China AI bubble bursts: $480 billion gone, can US tech hold up?

China’s AI rally ran far hotter than the one on Wall Street. AI stocks grew to make up 42% of China’s stock market value and 57% of its trading across mainland and Hong Kong shares, according to Goldman Sachs. When a single theme drives more than half of all trading, a reversal hits almost everyone, including small investors.

Some of the most hyped names show the damage:

• Robot maker Unitree has lost about half its value since listing in August, though it has bounced back somewhat in recent days.
• Chipmakers MetaX and Moore Threads have each fallen by about half since July.
• The broader CSI 300 index of China’s biggest companies is down about 20% from its June peak, after ten straight weeks of losses, which puts it in bear market territory.

Beijing-based fund manager Dong Baozhen summed up the mood, saying the market had turned into a speculative bull run and that the bubble was now bursting.

There was no single trigger. Analysts point to a mix of pressures building at once: prices that had run far ahead of earnings, tax bills, and the threat that the US and other countries could ban Chinese equipment from their data centers.

Valuations are the key number. At its peak, the STAR50 traded at nearly 250 times the past year’s earnings. Even after the crash, it sits around 100 times. The Nasdaq 100, home to Apple, Microsoft and Nvidia, trades at about 30. In plain terms, investors are still paying far more for each dollar of profit in China’s AI stocks than in America’s, which is why some analysts think the selling isn’t over. Analysts at Morgan Stanley think Chinese shares traded offshore could drop a further 10% to 15% before hitting bottom.

China’s government is trying to stop the slide. State-backed funds, known as the national team, poured 128.7 billion yuan (about $19 billion) into nine exchange-traded funds last week. That was the heaviest buying since April 2025, when tariffs rattled global markets and nearly 140 billion yuan went in.

Regulators have also started collecting ideas to calm the market, including curbs on borrowed money for trading and on high-frequency trading. The crash is likely to make Chinese tech companies think twice about new stock listings, which could slow fundraising for AI startups there.

So far, US tech stocks have taken a much smaller hit. Three weeks after hitting a record, the Nasdaq sits only about 6% lower. But this week brings a real test:

• Wednesday: Microsoft and Alphabet report earnings after the market closes, a few hours after the Federal Reserve’s interest rate decision.
• Thursday: Amazon reports.

Investors are watching how much these companies spend against how much their cloud businesses earn back. For 2026, Alphabet has forecast data center and chip spending of between $195 billion and $205 billion, while Amazon’s figure is around $220 billion. In both cases, that outlay is now bigger than the cash their operations generate. Microsoft has guided for 44% to 45% growth at Azure, its cloud unit, and Google Cloud grew 82% last quarter. If those growth numbers disappoint, the spending suddenly looks a lot harder to justify.

If you hold a US index fund or a tech-heavy retirement account, you’re already exposed to the AI trade, because a handful of AI-linked giants make up a big share of those funds. China’s crash doesn’t mean the same thing will happen here, and the US picture is different: American AI leaders have big, profitable businesses behind them, and valuations are much lower.

What it does show is how fast sentiment can turn when prices get far ahead of profits. Some practical points, as general information rather than personal advice:

• Check how much of your portfolio sits in a few big tech names. Many people own more than they realize through index funds.
• Expect sharp moves around this week’s earnings reports. Short-term swings are normal and not a reason by themselves to panic-sell.
• Be wary of anyone pitching Chinese AI stocks as a bargain after a 40% fall. Cheaper doesn’t mean cheap when prices still sit at about 100 times earnings.