We didn’t see it coming. But the data aggregates told a different story.
A cluster of Chinese hedge funds—previously heavy in Nvidia and the US hyperscalers—has rotated. The call? AI is a "super bubble." Not a correction. Not a pullback. A structural overvaluation of the entire infrastructure layer. The signal is sharp: sell the picks-and-shovels, buy the broader tech ecosystem. But what does "broader" mean when the narrative is still rewriting itself?
This isn’t a tweet. It’s a capital flow signal. And liquidity pools don’t lie.
Context: The Infrastructure Mirage
Since 2023, the AI narrative has been a one-way bet: buy the compute, buy the cloud, buy the GPU. Nvidia’s datacenter revenue grew 100%+ year-over-year for multiple quarters. The four hyperscalers—Microsoft, Google, Amazon, Meta—pushed combined annualized capital expenditure above $200 billion. The market priced in a future where AI adoption would be exponential, endless, and profitless.
But the code is law, and liquidity is the final arbiter. The capital flowing into GPU clusters and cloud data centers was not generating proportional returns. AI-related revenue for these hyperscalers remained in the single digits as a percentage of total revenue. The gap between expectation and reality widened. The "super bubble" label emerged not from price action alone, but from the dissonance between narrative pricing and cash flow reality.
China’s hedge funds are often early movers. They have a track record of rotating before the broader market catches up. In 2020, they sold Chinese tech before the crackdown. In 2021, they exited NFT plays before the crash. Now, they are signaling a shift in the AI narrative. The question is: are they right, or are they just front-running a narrative shift that hasn’t happened yet?
Core: The Narrative Mechanics of the Super Bubble
To understand the rotation, we must map the resonance of the super bubble narrative. It’s not just Nvidia’s P/E ratio (which, at its peak, exceeded 100x trailing earnings). It’s the behavioral resonance of everyone believing the same story: that AI infrastructure is the only safe bet.
Congestion and Contagion: The market became crowded. Institutional investors piled into the same names. The top 10 stocks in the S&P 500 became dominated by tech mega-caps, with combined weight approaching 30%—a level not seen since the 2000 dot-com bubble. The Chinese hedge funds recognized that the marginal buyer was exhausted. When everyone is long, the only direction is down.
Historical Resonance: The 2000 telecom bubble is a mirror. Then, the narrative was "fiber is the new oil." Companies like Cisco and Lucent rode the infrastructure buildout. But the actual demand for bandwidth lagged by years. The infrastructure providers saw their stocks collapse by 60-90% before the application layer (e-commerce, streaming) finally monetized the pipes. The AI super bubble follows the same blueprint: compute infrastructure is the new fiber, and the application layer is still a ghost.
The Geopolitical Layer: Chinese hedge funds face a dual constraint: capital controls and geopolitical risk. The rotation out of US AI stocks is not purely a valuation call. It’s also a hedge against potential escalation of US-China tech decoupling. By selling Nvidia, they reduce exposure to a single point of geopolitical friction. But the nuance is critical: they are not leaving tech. They are diversifying into what they believe is the next wave: AI application, edge computing, AI security, and even domestic Chinese AI chips (like Huawei’s Ascend). The shift is structural, not bearish.
The "Super Bubble" Discourse: The term "super bubble" is a narrative weapon. It signals that the overvaluation is not just a PE expansion but a systemic misallocation of capital across the entire AI ecosystem. The Chinese hedge funds are using it to justify their rotation—and to trigger a cascade. If other funds follow, the sell-off becomes a self-fulfilling prophecy. The bug wasn’t in the code. It was in the narrative.
Contrarian: The Liquidity Is Still There, But It’s Moving
Here’s the counter-intuitive truth: the rotation out of infrastructure doesn’t mean the AI bubble is bursting. It means the liquidity is relocating to a different layer of the stack. The super bubble label is a misdirection if applied to the entire AI sector. The real bubble is concentrated in the narrowest part of the value chain: the hardware and cloud providers that are pricing in a decade of growth in two years.
The Application Layer is Undervalued: AI SaaS companies like Adobe, ServiceNow, and even smaller players like C3.ai are still trading at multiples that reflect skepticism. The Chinese hedge funds are betting that the next wave of AI value creation will happen at the application layer—where AI actually drives revenue and margin expansion. They are not fleeing AI. They are repositioning for the next narrative cycle.
The Infrastructure Sell-Off Is a Second-Order Effect: If Nvidia and the hyperscalers correct, the capital freed up could flow into AI startups. The best time to build infrastructure is during a bear market. The 2000-2002 period saw the birth of Amazon Web Services (AWS) and the rise of Google’s search dominance. The current AI infrastructure sell-off could be the seed for the next generation of AI applications. The narrative is not dying; it’s adapting.
The Bug Wasn’t the Valuation. It Was the Consensus: The real error was not overpaying for Nvidia. It was assuming that the infrastructure layer would maintain its pricing power indefinitely. The bug was in the market’s cognitive bias: treating a temporary supply bottleneck as a permanent competitive advantage. As CoWoS packaging capacity expands and AMD’s MI300X gains traction, the scarcity premium evaporates. The Chinese hedge funds are simply recognizing that the code of the market is rewriting itself.
Takeaway: The Next Narrative Is Already Being Written
Where does the liquidity go next? Into the unknown. The Chinese hedge funds are signaling that the AI narrative is fracturing. The infrastructure layer is losing its monopoly on attention. The application layer, the edge, the security stack—these are the new frontiers. But the timing is uncertain. The super bubble label might be premature, or it might be exactly right.
The question is not whether AI is a bubble. It’s whether you are positioned for the next narrative or stuck in the last one. The liquidity pools are moving. Are you?