Invest Like the Best with Patrick O'Shaughnessy

Gavin Baker - Watts and Wafers - [Invest Like the Best, EP.473]

1h 18min

Gavin Baker and Patrick O'Shaughnessy explore the unprecedented AI infrastructure buildout, focusing on how Anthropic's rapid revenue growth (adding $11B ARR in a month) signals a historic moment in capitalism. Baker argues that March 2024's market drawdown was a buying opportunity, driven by AI fundamentals, GPU shortages, and the potential for orbital data centers to solve energy constraints. He warns of a potential bubble if TSMC expands wafer capacity too aggressively, though he notes that current valuations for high-quality AI companies remain reasonable.

Summarized by Podsumo

🎧 Listen 🎙️ Ask about this episode

✨ Key Takeaways

1

Anthropic added $11B ARR in one month, exceeding the combined 10-year ARR of Palantir, Snowflake, and Databricks, marking an unprecedented moment in capitalism.
2

Gavin Baker argues that March 2024's market selloff was a 'pent-up alpha' opportunity, as AI fundamentals (e.g., GPU rental prices rising) contradicted negative price action.
3

The disaggregation of pre-fill and decode in AI inference could extend GPU lifespans to 10-15 years, benefiting private credit markets by lowering financing costs.
4

Baker predicts orbital compute (racks in space, not giant data centers) will alleviate Earth's energy and zoning constraints, with SpaceX leading the effort.
5

The ‘bitter lesson’ (more compute beats human ingenuity) may temporarily break with ASI, but Baker remains bullish on frontier models capturing most economic value.
6

TSMC's capacity decisions are the single most important variable to watch; if they expand too fast, it could trigger a bubble; if constrained, it sustains the bull case.
7

New chip companies must make 'hard and different' trade-offs (e.g., Cerebras with wafer-scale computing) rather than trying to build a better GPU.
8

Usage-based pricing (pay-per-token) is replacing all-you-can-eat plans, potentially driving OpenAI/Anthropic ARR beyond $200B, but creating a 'lobotomized' experience for budget users.
9

Continual learning (models updating weights in real-time) could enable a fast takeoff to ASI, but remains unsolved; current RL fine-tuning is a crude approximation.
10

Personal and geopolitical risks rise with AI; Baker worries about political violence targeting AI leaders and advises 'safari safaris' to avoid cybercrime via AI deepfakes.

💬 Notable Quotes

"Anthropic added $11B of ARR in one month. Nothing like that has ever happened in the history of capitalism. Forget my career. Just the flat out history of capitalism."
"The disaggregation of inference means that I think these GPUs are going to have 10 or 15 year lives. [...] you can extend the useful life of that GPU until it melts."
"The returns to the frontier. All the economic returns to AI at the model layer [...] have been at the frontier, which is surprising to me. And I think it's been surprising to a lot of people."