Key Points
- Chinese AI stocks have lost 40% since June, with recent IPOs like Unitry down 50% from debut prices, as investor appetite collides with limited domestic use cases.
- China's $100 billion AI market lacks the capital infrastructure of America's $4–5 trillion ecosystem, with capital controls and real estate-focused wealth limiting foreign and domestic funding.
- Chinese manufacturers show minimal productivity gains from general-purpose AI compared to white-collar work, making language models a harder sell in an economy less dependent on knowledge work.
Summary
Chinese AI Stocks Crater 40% as Domestic Bubble Deflates
Chinese AI companies have lost roughly 40% of their value since June, signaling a sharp reversal in what appeared to be irrational exuberance around the sector. The selloff has hit recently public companies particularly hard. Unitry, which IPO'd a few months ago, is down roughly 50% from its debut price.
The collapse reflects several structural headwinds that differ markedly from the US AI narrative. Chinese AI companies lack the American advantage of distributed global capital—venture firms, sovereign wealth funds, pension funds, and retail investors all feeding into a $4–5 trillion combined market cap for American AI labs. China's equivalent sits at roughly $100 billion, a 40- to 50-fold gap.
Domestic capital flows are thinner by design. Most Chinese wealth remains tethered to real estate rather than equities, which removes a key wealth-effect pump that exists in the US. Capital controls also make it materially harder for international funds to deploy fresh dollars into Chinese AI companies, and that foreign money has largely dried up anyway. Raising even $100 billion in Chinese equity, as one speaker notes, is far harder than mobilizing that sum from American, European, Canadian, and Middle Eastern sources.
The more fundamental question is whether Chinese AI adoption will follow the American pattern at all. The US AI narrative rests on white-collar productivity—legal work, finance, spreadsheets, word documents. China's economy remains more manufacturing-driven, which has proven stubbornly resistant to language-model diffusion. While machine learning and image recognition have already embedded into manufacturing workflows, a general-purpose chatbot's productivity gains are marginal in an economy where white-collar work is a smaller share of GDP.
Whether Chinese labs face a ban or simply face worse unit economics from open-source competition remains unclear. Either way, the gap between investor appetite and actual diffusion use cases appears to have finally closed.
Every deal, every interview. 5 minutes.
TBPN Digest delivers summaries of the latest fundraises, interviews and tech news from TBPN, every weekday.