Yahoo's research divisions do great work that basically never makes it out into the wider company past the prototype stage. It is an awesome, wonderful waste of money.
Two layoffs ago they cut 'brick house' which was a hands-on innovation cell, purveyors of live.yahoo.com (a 2007 version of chatroullete with many more features and 5-way video chat) and fire eagle, the GEO standard, and other nice stuff.
man, it's depressing to count a company history in 'layoffs'
(1) hadoop enhancements (a friend works there on exactly that), which are a nontrivial cost savings given the size of Y's deployed clusters
(2) ad matching tech, which should be considered of the highest priority, given that Yahoo's business model is to make money off advertising. Small improvements in matching <users, stories, ads, context> can easily lead to material improvements in earnings
Why Y! doesn't consider the two above things important is beyond me. Also, one of their recruiters sent me an email this week! Dude has big old brass ones, that's for sure.