AI CapEx is a highly concentrated bet
The guest argued that the AI build-out is not a broad circular economy but a highly concentrated bet among three cohorts: semiconductor makers, frontier models, and hyperscalers.
The argument
These companies are funding massive compute capital expenditures through debt, private credit, and canceling share repurchases. The guest expressed skepticism about whether this massive compute spend will ultimately deliver a viable return on investment from end customers.
The thesis, stress-tested
✓ What validates it
- ✓Hyperscalers report flat or declining capital expenditure guidance
- ✓Frontier model startups struggle to raise subsequent funding rounds
▸ Risks discussed
- ▸Compute demand remains insatiable, driving sustained high margins for chipmakers
- ▸New AI applications rapidly monetize, proving high ROI for hyperscalers
Hear it yourself
"But it wasn't until, the Zirp, then an easing post the, the, the, repo crisis in, the fall of twenty nineteen and then COVID where government bond prices went parabolic. And so there's this underlying conditions which are bullish. There's the big change, Netscape Navigator, AI, COVID, the LBO frenzy, CDOs, and high leverage in the, great financial crisis that then get used and cause a significant escalation. And so price escalation is one thing. Parabolic moves, typical in a bubble. High valuations, very typical in a bubble. This one doesn't actually have that, and a lot of people are confused about that. You know, if we're in a bubble, why aren't we suing crazy PEs? That's a nuanced conversation we can get into it."
07:15
AFFILIATE LINK · ZORTIX MAY EARN A COMMISSION · NEVER A RECOMMENDATION TO TRADE