scripod.com
Dylan Patel - The Infinite Demand for Tokens, Claude Mythos, and Supply Constraints - [Invest Like the Best, EP.468]

Overview

Shownote

Highlights

Transcript

Chapters

Pins

Dylan Patel - The Infinite Demand for Tokens, Claude Mythos, and Supply Constraints - [Invest Like the Best, EP.468]

Invest Like the Best with Patrick O'Shaughnessy

2 DAYS AGO
Invest Like the Best with Patrick O'Shaughnessy

Invest Like the Best with Patrick O'Shaughnessy

2 DAYS AGO
OverviewShownoteHighlightsTranscriptChaptersPins

Shownote

This is my second conversation with Dylan Patel. Dylan is the founder and CEO of SemiAnalysis, where he tracks the semiconductor supply chain and AI infrastructure buildout. This conversation is about the supply and demand of tokens. On demand, Dylan desc...

Highlights

In this episode, Patrick O'Shaughnessy revisits Dylan Patel of SemiAnalysis to dissect the rapidly escalating dynamics of AI token supply and demand — a conversation grounded in real-world usage, infrastructure constraints, and strategic implications for businesses and labs alike.
02:29
Frontier models have near-unbounded willingness to pay for tokens
03:14
Current AI spend is over 25% of salary expense, and could exceed 100% by year-end
10:04
Investment firms buy data instead of building it because purchasing is cheaper and faster than internal development
11:44
Access to the most intelligent tokens is crucial for businesses to generate value, and some may be priced out
14:53
The speaker begged an Anthropic co-founder for access to Claude Mythos
15:36
Claude Mythos is potentially the biggest step up in model capabilities in two years
20:54
Ken Griffin got early access to advanced AI models, giving him a significant market advantage
23:45
The concept of a software-only singularity is a blip, as the majority of the world is physical
26:03
Claude Mythos is a larger model proving scaling laws still work
29:42
Failing to use more tokens to generate and capture economic value risks falling into the permanent underclass as model capabilities rise.
31:33
Memory margins have skyrocketed, and companies are making large prepayments.
35:57
TSMC may spend $100 billion on CapEx in 2028, risking downstream supply chain shortages
36:45
CPUs are crucial for reinforcement learning environments, where they grade the model's attempts, and for deploying models, as the output from GPUs or ASICs goes through apps running on CPUs
40:17
Actual token usage often exceeds estimates
42:20
Predict large-scale protests against AI in three months due to growing public fear and backlash

Chapters

Welcome to Invest Like The Best
00:00
Intro: Dylan Patel
02:29
Semi Analysis AI Spend: Zero to $7M
03:09
Real-World Examples of Claude Code
05:16
Token Demand: “Completely Explosive”
11:41
Why Everyone Wants the Frontier Model
14:48
Mythos: Biggest Model Capability Jump in Two Years
15:36
Fear of Rapid Model Progress
20:54
Robotics as the Next Demand Wave
23:45
Scaling Laws & Compute Efficiency
26:03
OpenAI vs. Anthropic
27:24
Supply Side: Bottlenecks Across the Stack
31:33
TSMC CapEx Could Cause a Shortage
33:26
CPUs, ASICs, and FPGAs
36:45
Tokenomics
40:12
Protests & AI Backlash
42:20

Transcript

Speaker 1: Most software companies try to maximize your time on their app to juice engagement. Ramp does the exact opposite. Ramp understands that no one wants to spend hours chasing receipts, reviewing expense reports, and checking for policy violations. ...