In this week's video, I argue that markets are at a clear inflection point in the AI trade, moving away from software abundance and toward physical scarcity. While investors are trying to short semiconductors and bottom-fish software, leadership has shifted decisively to energy, materials, and industrials, especially silver, copper, and critical minerals. These are not cyclical trades, despite their history, but structural ones, driven by the physical requirements of AI: power generation, grid infrastructure, advanced chips, memory, and industrial inputs and the Jensen Huang declaration of $85T of infrastructure buildout over the next 15 years. Silver, in particular, is reframed not as a precious metal but as a critical industrial input across AI factories, data centers, drones, robotics, EVs, and military systems, making demand price-inelastic. This physical upgrade cycle is still in its early innings and is expected to dominate returns for years.
Hardware bottlenecks and geopolitics define the next decade
AI’s true constraint is not algorithms but the conversion of raw energy into intelligence, which occurs through two bottlenecked pathways: power infrastructure (turbines, transformers, switchgear) and compute infrastructure (GPUs, high-bandwidth memory, advanced packaging). Massive global investment, potentially tens of trillions of dollars, is required to build this stack, and shortages in copper, silver, energy equipment, and memory will persist for a decade or more. This dynamic is deeply geopolitical, driving competition over critical mineral supply chains in regions like Greenland, Venezuela, Brazil, and emerging markets broadly. As a result, materials, energy, EM equities, and small caps are structurally under-owned and mispriced relative to over-concentrated U.S. large-cap tech.
Software disruption, agentic AI, and delayed monetization
On the software side, agentic AI and “vibe coding” are exploding application supply, undermining the traditional SaaS model built on standardized workflows and seat-based pricing. Enterprises increasingly demand fully customized workflows, analogous to “one drug per patient”, which rigid SaaS platforms struggle to deliver. This creates structural margin pressure, longer sales cycles, and valuation risk, even if revenues don’t collapse immediately. At the same time, AI capability is advancing far faster than enterprise adoption, creating a near-term monetization air pocket despite enormous long-term potential. The winners will be firms that deeply integrate AI into core workflows (the top ~12% of adopters), while investors should favor scarcity-driven assets tied to the physical AI buildout and remain patient on crypto, which the speaker believes will eventually mirror the parabolic moves seen in commodities.
Timestamps
00:00 — Why this week marks a turning point in the AI trade
02:05 — Markets are flat, but leadership is quietly shifting to materials and energy
05:15 — Software abundance vs physical scarcity: the core investment framework
07:18 — Critical minerals, geopolitics, and the new AI arms race
09:52 — Silver is not a precious metal anymore — it’s an AI industrial input
12:07 — Jensen Huang on the $85T AI infrastructure buildout
14:44 — Why materials and energy are structurally under-owned
16:27 — Emerging markets and Brazil as critical-mineral winners
19:43 — The two bottlenecks: power infrastructure vs compute infrastructure
23:52 — Why agentic AI and vibe coding threaten traditional SaaS
30:12 — Adoption, not technology, is the real AI monetization bottleneck
48:32 — Bitcoin, crypto structure, and why patience matters