scripod.com

Dylan Patel – Anthropic & OpenAI will have most of the world’s compute by 2028

Dwarkesh Podcast

Shownote

Had a lot of fun chatting again with my twin brother Dylan Patel. We went through lab economics over the next few years - the shift from inference to training as RSI draws near; and how Anthropic and OpenAI are on track to control most of the world’s usab...

Highlights

This conversation examines how the economics of advanced AI could reshape the technology industry and the global economy.
05:52
Compute efficiency may matter more than power ownership
12:09
AI infrastructure spending may exceed lab returns for years
15:39
Automated AI workers could generate immense value per gigawatt
20:48
Compute scarcity shifts power to infrastructure owners
27:39
Regulation may hide progress, not stop it.
29:47
Inference increasingly fuels the race toward AGI
46:13
AI expansion may reshape the global economy
1:00:00
AI growth could push interest rates into the tens of percent
1:10:53
Scaling advantages could centralize the future of AI

Chapters

Two labs will soon control most of the world’s compute
00:00
$6 billion in fab capex enables $1t+ of end revenue
07:01
Compute prices will rise if the labs outbid everyone
13:08
Which layer will capture most of the surplus?
18:22
What could slow down progress?
25:40
Labs are shifting compute from inference to R&D
29:43
China gets less than 10% of new compute, but its labs need less
33:27
Will AI cause a sovereign debt crisis?
48:48
Will the world’s future workforce belong to a few companies?
1:07:52

Transcript

Dwarkesh Patel: Okay, I'm back with Dylan Patel, founder of Semi-Analysis. Our version of a family Thanksgiving dinner is a regular, yearly podcast. We are not actually related. Dylan Patel: Don't tell the people this. Dwarkesh Patel: We'll destroy the m...