The alpha isn't in the timeline — it's in the layers.
Dhaval Joshi, chief strategist at BCA Research, just dropped a bombshell on the AI narrative. His core thesis: there is no single AI bubble about to blow. Instead, we're witnessing a "rolling bubble" — a sequence of local overheatings that shift from one tech layer to the next. Capital misallocation is the real risk, not a sudden crash.
Let me slow down and unpack why this matters for every crypto operator watching the AI-crypto convergence.
Context: Why This Analyst? Joshi isn't some retail Twitter influencer. BCA Research has been serving institutional investors for over 50 years. When a macro strategist of his caliber warns about capital misallocation in AI, the crypto market should listen — because the same structural dynamics play out in DeFi, NFTs, and L1s. I've seen this pattern before. Back in 2017, during the ICO boom, I audited whitepapers for BatCoin and saw capital flow sequentially from infrastructure projects (Ethereum killers) to application layers (gambling dApps). The rolling bubble phenomenon is not new; it's just dressed in AI clothes.
Core: The Four-Layer Rollercoaster Joshi's model maps neatly onto the AI tech stack: - Infrastructure (GPUs, data centers, chips) - Model (foundation LLMs like GPT, Claude) - Tooling (frameworks, middleware) - Application (SaaS, vertical solutions)
Each layer gets its own bubble phase. Capital pours in, valuations inflate, then the hype migrates to the next layer. The first wave was Nvidia and GPU providers. Then OpenAI and Anthropic. Now we're seeing tools and platforms. What's next? Applications. The capital misallocation Joshi warns about means that at any given moment, some layer is overvalued while another is undervalued — but the total system doesn't collapse because the hot money keeps rotating.
From my experience running DeFi meetups in Tallinn during Summer 2020, I saw the same rotation: first liquidity mining (infrastructure), then lending protocols (models), then yield aggregators (tools), then NFTs (applications). Each phase created a local bubble, but the whole ecosystem survived because the narrative kept shifting. The AI parallel is uncanny.
Contrarian: The Real Risk Isn't a Crash — It's a Slow Bleed Here's the counter-intuitive angle: The rolling bubble actually delays the big crash. But it doesn't eliminate it. What it does is create a false sense of security. Investors see one sector falling while another rises, so they think "the market is fine." But the underlying capital misallocation accumulates. When the rotation eventually stops — because the macro environment shifts (rate hikes, geopolitical shock) or because the next layer can't absorb enough capital — all layers correct simultaneously. That's the real tail risk.
Moreover, the crypto market is tightly coupled. AI bubbles rolling means capital could spill over into crypto as the next "application layer" for AI agents, decentralized compute, or tokenized models. But it also means that when the AI bubble finally rolls into crypto, the euphoria will be intense but short-lived. I've seen this happen in the NFT hype of 2021: celebrity endorsements drove a 300% surge in BAYC volume, but the underlying social sentiment was fragile. The alpha isn't in the timing — it's in understanding the layer position.
Takeaway: What to Watch Next Joshi's framework gives us concrete signals to track: 1. GPU rental prices (H100 spot) — real-time thermometer for infrastructure layer. 2. Model company funding rounds — if OpenAI struggles to raise at a higher valuation, the model layer is topping. 3. Cloud CAPEX growth — if Microsoft/Google slow down, the rotation is finishing.
The next hot layer? Application AI. And that's where crypto-native projects like decentralized AI inference or tokenized models could get their moment. But be careful: the bubble is rolling, not exploding. It's a slow bleed, not a sudden pop.
The alpha isn't in the timeline — it's in the layers. Watch the rotation, not the noise.
