In terms of running a data lake or keeping stuff for a long time, it's great, but of course they're banking on you moving a bunch of data to AWS to either train ML off of (compute costs) or keep it there and rack up storage space charges.