causal
Hardware shapes the optimal model architecture bidirectionally
The relationship between chips and models is not one-way; hardware heavily influences which model architecture is optimal, so the best model to run on Google's TPUs given fixed dollars and compute is architecturally different from the optimal one for OpenAI on NVIDIA - down to networking and data-center design choices - meaning architectures diverge based on the underlying hardware constraints.
causalpending
Speaker
Dylan PatelEvidence Quote
“the optimal solution—X amount of compute of TPU vs. GPU compute optimally—will diverge in what the architecture is.”
Source
@Asianometry & Dylan Patel — How the semiconductor industry actually works— Dwarkesh Patel PodcastCreated: 6/13/2026, 3:26:32 AM
Your notes
Loading notes...