The AI Super Bubble Is Ready to Burst

The warnings are no longer coming from fringe contrarians. As of late June 2026, two of China’s most prominent hedge fund managers, the Bank for International Settlements, and a growing roster of institutional investors have reached a shared and uncomfortable conclusion: the artificial intelligence investment boom has entered the territory of an unsustainable asset bubble — and the collapse, when it comes, may be swift and severe.

Collapse Point May Not Be Far Away

Wealspring Asset, founded by Yang Dong — a manager celebrated in China for correctly calling the market top in 2007 — declared in a June 2026 investor letter seen by Bloomberg that global AI stocks have become a “super bubble” and that the “collapse point may not be far away.” The firm joins a growing chorus of voices that include another prominent Chinese hedge fund manager who issued similarly stark language to clients.

The warning arrives at a time when AI-related equities have driven extraordinary market gains. Yet the financial infrastructure underpinning that boom is attracting unprecedented scrutiny from the world’s most important monetary institutions.

The BIS Names AI Its Number One Risk

In its 2026 Annual Economic Report, published on June 28, the Bank for International Settlements named an AI capital expenditure bust as one of its top-tier threats to global financial stability — placing it alongside sovereign debt fragility and runaway inflation in a single integrated risk framework. It marked the first time the BIS elevated AI financial risk to a flagship annual publication, rather than a working paper.

The report identified two interconnected vulnerabilities. First, a concentration of AI infrastructure financing in private credit markets, where loans to AI-related companies surged from roughly $3 billion in 2010 to over $40 billion in 2025, according to BIS Bulletin No. 120. Second, a phenomenon the BIS calls “circular financing” — arrangements that blend equity stakes, debt instruments, and supplier contracts in ways that obscure actual leverage.

In these deals, chipmakers and cloud hyperscalers take equity positions in AI laboratories or neocloud providers, which in turn commit to multi-year purchases of chips or computing capacity. Data centre construction is outsourced to third parties that lease facilities back to hyperscalers on long-term contracts. Assets, the BIS warned, may be “pledged multiple times” across these overlapping structures.

“Disappointment in returns could trigger a sudden pullback in financing and turn the capex boom into a protracted investment bust, with potential knock-on effects on financial conditions,” the report stated.

The Scale of the Problem

Independent research has put harder numbers on the exposure. AI capital expenditure drove roughly 74 percent of US GDP growth in Q1 2026, based on analysis of Bureau of Economic Analysis sub-components. That degree of concentration transforms a potential AI capex reversal from a sector-level correction into a macroeconomic event.

The private credit market is showing early stress signals. Many listed Business Development Companies — publicly traded funds that lend heavily to mid-sized technology companies — are now trading 15 to 20 percent below the stated value of their underlying assets. A significant portion of loans held by these funds use payment-in-kind structures, meaning borrowers are rolling unpaid interest into growing debt balances rather than servicing it in cash. The share of PIK arrangements in private credit doubled between 2022 and 2025.

S&P Global has flagged a refinancing cliff ahead: leveraged debt owed by weaker borrowers is projected to surge from $56.6 billion in 2026 to $215 billion in 2028. If those companies cannot roll their debt, the forced asset sales could cascade through markets in ways that resemble — but may exceed — the 2008 shadow banking unwind.

A Transmission Mechanism to Sovereign Debt

The BIS report identified what makes the current configuration particularly dangerous: the same leveraged hedge funds that dominate sovereign bond markets through basis trades are deeply exposed to private AI credit. A shock to non-bank financial intermediaries could force fire sales in government bond markets, creating a feedback loop from tech sector bust to sovereign debt crisis.

Polymarket, the world’s largest prediction platform, placed the probability of the AI investment frenzy bursting before the end of 2026 at 26 percent in mid-June — a figure that has been climbing steadily. An equity crash of the scale comparable to the early 2000s dot-com unwinding would, at current valuations, erase approximately $33 trillion of value, more than the entirety of US GDP.

A Question of When, Not If

Man Group, the London-based alternative investment firm, has put the case with unusual directness. The AI boom is real, it argues, but the financial architecture supporting it is expanding faster than any credible adoption curve can justify. Every major technological revolution — railroads, electrification, fibre optics, the dot-com era — saw the technology endure while the financing cycle broke.

The recursive demand loops in today’s AI ecosystem share structural features with those prior cycles. Training costs are rising exponentially. Marginal improvements increasingly require reinforcement learning approaches that generate more tokens per query, worsening unit economics. The foundational leap that came from scraping the public internet is, by definition, not repeatable.

What remains is a market betting trillions of dollars that commercial AI applications will scale fast enough to justify the infrastructure already built — and the far larger infrastructure still being financed.

The BIS, Man Group, and Chinese hedge fund managers may hold different views on many things. On this, they agree: the bet is far from certain, and the financial system is not prepared for it to fail.