Gavin Baker, whose portfolio holds NVIDIA, Micron, Cerebras, and Astera Labs, argues AI is a full supercycle, and the money is in infrastructure, not software. His thesis: chips, memory, and power generation are the constrained resources, and whoever controls those bottlenecks controls the cycle. The episode maps four specific AI constraints starting at the 11:20 mark, which is where the analytical substance begins.
The hosts position Baker against Leopold Aschenbrenner, who holds a more aggressive version of a similar overarching thesis. That tension is worth your attention. Baker's track record, covered at 1:47, gives weight to his infrastructure-over-applications conviction. His Unity Software position, discussed at 5:26 in the context of world models, is the most contrarian holding and the one that demands explanation.
The dot-com comparison gets addressed directly at 19:08, and Baker's argument for why this cycle is structurally different from 2000 is the section most worth reading in full context. Supply constraints at 24:10 close the loop on why the infrastructure bet is time-sensitive, not evergreen. This is not a hype piece. It is a capital allocation framework from someone with skin in the trade.
[WATCH ON YOUTUBE →]