AI infrastructure overinvestment threatens capital returns
The guest argued that massive capital expenditure in AI infrastructure is mirroring historical tech bubbles where overinvestment ultimately destroys the return on investment.
The argument
The guest pointed out that while transformational technologies like railroads and fiber changed the world, they yielded poor ROI due to overcapacity. He highlighted complex circular financing in the current AI cycle, such as OpenAI's warrant-heavy deal with AMD and Broadcom's backing of Anthropic's TPU leases, as signs of a frothy market.
The thesis, stress-tested
✓ What validates it
- ✓Hyperscaler debt ratios continue to rise rapidly over the next few quarters
- ✓Defaults or restructurings occur in AI-related leasing agreements
▸ Risks discussed
- ▸If electricity constraints halt data center buildouts, it may save capitalists from overinvesting
- ▸Hyperscalers currently remain high-quality companies with strong balance sheets
Hear it yourself
"tricky thing about today is we worry that this may be an earnings bubble. 2000 was a little bit of that, but mostly evaluation bubble. We saw earnings bubbles in Europe in, you know, in 2007, 2008, where, like, the earnings had just gone up a 100% over four years. And even today, they are struggling to make it back up on an index basis to those levels. We are likely to be in a situation where over the next twelve months, we see more supply come into The US stock market than has been the case in living memory."
AFFILIATE LINK · ZORTIX MAY EARN A COMMISSION · NEVER A RECOMMENDATION TO TRADE