StarZero All shows Try the podcast clipper →

KV cache

2 statements across 1 episodes · 0 bullish · 1 bearish · 1 people on the record · first statement Aug 25, 2026 by Neil Movva · across every show →

Everything said about KV cache, oldest first

Aug 25, 2026 negative
Opinion
Movva: KV cache memory usage is inefficient by 1-2 orders of magnitude
“Specifically, the KV cache is quite uncompressed right now. I think if you look at the entropy in a KV cache, it's nowhere near, it's not earning its keep. Like, we're storing many kilobytes of data in the KV cache per token. And that's probably off by an orde…”
Neil Movva Aug 25, 2026 ▶ 1:04:16 Ex-NVIDIA Engineer: Why AI Is About to Get 1000x Cheaper
Aug 25, 2026
Assertion Supported
Movva: KV cache frequently exceeds model weight size
“You have to store a representation for every token that we sent through the language model. And it frequently Gets to be larger than the weights of the model themselves.”
Neil Movva Aug 25, 2026 ▶ 29:18 Ex-NVIDIA Engineer: Why AI Is About to Get 1000x Cheaper
Made with StarZero

Turn any episode into a week of clips.

This entire site, over 60 episodes transcribed, diarized, checked and made playable, runs on the StarZero media pipeline. Drop in your own episode and the podcast clipper finds the moments worth sharing, cuts them, captions them, and reframes them for every feed.

Try the podcast clipper Learn more