The AI Bubble Isn't Bursting—It's Rolling: What Dhaval Joshi's Framework Means for Crypto

The fog is thick this morning. Over the past 72 hours, I've been chasing the alpha through the whispers of Silicon Valley, and what I'm hearing is not the usual 'AI is a bubble about to pop' chorus. Instead, a different narrative is crystallizing—one that maps directly onto the liquidity veins of both the AI and crypto ecosystems. BCA Research's chief strategist Dhaval Joshi has been quietly circulating a framework that flips the conventional panic on its head: the AI bubble isn't a single, monolithic explosion waiting to happen. It's a rolling sequence of mini-bubbles, each inflating and deflating across different layers of the tech stack. And for those of us who live in the crypto wild west, where speed meets substance, this is the signal we've been waiting for. Let me rewind to August 2017. I was a junior financial analyst in Madrid, fresh with an MS in Economics, auditing a whitepaper called 'SkyNet Chain.' I spotted a glaring discrepancy between their projected tokenomics and real-world utility within 48 hours of the presale. I published a data-driven exposé on Medium that went viral, knocking 30% off their presale volume. That experience taught me something: the market's most dangerous illusions are not the ones that are obviously false—they are the ones that are partially true, but only for a certain layer of the stack. The same principle applies to AI today. Joshi's argument rests on the observation that AI's valuation explosion is not a uniform wave. Instead, it's a series of overlapping bubbles that roll through four distinct layers of the AI technology stack: infrastructure (chips, data centers), models (foundation models like GPT-4, Claude), tooling (frameworks, middleware), and application (industry solutions). Each layer attracts capital in a frenzy, builds a narrative, and then—when the next layer's narrative becomes hotter—the capital rolls on, leaving the previous layer to deflate slowly. This is not a crash; it's a rotation. And if you've been in crypto long enough, you've seen this movie before. During DeFi Summer 2020, I was at the Ethereum Community Conference in person, intuitively sensing the momentum behind Compound Finance. I compiled a real-time dashboard tracking collateral ratios and APY spikes, sharing it on Telegram channels. The liquidity flowed from yield farming on Uniswap to lending on Compound to insurance on Nexus Mutual, each layer pumping and then cooling as the next narrative took over. The same rolling pattern is now playing out in AI. Where liquidity flows, value finds its home. Right now, the infrastructure layer—dominated by NVIDIA's GPUs and hyperscale data centers—is the hottest. But Joshi warns that capital misallocation is the hidden risk. I've seen this before. In 2021, during the NFT boom, I focused on the Bored Ape Yacht Club community, not the art. I hosted a Twitter Space with three influencers, discussing floor price dynamics. The social capital was the real asset, and the capital chasing it was misallocated to the art itself. The same is happening in AI: billions are pouring into GPU clusters that may not see full utilization for years, while application-layer startups with real revenue are starved. The rolling bubble hides this misallocation because each layer's temporary boom gives the illusion of endless growth. But here's the contrarian angle that the mainstream financial press is missing: if the AI bubble is rolling, it means it won't pop all at once. That's good news for crypto. Why? Because the rolling bubble creates a 'capital overflow' effect. When the infrastructure layer starts to cool—as it inevitably will when GPU supply catches up with demand—the capital that was chasing chip narratives will look for the next big thing. And that next big thing could very well be crypto-native AI, or even just crypto as a general risk-on asset class. I've seen this pattern before: during the Terra collapse in May 2022, the market was in a deep bear phase. Instead of technical despair, I organized a 'Crypto Survival BBQ' in Madrid, using humor and social connection to cope with a -90% drop. While peers were paralyzed, I wrote a piece on psychological resilience. The point is that capital flows are emotional, not just rational. When one narrative loses its luster, the capital doesn't disappear—it rotates to the next narrative that offers both excitement and potential. AI's rolling bubble will eventually push capital toward crypto, especially DeFi and decentralized AI projects, because they offer the same 'speculative upside' with a fresh narrative. Let me ground this in data. Over the past 12 months, I've been tracking the 'NVIDIA GPU utilization rate' across major cloud providers—a metric I've dubbed the 'Infrastructure Pulse.' Based on my audit experience with ICOs and DeFi protocols, I've learned that the best leading indicator of a bubble top is when capital expenditure significantly outpaces revenue growth. In Q1 2024, the top four cloud hyperscalers (Microsoft, Google, Amazon, Meta) collectively spent over $200 billion on AI infrastructure, while AI-related revenue growth, though strong, still lags at roughly 30% of that CAPEX. This is a classic sign that the infrastructure layer is overheating. But here's the twist: the same thing happened in DeFi in 2021. Total value locked (TVL) surged to $200 billion, but the actual fee revenue from lending and DEXs was a fraction of that. The bubble rolled from DeFi to NFTs to GameFi, and each time, capital found a new home. The infrastructure layer in AI will eventually roll, and crypto stands to benefit. But I'm not here to sell you a simple narrative. The rolling bubble also carries a risk for crypto: if the macro environment shifts—say, interest rates spike or a geopolitical crisis hits—the rolling bubbles could all deflate simultaneously, creating a synchronized crash. That's the 'tail risk' that Joshi's framework doesn't fully address. I've seen this happen in crypto: when the Terra collapse triggered a systemic liquidity crisis, all bubbles—DeFi, NFTs, L1s—popped at once. The rolling nature of AI bubbles might delay the day of reckoning, but it doesn't eliminate it. The key is to watch the 'capital misallocation' metric. If the ratio of AI CAPEX to AI revenue exceeds 5:1 for more than two consecutive quarters, the roll will accelerate, and the deflationary phase will hit hard. So what does this mean for the reader? For the crypto investor sitting on the sidelines, this is the moment to position yourself for the next roll. The AI infrastructure bubble is nearing its peak. The next layer to catch fire will be AI application layer—specifically, decentralized AI (DeAI) projects that combine blockchain's transparency with AI's intelligence. I've been mapping the liquidity veins of this emerging ecosystem, and I see patterns: projects like Bittensor, Render Network, and Akash Network are already showing signs of capital inflow. They are the 'Compound Finance' of this cycle—the early movers that will capture the narrative when the roll happens. To be clear, I'm not predicting a mass exodus from AI to crypto. Rather, I'm reading the pulse of the digital art market, as I did in 2021, and seeing the same psychological patterns. The market is a living organism. It breathes in narratives and exhales capital. The AI bubble is not a disease; it's a digestive process. The capital that is currently bloating the infrastructure layer will eventually be absorbed by the application layer, and if that application layer is decentralized, crypto will be the beneficiary. Let me leave you with a forward-looking judgment. The rolling bubble framework is powerful because it changes the question from 'When will AI crash?' to 'Where is the next roll?' Based on my analysis of on-chain data, GPU spot prices, and venture capital deal flow, I believe the next roll will happen within the next 6–9 months. The infrastructure layer will cool, and the application layer—especially those applications that are built on decentralized infrastructure—will heat up. The crypto projects that are already building AI compute marketplaces, decentralized inference networks, and data provenance layers will be the first to capture the capital overflow. But I'm not just a cheerleader. I'm a skeptic who has been burned by the fog of ICO whispers and the noise of DeFi summer. I know that every bubble, rolling or not, eventually leaves casualties. The question is whether you are positioned to catch the roll before it becomes obvious. The silence before the next pump is the loudest signal. Listen closely.