In the wake of Google Android, people from three different grid vendors have written about mobile grids: me from Terracotta, Bob Lozano from Appistry, and Nikita Ivanov from GridGain. I think Bob and I both had the reaction of..."cool, but what would you do with it?"
And indeed when taking from a pure cycles High Performance Computing point of view, I don't see that it makes much sense. Even with the rapid improvement of mobile processors, I don't think the cycles you're getting are going to rival servers, even at massive quantities. Not to mention that running a (necessarily) compute-intensive task on your phone would kill the battery time and/or make it unusable for other applications like, you know, calling people.
But then I read Charlie Stross's new novel Halting State over the holiday break. This book is set in near-future (2012) and features as context a series of massively multi-player games that run over mobile phones in a distributed non-centralized grid. State is shared redundantly but not in a centralized manner and secured with encryption.
This item of synchronicity let me think about it a bit differently. There are absolutely applications for non-centralized shared-state grid computing on mobile. They just aren't the kinds of compute-intensive HPC apps that grid vendors focus on. I think there is a lot of interesting territory here and it seems likely to me that location-aware and environment-aware distributed computing apps are going to be ever more prevalent in the future as the devices become ubiquitous and standards emerge.
BTW, I'll have to drop a plug for Stross in general - he has rapidly become one of my favorite science fiction authors. A great place to start with him is his book Accelerando, which just happens to be available free via Creative Commons as an ebook.