YC-Backed Lamb Labs Claims 63x Higher Efficiency for AI Inference Chips vs GPUs

ycombinator · x · 2026-08-05

Lamb Labs (YC S26) is developing custom AI inference chips claiming to achieve 20,000+ tokens/s with 63x higher Intelligence Per Watt (IPW) than traditional GPUs. The team argues that GPUs became the default for AI simply because they were available, not because they are the right hardware architecture.

Original post →

More from Venture

Venture channel →