AI bottlenecks shift focus to physical infrastructure
The hosts argued that the best way to play the AI theme is through physical supply-and-demand bottlenecks like energy, utilities, and hardware components rather than the software models.
The argument
They noted that massive CapEx from hyperscalers is hitting severe physical constraints, including electricity grid capacity, transformers, and copper. While TSMC continues to post stellar margins, the physical backlog of deploying these chips will ultimately limit the speed of the build-out.
The thesis, stress-tested
✓ What validates it
- ✓Further margin expansion or positive guidance from TSMC
- ✓Rising power demand and utility rate increases driven by data center clustering
▸ Risks discussed
- ▸Rising electricity costs are becoming a hot political issue, which could lead to regulatory intervention or cost-sharing mandates that hurt margins
Hear it yourself
"You ride the supply and demand bottlenecks. And if you look at certain parts of the energy market, utilities, the the chips, that's where the supply and demand bottlenecks are, and you have to, like, open those up with huge amounts of CapEx."
00:00 / 00:19
AFFILIATE LINK · ZORTIX MAY EARN A COMMISSION · NEVER A RECOMMENDATION TO TRADE