Via nvidia.com
Prominent funds including Wealspring Asset warn investors that conditions for an AI stock collapse have already begun to materialize
Two of China’s most vocal hedge funds are telling their investors to brace for impact. Wealspring Asset, managing roughly $1.4B, described the current AI trade as a “super bubble” in a recent letter to clients, warning that its collapse could arrive sooner than most expect.
Shanghai Banxia Investment Management Center, a smaller fund with approximately $294M under management, reached a similar conclusion through different evidence. The firm pointed to slowing revenue growth at Anthropic as one of several signs that the conditions necessary for the AI bubble to burst have already started forming.
The great rotation #
Both funds have been reducing their exposure to the stocks that defined the AI trade: Nvidia and the US hyperscalers, meaning Microsoft, Amazon, Alphabet, and Meta. The pivot isn’t necessarily out of AI entirely, but rather into the wider ecosystem of data center enablers and supply-chain companies that sit further from the spotlight.
AI-focused hedge funds saw their average excess returns decline significantly during the first half of 2026, according to industry data.
Wealspring’s founder, Yang Dong, carries a particular credibility on bubble calls. He previously predicted the peak of the Chinese stock market in 2007.
July’s drawdowns sharpened the pain #
The warnings didn’t arrive in a vacuum. July 2026 brought steep drawdowns for Chinese hedge funds with heavy AI stock exposure, as prices in the sector whipped around with unusual violence. Both quantitative and discretionary managers reported significant losses during the sell-off.
Some funds responded by cutting AI positions outright. Others chose to inject proprietary capital to stabilize their portfolios, essentially using the firm’s own money to avoid forced liquidations at the worst possible prices.
Diversification as a thesis, not just risk management #
The rotation these funds are describing isn’t purely defensive. Data centers require power systems, specialized construction, networking equipment, and physical real estate. The supply chains feeding into AI deployment touch semiconductor packaging, advanced memory, liquid cooling technology, and electrical grid infrastructure.
Neither fund disclosed specific trade details or dollar amounts, so the scale of the rotation remains unclear.
Disclosure: This article was edited by Editorial Team. For more information on how we create and review content, see our