Sharon Zhou: As FLOPs get cheap vs HBM, algorithms will trade compute for memory

realSharonZhou · x · 2026-09-02

Sharon Zhou (founder of Lamini) argues that with the relative cost of FLOPs per token versus HBM per token dropping, a large wave of algorithms, architecture changes, and performance optimizations that trade more FLOPs for less memory is coming — and she's excited to see what gets invented.

Original post →

More from Infra

Infra channel →