They probably were negligible. Billions of links, maybe some content cache that they already had since Google indexes and caches most of the web. Bandwidth was born by the user and content server. Checking RSS feeds is mostly a bunch of 304s.
The newsblur dev is in this thread and he runs a good reader and could probably say what the storage and bandwidth was.
Storage and bandwidth are pretty negligible. I'm under Digital Ocean's quota for bandwidth and NewsBlur is pretty aggressive in keeping feeds up to date. I don't keep an archive of every story ever published, but most feeds don't publish all that much and I could probably invest in storing everything with a cheaper archival DB.
Although the 80/20 rule is in effect, and when I query the DB I find that 80% of feeds have only a single active subscriber and 20% of feeds have 2+ active subscribers (activity in the past 30 days). But that's still quite a bit one-to-many fetching savings. And the majority of fetches are 304s.
The costs of storage and bandwidth of something as popular as Google Reader probably weren't negligible.