trillion in August. The latest figure remained below June’s $1.502 trillion, rather than setting a fresh record.
Borrowing can amplify losses when declining collateral values force investors to add cash or sell securities. However, a nominal debt total does not measure every form of market leverage, identify which positions are financed or predict when forced selling will occur.
Valuations remain elevated too. Multpl’s Shiller tracker showed the S&P 500’s cyclically adjusted price-to-earnings ratio at 40.66 at the October 26 close, below its December 1999 peak of 44.19.
CAPE uses ten years of inflation-adjusted earnings. It cannot be compared directly with the STAR50’s trailing annual earnings multiple, and a high reading is not a timetable for a crash.
Anthropic’s IPO Adds a Financing Test
Anthropic has added another major test of investor appetite by filing for an IPO under the proposed Nasdaq ticker ANT. Its October 26 announcement did not specify an offering price or share count, leaving its eventual public-market valuation unsettled.
The filing reported a first-half net loss of approximately $80.7 billion, but that number is not equivalent to cash spent running the business. As we reported, roughly $72.4 billion related to fair-value adjustments involving financing instruments.
The central question across both markets is whether earnings and cash generation can justify the capital committed to AI. China’s losses demonstrate the consequences when expectations unravel. They raise the stakes for Wall Street’s earnings and financing tests without determining their outcome.
