Two things - broadcasting and aggregation.
Windows Live Messenger, AOL Instant Messenger, Y! Messenger and Google Talk and several others are hugely popular Instant Messaging applications. Almost all offer rich client experiences as well as web-based versions. I remember reading somehwere that IM apps are the most used apps on computers, and I am pretty sure they will continue to be for the near future. Demand for IM applications on the web is pretty strong too, with sites such as Meebo growing in popularity.
IM clients have been very innovative over the years, starting from simple 2-person chat to having multi-party conversations, voice support, application sharing, file transfers, etc. Innovation in this space seems to be slowing down though - I haven't seen any new game changing features in any of these IM applications. They all seem to have missed the move to a more social web. In many ways, these apps should have been the ones driving the social web, with their knowledge of social networks via friend-lists. However, websites like Facebook, MySpace and Orkut have firmly established themselves as the drivers of innovation in the social space.
Consumer interactions are evolving from being specifically directed to an individual or small group to being widely broadcast. Blogs, tweets and public profiles are visible to anyone and often are used as a broadcast medium. Increasingly, status messages on social websites and, more importantly, micro-blogging are encroaching on the IM space. Twitter clients such as Twhirl are a great example of this encroachment - users interact with it if it was an IM client but it's really just an app that allows broadcasting messages. YouTube is another hugely successful broadcast medium.
Ofcourse, users also want the ability to have personal, directed conversations along with the ability to broadcast - both to a select group (their network) as well as broadly to the world. For IM application to remain relevant and to fulfill this need, they need to include broadcasting features in addition to preserving the existing communication & collaboration experiences. There are multiple potential approaches to do this - use existing APIs of popular social platforms to allow users to broadcast to a specific network or to farm out to multiple social networks
With the desire to broadcast comes the desire to learn about what others are boadcasting. Most (all?) social websites have news feeds and services can aggregate several such feeds to present a unified view of a users online broadcasts/activity like how FriendFeed does. An IM application would be an ideal home for such an aggregated feed.
In many ways, Facebook is almost there with regards to being an ideal IM application. It has the advantage of being a really popular social platform, which certainly helps. Facebook recently complemented their broadcast communication feature (status messages) with IM-like functionality to enable personal communications. And of course, Facebook pioneered the news feed. Now if only they could go broadcast to and aggregate from competitor sites too... ;)
The arguments above apply even to corporate IM applications - this is not just a consumer need.
Unless traditional IM applications evolve to fill this need and compete with the Facebooks of the world, I see them dying a long, slow death.