scripod.com
Gavin Baker - Nvidia v. Google, Scaling Laws, and the Economics of AI - [Invest Like the Best, EP.451]

Overview

Shownote

Highlights

Transcript

Chapters

Pins

Gavin Baker - Nvidia v. Google, Scaling Laws, and the Economics of AI - [Invest Like the Best, EP.451]

Invest Like the Best with Patrick O'Shaughnessy

2025/12/09
Invest Like the Best with Patrick O'Shaughnessy

Invest Like the Best with Patrick O'Shaughnessy

2025/12/09
OverviewShownoteHighlightsTranscriptChaptersPins

Shownote

My guest this week is Gavin Baker. Gavin is the managing partner and CIO of Atreides Management, and he has been on the show many times before.  I will never forget when I first met Gavin in 2017. I find his interest in markets, his curiosity about the wo...

Highlights

In this compelling conversation, Patrick O'Shaughnessy reunites with veteran investor Gavin Baker to explore the rapidly evolving landscape of artificial intelligence, from foundational technologies to long-term strategic implications. With deep expertise in tech investing and a track record of foresight on companies like Nvidia, Baker offers a nuanced perspective on how AI is reshaping industries, infrastructure, and investment opportunities.
06:00
Gemini 3's release confirmed the intactness of pre-training scaling laws
09:05
Post-training scaling laws have enabled breakthrough AI progress since October 2024
12:14
The first models trained on Blackwell are expected in early 2026, likely from XAI.
24:36
Gemini 3 successfully makes a restaurant reservation, showcasing near-term AI assistant potential
33:45
Booking a vacation with family preferences is harder and more economically useful than booking a restaurant.
34:38
VCs are more bullish on AI due to observed productivity gains with fewer employees
42:35
Reasoning enables a flywheel effect in AI, where good answers improve the model iteratively.
46:35
Chinese open-source can serve as a checkpoint, but lack of Blackwell access widens the AI gap with U.S. labs.
53:54
Space data centers using laser-linked satellites enable faster, more efficient AI inference than Earth-based systems
56:49
AI constantly requires compute, making sustained demand different from past tech cycles
1:00:18
Power as a constraint makes the price of compute less relevant.
1:05:57
The semiconductor ecosystem must evolve together to sustain annual cadence gains
1:09:54
AI companies can generate cash earlier despite sub-40% gross margins due to fewer human employees.
1:11:10
SaaS companies have a structural advantage over AI-native startups in adopting AI agent strategies due to existing cash flow and customer data.
1:19:19
Taking risks, changing minds, and admitting wrongness are crucial for learning.
1:25:23
Investing is a game of skill and chance where history and current events create an edge.
1:26:56
Investing is the only thing I'm good at

Chapters

Welcome to Invest Like The Best
00:00
Meet Gavin Baker
04:00
Understanding Gemini 3
06:00
Scaling Laws for Pre-Training
09:05
Google v. Nvidia
12:12
Google as  Lowest Cost Producer of Tokens
16:52
 AI Can Automate Anything that can be Verified
28:05
The AI Bear Case: Edge AI
34:30
Going from Intelligence to Usefulness
37:18
AI Adoption in Fortune 500 Companies
43:44
Frontier Models and Industry Dynamics
48:58
China’s Mistake and Blackwell’s Geopolitical Leverage
56:40
OpenAI’s Code Red
57:50
Data Centers in Space
1:00:46
Cycles in AI
1:07:13
Power as a Bottleneck
1:11:10
AI Native Entrepreneurs
1:14:17
Semiconductor VC
1:16:21
The Mistake the SaaS Industry is Making
1:20:41
Series of Bubbles
1:26:50
Whatever AI Needs, It Gets
1:28:56

Transcript

Patrick O'Shaughnessy: Here's an interesting question to think about. If your finance team suddenly had an extra week every month, what would you have them work on? Most CFOs don't know because their finance teams are grinding it out on lost expense report...