Our architecture is generally built in a way with self-hosted in mind but it's somewhere down the road and not something we'll have in the short to mid-term.
For my own part I struggle to find data that's strewn across frequently deprecated PDF's, salesforce, one-drive / sharepoint, mailing list archives, a home-spun knowledge base, yammer, an aging twiki instance, and possibly a few more information repositories. And I work in a relatively small and tech-savvy organisation. So I definitely feel the pain you're trying to cure. :)
'Cloud' is tricky. No two people agree what that word means for starters. In terms of a knowledge repository / index / retrieval system -- for many enterprises, when talking about core institutional knowledge, 'manage yourself' is not the opposite of 'in the cloud'.
My use of 'cloud' wasn't very exact, you're right. I suppose most enterprises have or will have their 'own' infrastructure in the cloud and manage it there themselves. I really meant cloud in the way of a SaaS-model.