Two Chinese Hedge Funds Warn AI Has Become a 'Super Bubble'

Two of China's best-known hedge fund managers have issued client letters warning that the global AI trade has crossed into "super bubble" territory, with one arguing that the trigger for its collapse has already appeared. Bloomberg reported that Wealspring Asset, founded by Yang Dong, who is well-known in China for calling the top of the 2007 bull market, wrote in an investor letter that AI stocks globally have become a "super bubble" and that "the collapse point may not be far away." Wealspring manages more than $1.4 billion. Shanghai Banxia Investment Management Center, with more than $294 million under management, went further: it said "the trigger for the AI bubble to burst has already appeared," pointing to pressure on Anthropic's revenue growth as the specific evidence.

Banxia's argument centers on Anthropic's annualized revenue run-rate, described as a closely-watched metric for AI bulls, falling short of market expectations. The firm predicted that large tech companies will recoil at the rising cost of tokens while competitors chip away at Anthropic's popularity among developers. That is a testable claim, though Anthropic is a private company and the underlying data is not publicly auditable. Wealspring added a China-specific dimension, comparing current buying in domestic AI infrastructure names to the "brainless buying" of China's 2015 bull market and predicting the hottest Chinese AI infrastructure stocks could crash more than 80%. The CSI Artificial Intelligence Index had been up more than 35% for the year heading into these warnings.

Market moves coincided with the letters circulating. Semiconductors led losses: the Nasdaq Composite fell 2.2% on June 23, South Korea's KOSPI sank almost 10% and tripped a circuit breaker, and Micron Technology plunged more than 13%.

The honest caveat is that Banxia's founder Li Bei was not speaking from undisturbed profits. The firm's flagship fund reportedly crashed 15.03% in the week of June 15 to 18, its worst single-week decline since inception. That does not make the warnings wrong, but it does mean they should be read as analysis from inside a painful drawdown, not detached calls from the sidelines. What the reporting does not give you is detail on how Banxia obtained its Anthropic revenue data, or how broadly other Chinese institutional managers are positioned relative to these letters.

For practitioners, the claim worth tracking is the revenue one: if AI infrastructure spending is not generating commensurate revenue at the application layer, current valuations carry real downside regardless of where individual managers sit. Two credible voices with real capital at stake are now saying the math does not hold up.