AI CapEx numbers are unsustainably stretched
The massive global capital expenditure on AI and semiconductors is fundamentally detached from realistic revenue generation, posing a long-term risk of a capital-cycle bust.
The argument
The guest argued that projected AI CapEx of $6.7 trillion by 2030 requires roughly $2 trillion in annual revenue to justify, which is double the size of the entire global advertising industry. While momentum remains strong, valuations and crowded positioning make the trade highly risky, drawing parallels to the telecom overbuild of the late 1990s.
The thesis, stress-tested
✓ What validates it
- ✓AI industry revenue failing to scale toward the trillion-dollar level
- ✓Hyperscalers scaling back CapEx guidance in quarterly earnings
▸ Risks discussed
- ▸High depreciation and rapid obsolescence of compute hardware
- ▸Crowded investor positioning in semiconductor names
- ▸Extreme valuation levels that assume unrealistic revenue growth
Hear it yourself
"But, again, you strip out those three semiconductor stocks and you get a very, very different picture. Okay. So it sounds like kind of the breadth of the international market is not very good, and and the breadth of The US market isn't so great either. But but but but is the international market perhaps even more reliant slash vulnerable to what happens with semis? So I think one key difference, perhaps, between so many international markets in The US has been the performance of financials. And I spent a lot of time looking at how banks are performing."
AFFILIATE LINK · ZORTIX MAY EARN A COMMISSION · NEVER A RECOMMENDATION TO TRADE