Here's Why Here's Why AI Costs Still Worry Investors
Nov 21, 2025
Tom Mackenzie, a technology journalist and host of Bloomberg Tech Europe, dives into the concerns surrounding AI investments. He explains why investors like Michael Burry are anxious about rising data center costs and depreciation risks of AI chips. Tom highlights NVIDIA’s defense regarding chip longevity and discusses the alarming revenue gap against infrastructure spending. He also examines the funding risks for smaller cloud providers and warns about the complexities of circular financing in the tech industry.
AI Snips
Chapters
Transcript
Episode notes
Depreciation Risk Could Create Stranded Assets
- Michael Burry worries AI chips depreciate fast and could leave hyperscalers with stranded assets.
- This risk echoes telecoms' huge infrastructure losses during the dot-com bust.
Older Chips Still Serve Useful Roles
- NVIDIA argues older chips like Hopper remain useful for inference and have a roughly six-year life.
- That versatility helps counter the fastest depreciation arguments from skeptics.
Revenue Gap Must Close By 2030
- Bain says hyperscalers must generate roughly $2 trillion by 2030 to justify current infrastructure spending.
- The key uncertainty is product-market fit and whether enterprises adopt AI at scale before investor patience fades.