AIANALYSISPODCAST
Macro investor Jordi Visser re bought Micron, gave Google's $514B AI bet 75% odds, and is watching the pending U.S. Clarity Act vote as crypto's next swing.
Visser told Anthony Pompliano on Podcast Alpha that the 7-8x returns AI infrastructure gave early investors are gone for good. He is not calling the AI top. He is calling a regime shift, from moonshot model bets to the physical bottlenecks underneath them.
The mechanism is the shift from model-level to physical-infrastructure-level exposure. The first wave of AI trades was about picking which model lab would win. That bet is now being repriced. As routers that pick the cheapest model for each query become standard, raw token demand plateaus. Anthropic, a leading AI model lab competing with OpenAI and Google, is decelerating rather than plateauing, Visser argues, because the industry is shifting from token-maxing to token-efficiency (around Jordi @ 17:56).
That re-pricing is pushing capital toward the physical layer. Visser re-bought Micron last week, framing memory as the real AI bottleneck and a way to play an AI-anxiety bottom in memory names. Micron's high-bandwidth memory (HBM) chips are the constrained supply line behind every training cluster and inference server; when model demand plateaus, memory demand does not (around Jordi @ 43:42).
Visser's own calibration of the largest AI bet on the board is more skeptical than his tone. Google's roughly $514B cloud backlog justifies the negative free cash flow, he says, but his AI audit assigns only a 75% probability that the hyperscaler capex bet pays off. He is not short the trade. He is sizing it like an options book: 75% odds, not 100% conviction (around Jordi @ 21:02).
The same bottleneck-to-verdict logic is already playing out in crypto. Visser frames Ethereum's breakout against Bitcoin's stall as a rotation facing a regulatory verdict. The Clarity Act, the pending U.S. digital-asset market-structure bill whose vote is the next swing catalyst for the whole complex, is the event that settles the rotation. The thesis is the same one applied to a different market: when the easy trade is gone, exposure migrates to the most constrained input and waits for the verdict that unlocks it.
Visser's calendar says the verdict is not far off. The Clarity Act vote is the next dated catalyst, and Micron's HBM supply is the next physical one. The AI trade is not ending. It is moving down the stack, from leaderboards to memory modules, and from moonshot multiples to options-book odds.