China Hedge Funds Warn Global AI Stocks Are a ‘Super Bubble’

Two of China‘s best-known hedge fund managers have told clients that the global rally in artificial-intelligence stocks has crossed from exuberance into what they are calling a “super bubble,” a warning that has already rattled semiconductor markets from Seoul to Santa Clara. Wealspring Asset, founded by Yang Dong — a manager credited in China with correctly calling the peak of the 2007 bull market — and Shanghai Banxia Investment Management Center issued the warnings in investor letters that quickly circulated beyond their client base.

The letters carry weight precisely because of who wrote them. Fund managers who navigated China’s own boom-and-bust cycles are now applying the same skepticism to a global AI trade that Western allocators have largely treated as a structural, multi-year growth story rather than a bubble in the classical sense.

The Case for a ‘Super Bubble’

Yang Dong‘s Ningquan Asset — the vehicle behind the most quoted warning — argued in its H1 2026 investment report that global AI stocks have formed a bubble condition with a collapse point that “may not be far away,” according to reporting from KuCoin’s news desk. The fund went further, projecting that a substantial share of the most popular AI-linked A-share stocks could fall by 80% or more once sentiment turns.

Wealspring, which manages more than $1.4 billion in assets, framed its skepticism around business fundamentals rather than pure valuation math. The firm argued that many of China’s AI infrastructure companies lack a durable competitive moat, run comparatively ordinary business models, and require continuous capital expenditure just to sustain current growth rates, according to Bloomberg’s original reporting carried by Yahoo Finance. The firm drew an explicit parallel to China’s 2015 equity bull run, describing current buying patterns in domestic AI infrastructure names as reminiscent of the “brainless buying” that preceded that crash.

Shanghai Banxia, a smaller fund managing roughly $294 million, took a different angle, pointing to a specific and testable trigger outside China’s borders: pressure on Anthropic‘s revenue growth trajectory. Banxia predicted that Anthropic’s annualized revenue run-rate — a metric closely tracked by AI bulls as a proxy for enterprise adoption — will fall short of market expectations as large technology companies push back against rising token costs and as competitors erode the company’s standing among software developers.

Market Reaction Has Already Arrived

The warnings did not stay confined to investor letters. Global chip stocks fell sharply in the days following the letters’ circulation, with the Nasdaq Composite dropping 2.2% on June 23 and South Korea‘s KOSPI sinking nearly 10% — a decline severe enough to trip a circuit breaker for the first time since March, according to analysis published by NAI 500. Micron Technology plunged more than 13% in the same window, and Nvidia slid as investors reassessed whether AI infrastructure capital expenditure could continue delivering earnings growth commensurate with its valuation.

The severity of the Asian sell-off reflects the region’s outsized exposure to the AI hardware supply chain. South Korea’s chip-heavy index had surged nearly 100% earlier in the year, powered by a rally in SK Hynix and Samsung Electronics, making it disproportionately vulnerable to a sentiment reversal. China’s own CSI Artificial Intelligence Index had climbed more than 35% year-to-date heading into the warnings, far outpacing the roughly 5% gain in the broader Chinese benchmark — a valuation gap the hedge funds argue is unsustainable.

At least four additional Chinese hedge funds expressed reluctance around AI exposure in a monthly summary of fund positioning compiled by CSC Financial Co., with only four funds registering a positive stance and seven declining to take one at all — evidence that the skepticism extends well beyond the two most-quoted names.

A Test of Who Is Early Versus Who Is Right

The central tension in the AI bubble debate is not whether artificial intelligence will reshape enterprise software and global productivity — most market participants, bullish and bearish alike, accept that premise. The dispute is whether current public equity valuations have already priced in an adoption curve, margin structure, and pricing power that has not yet been proven at scale. As framed by NAI 500’s analysis, the AI trade has moved from “look what this model can do” to “show us the business case” — a materially higher evidentiary bar for markets to clear.

Institutional voices remain split. The Bank of England warned in prior analysis that AI-linked equities had become a growing share of total US market capitalization, with some valuation metrics approaching dot-com-era extremes, while Morgan Stanley‘s 2026 outlook estimates that nearly $3 trillion in AI-related infrastructure investment could still flow through the global economy by 2028 — suggesting the capital expenditure cycle, whatever its near-term valuation risk, is far from complete.

Why the China Angle Matters Globally

What distinguishes this warning from generic bubble commentary is its origin. Yang Dong‘s track record calling the 2007 peak gives his current call outsized credibility inside China’s domestic investor base, while Banxia‘s Anthropic-specific thesis offers international investors a concrete, trackable metric rather than an abstract valuation argument. Because Anthropic remains a private company, the revenue data underpinning Banxia’s thesis is not independently auditable — a caveat that tempers, without eliminating, the weight of the warning.

For investors and strategists tracking Asia’s exposure to the AI capital cycle, the practical takeaway is that the region’s chip manufacturers, foundries, and AI infrastructure suppliers now carry two distinct risk vectors simultaneously: the conventional cyclical risk of semiconductor demand, and a newer, sentiment-driven risk tied directly to whether frontier AI developers can convert capital expenditure into durable revenue before investor patience runs out. The next disclosed revenue milestone from a major AI lab, whichever company reports it first, is likely to become the market’s de facto referendum on which side of this debate was correct.