PodcastDeep ValuesOctober 25, 2026
The AI Hardware Value Trap: Why 3x Earnings Can Lie | Stock Market
Cheap memory-chip multiples and NVIDIA/Broadcom's $40B gigawatt margin gap both face structural compression risk, not durable value.
- Memory margin mirage: HBM makers post 80-90% gross margins (up from negative just years ago) and trade at 2-4x forward earnings, but new fab capacity arriving ~2028 could collapse margins, making the 'cheap' multiple a cyclical value trap.
- Gigawatt cost math: Building a gigawatt of AI compute cost $40B in 2024, now $60B in 2026, but only ~$20B is raw silicon/packaging/memory cost — the remaining $40B is pure margin to designers like NVIDIA and Broadcom.
The full card is a free trial away.
The full summary and the deep dive on this episode, plus a brief every morning on the shows and companies you follow.
Read free — start your trial14-day free trial
Themes, sentiment and reasons are AI-written paraphrases, checked so they never quote an episode. They may contain errors. Not investment advice. Episodes remain the property of their creators.
