AI CapEx faces a return-on-capital reckoning
The massive capital expenditure boom in artificial intelligence is reaching a point where the required returns on deployed capital are unlikely to be met, threatening hyperscaler valuations.
The argument
The guest argued that a projected $5 trillion in total AI spend would require roughly $550 billion in earnings just to achieve a basic 10% return on capital, a threshold the market is far from proving. A key signal of this peak cycle is Google shifting from buying back shares to issuing shares, indicating a capital-intensive phase that may dilute shareholder returns.
The thesis, stress-tested
✓ What validates it
- ✓Hyperscalers reporting lower-than-expected ROIC on AI investments
- ✓Deceleration in AI CapEx guidance in upcoming quarterly reports
- ✓Further share issuances by major tech firms to fund CapEx
▸ Risks discussed
- ▸AI productivity gains create entirely new, highly profitable revenue streams
- ▸Customer captivity allows hyperscalers to extract high surplus margins
Hear it yourself
"I would just point out just how stressed the system was and how close we were to I don't think the word is overused collapse. And then, sorry, Tanner, the second part of your question. The second part of my question is exactly about that psychological makeup that one needs in order to actually invest during the times of crisis. I mean, that famous line from the 1986 letter that we are greedy when others are fearful and we are fearful when others are greedy type of thing, you know, which is exactly the right type of mental framework to have when it comes to investing in crisis."
AFFILIATE LINK · ZORTIX MAY EARN A COMMISSION · NEVER A RECOMMENDATION TO TRADE