Compute layer remains the primary AI bottleneck
The guest argued that the best way to play the AI and robotics transition is through the scarce compute infrastructure layer rather than picking application winners.
The argument
He noted that computer chips are the 'new oil' to AI. As the technology scales, the physical bottlenecks are increasingly memory-bound, making high-bandwidth memory (HBM), optics, and networking the critical points of value capture.
The thesis, stress-tested
✓ What validates it
- ✓Continued upward revisions in semiconductor backlog and earnings guidance
- ✓Increased adoption of high-bandwidth memory (HBM) in next-generation data centers
▸ Risks discussed
- ▸High cyclicality of semiconductor capital expenditure
- ▸Market skepticism regarding the sustainability of current growth rates
Hear it yourself
"And, you know, these companies, like, in terms of, like, Catherine, they would like, you know, be scarce in terms of how many companies can do them, like, the scale that's needed, like, the I actually think, like, you know, computer chips are, like, the new oil to to AI."
00:00 / 00:17
AFFILIATE LINK · ZORTIX MAY EARN A COMMISSION · NEVER A RECOMMENDATION TO TRADE