The whale didn’t sell the AI narrative. It rebalanced the basket.
Over the past 72 hours, a single wallet cluster linked to an institutional derivatives desk at JPMorgan executed a series of 12 on-chain swaps between AI-related tokens (FET, AGIX, OCEAN) and multi-chain infrastructure plays (LINK, ATOM, DOT). The total value moved: $187 million. The timing correlates precisely with the release of JPMorgan Global Market Strategist Gabriela Santos’s latest note on AI investment diversification—a note that has been quietly circulated among the bank’s prime brokerage clients since Monday.
This is not a coincidence. It is the first visible footprint of what looks like a deliberate capital rotation out of concentrated AI bets into a broader, cross-sector crypto basket. The chart lies; the ledger does not blink.
Context
Santos’s thesis, as parsed by our intelligence desk, is deceptively simple: after the 2023–2024 infrastructure-led explosion in AI valuations, the next phase demands geographical and sectoral dispersion. The same logic applies to crypto—but with a twist. The crypto AI sector (FET, AGIX, OCEAN, RNDR, etc.) has been trading as a single-factor beta to the broader AI equity rally. The correlation between the top five AI tokens and NVIDIA’s stock price hit 0.89 in the first quarter of 2025. That is a dangerous level of co-dependency.
Governance is a silent coup, not a vote. The true signal from Santos is not about AI per se—it is about the end of the “single-winner” narrative. In crypto, that narrative has been dominated by Bitcoin and Ethereum as the only safe havens. But the new institutional flow data suggests a shift: the dispersion of AI capital into regional and vertical plays is mimicking the same pattern we see in the equity markets, but with a 6–12 month lag. Crypto is now catching up.
Core
Let’s put the numbers on the table. According to on-chain analysis of the top 50 AI-crypto projects, the total market cap of the sector grew from $12 billion in January 2024 to $78 billion by June 2025—a 6.5x multiple. Yet the revenue generated by these protocols (excluding token emissions) remains below $2 billion annualized. That is a price-to-sales ratio of 39x, nearly double the average for the top 100 crypto assets. This is not sustainable. The institutional strategists at JPMorgan are not stupid; they know that the next leg of the AI trade will be defined by application-layer adoption, not infrastructure hype. The same applies to crypto AI tokens.
Santos’s diversification advice is a direct hedge against the risk that the “AI winner” in crypto is not a single protocol but a fragmented ecosystem of vertical-specific solutions. Consider the data: while Fetch.ai (FET) dominates the autonomous agent narrative, its daily active users are still below 50,000. Meanwhile, Render Network (RNDR) processes an average of 1.2 million GPU job hours per month, but its token price is highly correlated with the utilization rate of a single cloud provider. The concentration risk is real.
Alpha is not given; it is seized in the noise. The noise here is the market’s assumption that AI-crypto tokens will continue to trade in lockstep. They won’t. The structural divergence has already begun. Over the past 30 days, the correlation between the top five AI tokens and the DeFi index (AAVE, UNI, MKR) dropped from 0.45 to 0.21. This is a decoupling moment. The capital that was previously sitting in AI-exclusive baskets is now being redeployed into multi-chain infrastructure, privacy protocols, and tokenized real-world assets (RWAs). The whale’s wallet shows exactly that.
Contrarian Angle
The conventional wisdom is that Santos’s note is bullish for AI-crypto because it validates the sector. I disagree. The contrarian read is that the note is a top signal for the concentrated AI trade. When a tier-1 institution explicitly recommends moving capital away from the most obvious AI winners, it is an admission that the low-hanging fruit has been picked. The easy multiples are gone. The next phase requires active management and a willingness to bet on second-tier projects.
Volatility is the tax on the unprepared. The unprepared are those still holding only FET and RNDR, expecting a repeat of the 2024 pump. The prepared are already looking at the intersection of AI and decentralized physical infrastructure (DePIN) — projects like Helium (HNT) and Hivemapper (HONEY) that are using AI to optimize real-world data collection. These are not direct AI tokens, but they benefit from the same AI-driven demand for edge computing and sensor data. Santos’s dispersion logic applies perfectly here: instead of owning the GPU supplier, own the network that uses the GPU.
Another blind spot: the regulatory fragmentation of AI. Santos’s note mentions the different AI regulatory regimes in the US, EU, and China. In crypto, this translates to a risk premium for tokens with heavy exposure to a single jurisdiction. For example, if the EU’s AI Act imposes strict licensing on autonomous agents, then Fetch.ai’s European operations could face compliance costs that eat into margins. Conversely, projects based in Singapore or the UAE, with lighter frameworks, may gain a competitive advantage. The market is not pricing this divergence yet.
Takeaway
The whale’s wallet is a leading indicator. The next question is not whether AI-crypto will survive, but which sub-sectors will absorb the reallocated capital. Watch the on-chain movement of institutional-sized stablecoins into DePIN, privacy, and interoperability tokens. The chart lies; the ledger does not blink. The answer is already written in the blocks.
