There's no fundamental reason that things should play out either way (consolidation in a data center, or personal hardware). It will come down to economics, performance and convenience. Up until now performance (latency) has favored personal hardware, but it appears that may soon change, at which point the economies of scale of data centers, and the convenience of not needing to upgrade/store/maintain/move personal hardware, could very easily sway things in favor of consolidation.
Well, to be fair, there is a lot of spare sillicon in most households. One giant powerserver per family, is much more efficient and should provide more bang for your buck.
Agreed. But the combined sillicon of the xbox, desktop, ps3, mediacenter and multiple laptops in my house, had a seriously high price attached, but using all that sillicon to do one task fast, isnt even an option.
We need an open standard to share spare cpu cycles within the local network. Imagine updating the performance of all your appliances at once, just by pluggin a server in your home network.
But yes, it kind of defeats the purpose of OnLive as a closed product. On the other hand, local ISP can start selling not just bandwidth, but also cpu cycles with low latency. (they could hook up the servers straight in your neighbourhood)
The problem with OnLive is that its not a standard, and its focus is not width enough. They get to choose what kind of computation i want to do.