Photo: Larry Qian / Wikimedia Commons (public domain)

Did the AI Bubble Burst in Asia? China's AI Stocks Lost 40%, Korea Crashed 39%, and Japan Barely Flinched

The short answer: true in China, half-true in Korea, false in Japan

"The AI bubble has burst in Asia" is one of the most searched market claims this month. We checked it market by market. The honest answer depends on which Asian market you mean.

Sources: Reuters, Korea Exchange data via Korea Economic Daily and JoongAng Daily, Macrotrends, AP. Figures rounded.

A quick definition helps. A bubble is when prices rise far beyond what the underlying profits can justify, driven by the fear of missing out. It bursts when prices fall hard and don't quickly come back. A 20% drop is a bear market. A 40% drop in a few months, with traders saying the story is over, is what a burst looks like.

China: the bubble that actually popped

China is the clearest case. Reuters ran the headline outright on October 26: "In China, the AI bubble has already burst."

• The STAR50, Shanghai's index of AI and semiconductor leaders, fell about 40% in less than four months, erasing roughly $480 billion of market value.
• At its peak the index traded at nearly 250 times earnings. Even after the crash it's near 100 times, compared with about 30 for the Nasdaq-100.
• GPU makers Moore Threads and MetaX, which jumped 400% and 700% on their debuts last year, have each roughly halved since July.
• Optical-parts makers Zhongji Innolight and Eoptolink are down about 60% from their summer highs as Washington moves toward banning them from US data centres.
• Robot maker Unitree, whose IPO was about 8,000 times oversubscribed by retail buyers, has roughly halved since it started trading in August.

How big was the mania? According to Goldman Sachs, AI stocks make up 42% of China's market value and 57% of its trading volume. A Shanghai fund manager told Reuters bluntly that "all bubbles inevitably burst."

Beijing helped let the air out on purpose. Regulators clamped down on leverage and high-frequency trading, and state-backed investors sold into rallies. And on October 27, when Anthropic's IPO filing lifted chip stocks everywhere else in Asia, mainland China and Hong Kong fell. As one economist told Fortune, China sits "outside the main U.S. AI supply chain."

Verdict: burst.

South Korea: a crash, but not (yet) a burst

Korea had the most dramatic ride of 2026.

The Kospi started the year at 4,214. Driven by Samsung Electronics and SK Hynix, which make the high-bandwidth memory chips inside AI servers, it shot past 9,000 in June. SK Hynix closed at a record 2.52 million won on June 17.

Then it unravelled. By October 15, the Kospi had closed at 5,554.22, down 7.57% in a single day and 39% below its June record. Over seven straight losing sessions, Korean stocks lost 1,193.5 trillion won (about 21%) of their value. Trading was halted for 20 minutes by circuit breakers. One Korean headline captured the mood: retail investors were "too scared to open an account."

But here's where Korea differs from China: the profits are real.

On October 28, SK Hynix reported third-quarter results:

• Revenue: 104.6 trillion won, up 328% from a year ago, its first quarter above 100 trillion won.
• Operating profit: 83.3 trillion won (about $62.5 billion), well above the roughly 77 trillion won analysts expected.
• Shareholder returns: a 2026 programme worth more than 100 trillion won (about $75 billion).
• Outlook: the company expects AI memory demand to outpace supply through 2030.

The stock jumped 10.9%, Samsung rose 9%, and the Kospi soared 6.63% to 6,012.44. Foreign investors were net buyers.

Even so, SK Hynix at 1.49 million won is still roughly 40% below its June record. The Kospi is 34% below its peak while still being up about 43% this year. That's the strange math of a market that ran too far, too fast, on a trade that's still fundamentally profitable.

Verdict: a bubble-sized crash in prices, but not a burst in the underlying business. We covered the earlier stages in the Kospi bear market and the AI memory trade and CXMT's debut and the Kospi's 10.84% crash.

Japan and Taiwan: wobbles, not wipeouts

Japan's Nikkei 225 hit a record of about 72,366 on June 25. It fell 4.9% on October 15 and spent three weeks drifting lower on AI-spending worries. At around 64,000 on October 28, it's roughly 12% below its high. That's a correction. Japan's AI exposure runs through chip-equipment makers like Advantest and Tokyo Electron and investors like SoftBank, but the index is much more diversified than Korea's.

Taiwan's main index fell about 3.4% on October 15, yet TSMC, the world's biggest chipmaker, guided fourth-quarter revenue to $52.3–$54.3 billion and said AI chip demand remains strong.

Verdict: no burst.

What triggered the October selloff

The worst day, October 15, came from a pile-up of bad news that started in the US:

1. Hot inflation. US consumer prices rose 3.7% in September, above the 3.6% forecast. That raised odds of more Fed rate hikes after September's increase. Higher rates hurt expensive growth stocks most.
2. AI debt stress. Bonds of AI cloud provider CoreWeave fell into distressed territory, with some yields near 15%. Research firm Gartner warned that more than 80% of "neocloud" companies might not survive past 2030 at the current spending pace. See CoreWeave's bonds now yield nearly 15%.
3. Slower AI revenue growth. Reports that leading AI labs were growing more slowly than investors hoped hit sentiment, after OpenAI's revenue came in $20 billion below expectations earlier in the month.
4. Chinese competition. Progress by China's CXMT in advanced memory raised fears about future memory chip prices, a direct threat to Samsung and SK Hynix.

Asia fell harder than the US because its markets are more concentrated. When two companies make up a huge share of the Kospi, a memory-chip scare becomes a national market crash.

What it means for you

Asia isn't one trade. China's AI stocks were a domestic speculative mania. Korea's and Taiwan's are tied to real orders from US tech giants. A broad Asia or emerging-markets ETF holds both, which is why it fell far less than any single market.

Concentration cuts both ways. The same two stocks that pushed the Kospi up 75% in 2025 and past 9,000 in 2026 pulled it down 39%. If you own a single-country fund, know what's inside it.

A bounce is not a bottom. Two big up days don't confirm that the worst is over. Markets often rally sharply inside a longer decline.

Will this happen to the US?

This is the question most readers are really asking. So far, Wall Street has had a scare, not a burst. The S&P 500 is within about 1% of its all-time high, and the Nasdaq was about 6% below its record as of October 26. Reuters called the US move a "mere prick" compared with China.

There are good reasons the US looks different. Its biggest AI spenders, Microsoft, Alphabet, Amazon and Meta, earn enormous profits, unlike many of China's money-losing AI listings. US valuations, while high, are nowhere near China's 100-to-250 times earnings.

But there are warning signs too. US stocks are among the most expensive in history on long-term measures. Big Tech is increasingly paying for data centres with debt. And the Fed is raising rates, the classic trigger that ended past bubbles.

We go through all of it, the warning signs, the counter-arguments and what history says, in our companion piece: Will the AI bubble burst in the US stock market?

Frequently asked questions