Transistor Radio TR 43: TR Live with Jeremie from SemiAnalysis
68 snips
Jan 29, 2026 Jeremie Ontiveros, a SemiAnalysis analyst and data center expert, joins in-person. He talks semiconductors, AI agent tooling like Cloud Code, and trade-offs between major AI labs. Expect lively client anecdotes, token-cost debates, and why agentic workflows could change research and junior analyst work.
AI Snips
Chapters
Transcript
Episode notes
Three Labs, Three Strategies
- Jeremie frames OpenAI, Anthropic, and Google as three archetypes with different strengths and business models.
- Token efficiency and model size drive meaningful cost and margin differences between providers.
Token Efficiency Shapes Margins
- Token efficiency materially influences per-query economics and gross margins across providers.
- Smaller, efficient models can deliver better cost structures despite lower raw capabilities.
Mistaking Druck In A Hedge Fund Lobby
- Jeremie recounts arriving late to a Duquesne meeting and being escorted in by a man who introduced himself as Stan Druckenmiller.
- He admits he didn't recognize Druck's face or name in the moment.