China's STAR50 AI Index Falls Nearly 40% From July Peak, Erasing About $480 Billion in Market Value

Since U.S. tech stocks began wobbling earlier this month, attention has focused on Nasdaq and chip-stock swings. China's version of the same trade has already cracked much harder.

The STAR50 index, which tracks AI and semiconductor leaders on China's mainland, is down about 40% from its July high. Reuters puts the lost market value at roughly $480 billion in less than four months.

The damage is not limited to the speculative end. China's CSI 300 blue-chip index is down about 20% from its June peak and entered a bear market last week, per BeInCrypto's account.

The names that fell hardest

Retail favorites took the worst of it. Robot maker Unitree was 8,000 times oversubscribed by retail buyers at its August listing. Its shares have since halved.

Chipmaker CXMT has lost more than a quarter of its value in the past month. MetaX Integrated Circuits and Moore Threads Technology jumped 700% and 400% respectively on their debuts last year. Both have roughly halved since July.

Chen Wensheng, a small business owner from Hainan, told Reuters he bought into the rally because he felt Beijing was backing names like Unitree, which counted state-backed funds among its investors. He and his friends are now sitting on losses.

How concentrated the market had become

Goldman Sachs figures show AI-related stocks made up 42% of China's total market capitalization. They accounted for 57% of turnover across mainland A-shares and Hong Kong H-shares.

More than half of all trading was chasing one theme, in a market dominated by small investors.

Beijing-based fund manager Dong Baozhen described the run-up this way: "The stock market later evolved into a speculative crazy bull. And now we see the bubble bursting."

No single trigger set off the selling. Traders cited high prices, looming tax bills and the threat of the U.S. or other countries banning Chinese equipment in data centers.

Chinese regulators had flagged the risk. On Oct. 9, China issued guidelines meant to prevent investment bubbles in what it calls new productive forces, a category that includes AI.

The selloff also undercuts a stated policy aim. Reuters reports it has stalled Beijing's push for a "slow bull" market and dampened efforts to steer household savings into stocks to fund the AI buildout.

Even after the drop, the STAR50 trades at about 100 times trailing earnings, BeInCrypto calculated. The Nasdaq 100 trades at about 30 times. Morgan Stanley expects offshore Chinese stocks to fall another 10% to 15% before China's market bottoms.

The U.S. picture

U.S. AI stocks have taken a far smaller hit. The Nasdaq is about 6% below the record it set three weeks ago. Chip stocks rebounded today, Oct. 26, with the Philadelphia semiconductor index closing up 2.8% after Anthropic's IPO filing.

Still, the cracks are visible. On Oct. 22 the market-cap-weighted S&P 500 fell 1.2% while the equal-weighted version gained 0.5%. James Reilly, senior markets economist at Capital Economics, said a divergence that large had happened only twice since 1995: April 14, 1999, and March 30, 2000.

The 10-year Treasury yield fell below 5% that same day, its largest one-day drop since March 10, 2023.

On Oct. 15, the Philadelphia semiconductor index fell about 2%, and South Korea's KOSPI, heavily exposed to memory chips, plunged almost 8%.

The concentration risk is just as stark at home. A J.P. Morgan analysis found 42 AI-related companies accounted for about two-thirds of the S&P 500's gains since early 2024. Those companies are also borrowing hundreds of billions of dollars for data centers and power.

Not everyone sees a bubble

Dale Smothers, founder and CEO of RDS Wealth, argues the AI bubble is "far from popping" and still sees room for the S&P 500 to reach 8,000.

Hendi Susanto, a portfolio manager at Gabelli Funds, says the problem is not fundamentals: "It is the expectations that are fragile." He notes suppliers are expanding capacity because, in his words, "companies don't commit billions to new fabs unless they see the orders coming."

Economist Greg Rosalsky of NPR's Planet Money offers a caution. A technology can be truly revolutionary and still fuel a bubble that ends in a crash, as railroads did before the Panic of 1873. He adds that none of the current data proves the debt behind AI is a ticking time bomb.

What comes next

The test arrives Wednesday, Oct. 28. The Federal Reserve announces its rate decision at 2 p.m. ET. Microsoft and Alphabet report after the close. Amazon reports Thursday, Oct. 29.

Alphabet expects to spend $195 billion to $205 billion on chips and data centers in 2026. Amazon expects about $220 billion. Both now spend more cash than their businesses bring in.

Microsoft has guided for 44% to 45% Azure growth. Google Cloud grew 82% last quarter. Goldman Sachs analyst Gabriela Borges expects investors to focus on Azure's return on invested capital.

Alphabet fell 7.1% after its July report despite beating profit estimates. All three companies have trailed the S&P 500's 11% gain this year heading into results.

China's selloff has so far been driven by sentiment. Any formal restrictions on Chinese AI equipment exports could hit company revenues directly. Whether U.S. cloud revenue can justify $200-billion-plus capital budgets is the question those earnings reports will begin to answer.