Two of China’s best-known hedge funds are calling the AI boom an unsustainable “super bubble.” Wealspring Asset says the “collapse point may not be far away.” Shanghai Banxia says “the trigger to burst has already appeared.” Four more Chinese funds agree. The skeptics have arrived β from the country building its own AI stack.
What Bloomberg Reports
Wealspring Asset said global AI stocks have become a “super bubble” and that the “collapse point may not be far away.” Shanghai Banxia Investment Management Center said “the trigger for the AI bubble to burst has already appeared.”
At least four other Chinese hedge funds expressed reluctance around AI, with both Wealspring and Banxia advising investors to be cautious due to concerns over valuations and the lack of a long-term moat for some AI infrastructure companies.
The irony: These warnings come from China β the country that just committed $295 billion to AI infrastructure, trained DeepSeek V4-Pro on Huawei chips, and launched Seedance 2.5 for AI video. China is building the AI stack AND calling it a bubble. Belt and suspenders β or hedging the bet?
The Structural Read
THE “NO MOAT” CONCERN IS THE REAL SIGNAL
The hedge funds aren’t saying AI doesn’t work. They’re saying AI infrastructure companies lack durable competitive advantage. This week’s data supports it: Big 3 token share collapsed 72% β 33% in one year. Open-source is commoditizing the model layer. If models are commodities, the companies selling them at premium prices are overvalued.
THE CONVERGENCE WITH THE $270B ROUT
This Bloomberg piece arrived the same day as the $270B AI rout story. Chinese hedge funds calling bubble + US leveraged ETFs unwinding + SpaceX down 30% + OpenAI delaying IPO. Four independent signals all pointing the same direction: the market is repricing AI from hype to fundamentals.
BUT THE SUBSTRATE STILL GETS PAID
Apple raised MacBook prices because DRAM is up 98%. Xbox went up $150. SK Hynix is IPO’ing at $29.4B. Even if the model layer is a bubble, the substrate layer (memory, chips, power) is capturing real value. The bubble question is about WHICH layer, not whether AI itself is real.
The Bottom Line
Six Chinese hedge funds are calling AI a super bubble. The same week the $270B rout hit AI chip stocks, OpenAI delayed its IPO, and Tim Cook called memory costs “unlike anything in 40 years.” The consensus is fracturing. But the “no moat” concern is the sharpest critique β because the token share data (72% β 33%) confirms that the model layer IS commoditizing. The question isn’t whether AI is real. It’s whether the companies at the top of the stack can defend their margins when open-source alternatives cost 1/6th as much.
Business Engineer
The AI Supercycle β Which Layers Survive the Correction
Read the AI Supercycle βSource: Bloomberg β June 26, 2026









