# The shift from H100s to 4090s: "now we can incentivize home computers... eventually mobile phones"

- **Speaker:** Greg Osuri
- **Attribution:** attributed
- **Said:** 2024-08-02
- **Theme:** Local Compute — https://www.gregosuri.com/predictions/local-compute/
- **Source:** Greg Osuri, Marko Stokic, Michael Heinrich & Luki Song on Can User-Owned AI Compete with Big Tech? (Nebular)
- **Watch at:** 00:10:09 — https://www.youtube.com/watch?v=9bEi1i2c12E&t=609s
- **Canonical:** https://www.gregosuri.com/predictions/local-compute/2024-08-02-the-shift-from-h100s-to-4090s-now-we-can-incentivi-lt2a/
- **Cite as:** Greg Osuri, "The shift from H100s to 4090s: "now we can incentivize home computers... eventually mobile phones"," Greg Osuri, Marko Stokic, Michael Heinrich & Luki Song on Can User-Owned AI Compete with Big Tech? (Nebular), 2 Aug 2024, 00:10:09. https://www.youtube.com/watch?v=9bEi1i2c12E&t=609s

## Quote — verbatim

> For the first time we've seen a shift in what kind of compute we'll need — H100s were all the glory a few months ago, but now it's 4090s. 4090s, turns out, are actually pretty good for inference workloads, and you see gaming machines having quite a lot of 4090s. So the shift from data-center-grade compute to a consumer-grade computer — that's happening, and that's incredible news because now we can incentivize home computers... So there's a big shift happening from data centers to home-grade computers, eventually to mobile phones.

## Context — site annotation, not spoken

On incentive design; he also cites the Nvidia-to-AMD shift (Tinygrad's first box on Akash with 120GB-memory AMD chips "as a home setup") and Google DeepMind's DiLoCo paper implemented by Prime Intellect enabling heterogeneous distributed training. Quote spans to [00:11:36].