China’s AI bubble is facing a brutal reckoning. The STAR50 technology index has fallen approximately 40% from its July peak, erasing around $480 billion in market value in less than four months, according to Reuters’ October 26 report.
The reversal reflects concerns about stretched valuations, future AI demand and potential restrictions on Chinese technology exports. China’s broader CSI300 index has also dropped roughly 20% from its June highs, undermining hopes that the country’s stock market could deliver a more stable, sustained rally.
The selloff raises a question beyond China: can the profits generated by artificial intelligence justify the extraordinary sums being invested in it?
China’s AI Stock Crash Hits Chipmakers and Robotics Companies
The STAR50 tracks 50 large, liquid securities listed on Shanghai’s Science and Technology Innovation Board. It is a technology benchmark, rather than an index composed exclusively of AI businesses.
The damage nevertheless reaches prominent AI-linked names. Reuters reported that chipmakers MetaX and Moore Threads had lost around half their value since July. Robotics company Unitree had halved since its August debut, while memory-chip producer CXMT had fallen more than 25% over the preceding month.
The concentration was striking: Goldman Sachs figures cited by Reuters put AI-related stocks at 42% of market capitalization and 57% of trading turnover across mainland A-shares and Hong Kong H-shares.
That exposure helps explain why a reversal in one investment theme can spread across portfolios. Investors holding different companies may still be making essentially the same bet on continued AI expansion.
Why AI Bubble Fears Extend Beyond China
A technology can become indispensable while investments in it still lose money. The financial question is whether future earnings will be sufficient to support today’s share prices and infrastructure commitments.
INSEAD finance researchers identify several warning signs: investors extrapolating exceptional growth too far into the future, expensive startups without clear profitability paths, and financing arrangements linking suppliers with their customers. Their analysis also stresses that profitable technology leaders differ from many speculative businesses of the dotcom era.
Consider a hypothetical company whose profits remain unchanged while its valuation falls from 100 times earnings to 50 times. Its share price would halve even though the underlying business had not stopped growing previously or suddenly become worthless.
Competition creates another risk. Cheaper models and more efficient computing can benefit customers while weakening the pricing power of companies whose forecasts assume premium margins.
The bearish scenario therefore does not require AI adoption to collapse. Revenue growth merely falling short of expectations could force investors to reassess valuations.
Hundreds of Billions Are Flowing Into AI
The scale of investment is measurable. Stanford’s 2026 AI Index reports that global corporate AI investment reached $581.69 billion in 2025, including $344.66 billion in private investment. Private funding increased 127.5%, while generative AI investment grew more than 200%. These categories should not be confused with spending solely on data centers.
Infrastructure commitments add another dimension. In September, UBP estimated that Microsoft, Amazon, Alphabet, Meta and Oracle would collectively spend approximately $820 billion in capital expenditure during 2026, against around $750 billion in operating cash flow. These are research estimates, not finalized annual results, and the spending supports broader businesses as well as AI.
Meta’s own results provide a concrete example. The company reported third-quarter capital expenditure of $34.91 billion, including finance-lease principal payments, and negative free cash flow of $2.12 billion. Its full-year spending guidance stands at $135 billion–$145 billion.
Such investment can create valuable capacity. But it also raises the revenue threshold needed to generate attractive returns after equipment costs, electricity, maintenance and financing.
Circular Financing Could Amplify a Slowdown
Another concern involves companies financing businesses that subsequently purchase their products.
In September 2025, Nvidia and OpenAI announced a letter of intent covering at least 10 gigawatts of Nvidia systems. Nvidia said it intended to invest up to $100 billion progressively as deployment occurred. The announcement described a planned arrangement, not an immediate $100 billion payment.
INSEAD researchers highlight why this structure attracts scrutiny: supplier investment can help customers fund purchases from that same supplier. That does not establish wrongdoing or artificial revenue, but it makes independent, paying customer demand especially important.
If fundraising weakens, the effect could spread from AI developers to cloud providers, chip suppliers and infrastructure lenders. UBP similarly identifies timing risk where contractual payments become due before customers generate sufficient revenue.
The most vulnerable businesses would be those combining continuing cash losses, financing dependence and concentrated customers. Exposure to these conditions is more informative than simply labeling every AI company a bubble.
Does China’s Selloff Mean the Global AI Boom Will Fail?
The evidence does not establish that a worldwide collapse is imminent. Stanford reports that 88% of surveyed organizations used AI in 2025, demonstrating adoption alongside the investment surge. Adoption, however, does not automatically translate into profits for every supplier.
Meta also reported quarterly revenue growth of 26% and says AI is improving advertising performance. That claim offers a counterpoint to the idea that all AI spending lacks commercial value.
The next tests are financial: paid usage, margins, cash generation and whether infrastructure commitments can be funded without increasingly expensive capital.
China’s correction shows how quickly confidence can reverse when expectations become excessive. It is a warning about the prices investors pay and the returns companies must deliver—not proof that artificial intelligence itself has failed.