Skip to content

Mission

Distributed AI inference so that the megacorps do not get a total monopoly.

Running the best AI models takes datacentre hardware, and that concentrates AI in a few companies. Open-weight models changed half of that: anyone may run them. AIs@home tries the other half: serving them from volunteers’ GPUs to people who don’t have expensive GPUs themselves.

It is an experiment: we don’t know yet whether volunteer GPUs can serve AI well enough, safely enough and cheaply enough. The way to find out is to build it in the open and measure.

One volunteer maintainer. AIs@home is non-commercial: no company, no investors, no token, no ads. The software is open source under Apache-2.0; the name “AIs@home” and the logo are not licensed.