The consensus on AI valuations is a monolith: a single, overinflated balloon waiting to pop. But in a recent analysis, BCA Research's Dhaval Joshi dismantled this binary. His thesis — that AI is experiencing a 'rolling bubble' — is the most structurally sound perspective I've encountered since auditing the 2017 ICO whitepapers. The balloon doesn't burst; it deflates in layers, each one passing the torch to the next. And if you're watching from the crypto side, you already know the final stop on this roller coaster.
This isn't about a crash. It's about a cascade. The thesis held firm when the charts turned red.
Context
Joshi's framework is a direct response to the 'AI bubble imminent' panic that has gripped institutional investors since late 2023. The prevailing narrative was binary: euphoria then collapse. But the data told a different story. Capital flows weren't uniform across the AI stack; they were rotating. From Nvidia's GPU infrastructure (2023) to foundation model funding (OpenAI, Anthropic in 2024) and now into application-layer plays (Palantir, etc.). Each phase saw a surge, a peak, then a relative cooling as money pivoted to the next narrative. This is not a single bubble. It's a rolling sequence of localized overvaluations, each one feeding the next.
This structural pattern is familiar to anyone who tracked the 2020 DeFi composability risks. Just as flash loans cascaded across protocols, AI capital is cascading across layers of the tech stack. The 'capital misallocation' Joshi warns about is the systemic flaw — the equivalent of a smart contract vulnerability in the macroeconomic code.
Core
The core insight is the mechanism of the rolling bubble itself. It operates on a four-layer stack: infrastructure (chips, data centers), models (foundation LLMs), tools (development frameworks, middleware), and applications (industry solutions). Each layer has a lifecycle: capital inflow → valuation spike → narrative saturation → capital outflow to the next layer. The key is that the outflow doesn't crash the previous layer completely; it deflates it to a sustainable level, while the new layer inflates.
Based on my audit experience of tokenomics during the ICO boom, I see a direct parallel in the 'narrative rotation' of crypto cycles. 2017: ICOs → 2019: DeFi → 2021: NFTs → 2023: AI meme coins. The pattern is consistent: new narratives emerge to absorb excess liquidity, preventing a systemic collapse. The 'rolling bubble' is the market's way of delaying the inevitable reckoning by constantly finding new stories to believe in.
But the risk is in the compounding. Each rotation leaves behind a residue of capital misallocation. The GPU farms built in 2023 may be partially underutilized, but they still have salvage value. The real danger is when the narrative fails to find a new layer. According to my analysis, the current cycle is at the tail end of the model layer, with capital starting to rotate into applications. But the application layer is the most fragile — it requires proven ROI, which is still scarce. The next stop, if the application layer fails to deliver, could be a rotation into adjacent hype cycles: AI agents, autonomous systems, or — most likely — crypto.
Contrarian
The contrarian angle is that the rolling bubble is not a bug; it's a feature. The infrastructure layer (Nvidia, data centers) is overvalued relative to current demand, but those assets are not worthless. They have long-term use value. The 2000 internet bubble left behind a fiber optic network that eventually powered the next decade of growth. AI infrastructure could be the same. The true blind spot is the assumption that the bubble will end in a single crash. It won't. It will deflate layer by layer, and the most dangerous position is being short the entire AI sector — because a new narrative will always emerge to lift the next layer.
But here's the crypto twist: if the AI application layer cannot sustain its valuation, capital will seek the next narrative. And that narrative is the convergence of AI and on-chain economics. AI agents using crypto for autonomous transactions, decentralized verification markets, AI-driven DeFi strategies. This is not a pipe dream; it's a logical extension of the rolling bubble. The 's chaos.' of AI will meet the 's chaos.' of crypto. The market is already pricing this in — look at the recent surge in AI-related tokens (FET, AGIX, RNDR). The rotation is happening.
Takeaway
The rolling bubble implies that the final collapse, if it comes, will be a multi-step process, not a single event. For crypto investors, the signal is clear: the next narrative pad is already being prepared. The question is not whether AI valuations will correct, but which layer will absorb the next wave of capital. If you're not watching the on-chain activity of AI agent wallets, you're already behind. The narrative shift is imminent. Watch the volume.
s whitepaper vs. technical reality — the gap is where the opportunity lies.
