China's top hedge funds are now openly warning that what they're calling an AI "super bubble" may be entering its most dangerous phase.
This is with one prominent fund manager who correctly called the 2007 market top saying the collapse point "may not be far away."
It's a growing chorus of warnings, suggesting the AI trade may be approaching its breaking point and this time the alarm isn't coming from skeptical outsiders, it's coming from some of the people who have been closest to the money. Yet new figures circulating the same week suggest AI infrastructure spending may be clearing its first real accounting test, complicating the picture for anyone trying to call this top with confidence.
The warnings aren't isolated to China either. They're arriving alongside a pattern of insider selling, billionaire short positions, and a corporate decision from Apple that suggests the pressure has moved well beyond stock valuations and into the physical supply chain itself.
The most attention-grabbing warning comes from Wealspring Asset, a Chinese hedge fund whose founder, Yang Dong, has earned a reputation for calling market tops accurately, most notably in 2007, just ahead of the global financial crisis. According to reports circulating on social media, Yang Dong has now labeled the current AI investment cycle a "super bubble," warning directly that "the collapse point may not be far away."
That kind of language, paired with Yang Dong's track record, has clearly rattled segments of the investment community already nervous about how far AI valuations have run. A fund manager known specifically for identifying bubble peaks before they burst is not a voice the market tends to dismiss easily, regardless of how strong current AI sentiment has been.
If Wealspring's warning was alarming, Shanghai Banxia's was more specific. The fund went further than simply flagging general bubble concerns, stating outright that "the trigger for the AI bubble to burst has already appeared." Crucially, Banxia pointed directly at Anthropic as that trigger, predicting that the company's revenue run-rate will fall short of what the market currently expects.
Naming a specific company as the catalyst for a broader market correction is a notably bold call. It suggests Banxia's analysts believe Anthropic's growth trajectory has been priced into the broader AI ecosystem's valuation in a way that, if it disappoints, could ripple outward across the sector rather than remain contained to a single company's stock or private valuation. At least 6 of 11 Chinese hedge funds currently tracked have no positive stance on AI whatsoever right now, a sentiment shift that has been building for several months. Wealspring went as far as suggesting some of China's hottest AI-linked stocks could crash by more than 80% if the bubble bursts as anticipated.
What makes these warnings harder to dismiss as a regional or culturally specific skepticism is how closely they mirror sentiment building elsewhere. Michael Burry, the investor famously known for correctly predicting the 2008 housing collapse, has disclosed bearish positions against both Nvidia and Palantir. Burry has been particularly pointed about the circular funding arrangements connecting major AI companies, warning that what currently looks like a mutually reinforcing growth flywheel could later be remembered as something closer to "a picture of fraud, not a flywheel."
Insider trading activity adds further weight to the caution. Nvidia insiders have sold approximately $5.4 billion in stock, while Palantir insiders have sold roughly $7.2 billion, even as Palantir's own share price trades well off its recent highs. When insiders at companies sitting at the center of the AI boom are selling at this scale, it tends to raise more questions than reassurances about how those closest to the businesses actually view their own near-term prospects.
Just as the bearish narrative was building steam, a different piece of data emerged that cuts against the simplest version of the bubble story. According to analysis shared by Rohan Paul, AI revenue has crossed what amounts to its first serious accounting test: roughly $25 billion in quarterly sales now exceeds an estimated $21 billion in chip and data-center depreciation for the same period.
In plain terms, that means AI infrastructure is starting to pay for itself before power costs, labor, financing, and lease obligations even get factored into the equation. It's a meaningfully different signal than the pure spending-versus-revenue framing that dominates most bubble arguments, because depreciation specifically captures whether the hardware itself is generating enough return to justify its own wear and replacement cost, independent of the broader operating expenses still left to cover.
This doesn't resolve the debate. Power, labor, financing, and lease costs are not small line items, and clearing the depreciation hurdle alone doesn't mean the broader economics work yet. But it does complicate any narrative suggesting AI infrastructure spending is purely speculative with no revenue backing whatsoever. The hardware, at minimum, appears to be earning back its own cost basis at the current pace of sales.
Even with that nuance, the broader spending math across the industry remains difficult to ignore. OpenAI has openly acknowledged that its current spending trajectory doesn't yet make financial sense relative to its revenue. The company is committing approximately $1.4 trillion over eight years toward data center infrastructure, while currently generating just $13 billion in revenue.
That gap between committed capital expenditure and actual revenue generation is enormous by any standard measure of corporate financial health. It doesn't necessarily mean the underlying technology lacks value, but it does mean the current spending pace assumes revenue growth at a scale and speed that has yet to materialize, and that assumption is precisely what skeptics like Burry and the Chinese hedge funds are now challenging directly.
Perhaps the most telling signal of all has nothing to do with hedge fund commentary and everything to do with corporate behavior. Apple is reportedly lobbying the White House to purchase memory chips from a blacklisted Chinese company, specifically because U.S. chip prices have become too expensive for even Apple to absorb comfortably.
That detail matters enormously. Apple is not a struggling company searching for shortcuts, it's one of the most cash-rich corporations in the world. If Apple is actively seeking sanctioned alternatives to escape AI-driven chip pricing, that's a strong signal the inflation in AI-related costs has moved well beyond speculative stock valuations and into the actual physical hardware supply chain that the entire industry depends on.
This dynamic echoes what happened when DeepSeek delivered its shock to the market last year, proving that comparable AI output could be achieved for dramatically less money, a revelation that wiped roughly $600 billion off Nvidia's valuation in a single trading day. If the world's largest hardware maker is now actively searching for cheaper alternatives to escape AI-driven chip costs, that isn't a minor side story buried in supply chain trade publications. It's confirmation that the bubble, if it exists as described, has inflated far more than just stock prices, it has worked its way into the cost structure of the entire AI hardware ecosystem, even as the revenue side of the ledger shows its first genuine signs of catching up.
This is not trading or investment advice. Always do your research before buying any cryptocurrency or investing in any services. Follow us on X @nulltxnews
