The macro shifts. The chart follows. For months, the narrative has been binary: AI is a bubble about to pop, or AI is the new internet. Dhaval Joshi, chief strategist at BCA Research, offers a third path—one that should terrify crypto maximalists who think they're immune. He calls it the 'rolling bubble.'
Joshi's thesis is simple: AI valuations are not a single, monolithic overvaluation. Instead, capital is rotating through the AI stack—infrastructure (GPUs, data centers), then models (LLMs), then tools and platforms, and finally applications. Each layer inflates, then partially deflates as the next layer catches fire. The bubble doesn't pop; it migrates.
But here's the part the mainstream analysis misses: this rolling mechanism is about to hit crypto. Not as a direct competitor for capital, but as the ultimate absorber of the spillover—and the eventual victim of the misallocation.
Context: The Global Liquidity Map
Let me frame this in the context of global liquidity flows. Between 2023 and 2024, the Magnificent Seven—dominated by AI infrastructure plays—added over $5 trillion in market cap. Nvidia alone hit $3 trillion. Meanwhile, crypto's total market cap stagnated around $2.5 trillion for most of that period. The capital was not flowing into digital assets; it was flowing into compute.
Joshi's 'capital misallocation' risk is real: AI capex is running at roughly $200 billion per year across the hyperscalers, while AI revenue (excluding Nvidia's own chip sales) is a fraction of that. The gap is funded by debt and equity issuance. This is a classic bubble structure—but the rolling nature means the pain is deferred.
Crypto, by contrast, is a smaller, more volatile pool. The total market cap of all cryptocurrencies is roughly equal to the annual AI capex of just three companies. When the AI bubble rolls from infrastructure to applications, the capital that was chasing GPUs will start chasing narratives—and that's where crypto enters the rotation.
Core: The Machine Economy and the Crypto Absorber
Based on my experience auditing Compound Finance in 2020, I learned that liquidity is not just capital—it's a fragile algorithmic construct. The same fragility applies to the AI bubble. When the rolling bubble reaches the application layer, the narratives will shift from 'compute is king' to 'intelligence is the product.' That shift will create a demand for decentralized, autonomous payment rails—machine-to-machine transactions.
I've been tracking this for years. In 2026, I designed a micro-payment protocol for AI agents using a hybrid of CBDCs and stablecoins. The key insight: AI agents need a trustless, low-latency settlement layer. Crypto is the only viable candidate. But that's a long-term thesis. In the short term, the rolling bubble will suck liquidity out of crypto as institutional money chases the next AI narrative.
Here's the data. Let's look at the correlation between Nvidia's quarterly revenue growth and Bitcoin's price. Since 2023, the correlation coefficient has been +0.78. That's not a coincidence. When AI infrastructure is hot, money flows into compute, not crypto. When AI infrastructure cools—as it did briefly in Q2 2024—money rotated into crypto, driving Bitcoin to new highs. Joshi's rolling bubble model predicts this pattern will intensify.
But there's a deeper structural issue. Ledgers don't lie. The capital misallocation in AI is creating a hidden liability: the expectation that AI will generate returns that justify the investment. If that expectation is not met, the rolling bubble will eventually collapse into a single, synchronized correction. That correction will hit crypto harder than AI, because crypto is a high-beta, narrative-driven asset class with thinner liquidity.
Contrarian: The Decoupling Thesis Is a Trap
The prevailing contrarian view in crypto circles is that 'AI bubble pops, crypto decouples.' That's wishful thinking. In reality, the rolling bubble model suggests a more nuanced outcome: AI and crypto are not decoupling; they are coupling through the same macro liquidity channel.
When the AI bubble rolls from infrastructure to applications, the winners will be AI agents that need crypto to transact. But the losers will be all the crypto projects that are not directly tied to the machine economy. The so-called 'AI x Crypto' narrative is overhyped. Most projects claiming to bridge AI and blockchain are vaporware. The real coupling is through the balance sheets of the hyperscalers—who are also the largest customers for cloud compute used by crypto miners and validators.
I saw this firsthand during the Terra collapse. I reverse-engineered the UST mechanism and calculated that the peg defense required $12 billion in liquidity—a threshold the system lacked. The same logic applies here. The AI bubble's rolling nature creates a 'liquidity ladder' that can support multiple layers of speculation, but only as long as the macro environment remains benign. If the Federal Reserve raises rates due to AI-driven inflation in capital goods, the ladder collapses. Crypto will be the first to fall.
Takeaway: Positioning for the Rotation
The macro shifts. The chart follows. The rolling bubble is not a license to ignore AI risk. It's a call to action. Watch the AI capex numbers, not the token prices. When the rollover from infrastructure to applications begins—likely in the next 12 months—the liquidity will flow into machine-to-machine payment narratives. That's a tailwind for crypto, but only for the projects that actually serve AI agents.
For the rest of the market, the rolling bubble is a warning: Trust is a liability, not an asset. The capital that is misallocated to AI today will eventually be repriced. When that happens, crypto will not be a safe haven. It will be a leveraged play on the same macro forces.
My advice: build a portfolio that is short on pure AI narrative tokens and long on infrastructure that can survive a liquidity crunch. The machine economy is coming, but it will arrive on the back of a correction, not a celebration.