# The peer-to-peer machine learning cloud

- **Speaker:** Greg Osuri
- **Attribution:** attributed
- **Said:** 2021-03-24
- **Theme:** GPU Economics — https://www.gregosuri.com/predictions/gpu-economics/
- **Source:** AKASH NETWORK (AKT) - Decentralized clouds, Staking, Cosmos & more! - Interview with CEO Greg Osuri (Michaël van de Poppe)
- **Watch at:** 00:46:37 — https://www.youtube.com/watch?v=XY33CFNI-OE&t=2797s
- **Canonical:** https://www.gregosuri.com/predictions/gpu-economics/2021-03-24-the-peer-to-peer-machine-learning-cloud-zau3/
- **Cite as:** Greg Osuri, "The peer-to-peer machine learning cloud," AKASH NETWORK (AKT) - Decentralized clouds, Staking, Cosmos & more! - Interview with CEO Greg Osuri (Michaël van de Poppe), 24 Mar 2021, 00:46:37. https://www.youtube.com/watch?v=XY33CFNI-OE&t=2797s

## Quote — verbatim

> 20% of revenues for these machine learning companies is going to the cloud... We'd love for these companies to install Akash and earn the tokens when they're not using \[their servers\], and when they want to peak they can use these tokens to actually get compute. This is the use case where the provider becomes a tenant and vice versa, creating a truly peer-to-peer cloud.

## Context — site annotation, not spoken

The 2021 GPU roadmap item (a16z report on ML cloud spend cited): ML companies' office GPU servers idle ~21 hours a day; enabling this required solving GPU multi-tenancy. GPUs eventually shipped on Akash in 2023.