Bits + Bips Why Elon’s $25B Chip Fab TeraFab Is 'Not Real' and AI Layoffs Are an Excuse
Mar 25, 2026
They debate the realism of a $25B chip fab plan and whether space-based data centers make sense. They examine tech and crypto layoffs framed as AI-driven, questioning if AI is a convenient cover. They highlight big tech AI reorganizations and the downstream innovation that AI spending might spark. The conversation mixes skepticism of grand plans with curiosity about where AI funds will actually flow.
AI Snips
Chapters
Transcript
Episode notes
TeraFab Announcement And Ambitious Scope
- Elon announced TeraFab, a $25B joint fab between Tesla, SpaceX, and xAI targeting 2nm and space-directed compute allocation.
- Austin, Chris, and Ram flagged scale issues: no timeline, 80% orbital compute claim, and lofty ambitions like on-shoring chips.
Fab Economics Make TeraFab Implausible
- Ram calls TeraFab 'bullshit' because fabs cost $20–25B to start and require deep technical execution that only incumbents like TSMC have mastered.
- He cites TSM's $100B investment, Tesla's limited free cash flow (~$3–6B), and lack of CapEx planning as practical barriers.
Space Data Centers Are Not Economically Mature
- Ram and others argue data centers in space don't make sense today; Jensen Huang and Semi Analysis sources back this skepticism.
- The technology and economics for orbital AI compute remain immature, making Elon’s space-compute emphasis questionable.