You are currently viewing SemiWiki as a guest which gives you limited access to the site. To view blog comments and experience other SemiWiki features you must be a registered member. Registration is fast, simple, and absolutely free so please, join our community today!
There’s definitely going to be inference at the edge, on AI accelerated CPUs and even microcontrollers. But location/host hardware is going to depend on the size of the problem (and associated models) and where the training and reference (RAG and agent support) data is coming from.