For the foreseeable future, however, a vision is all it is.
Grids are nothing new, and many universities and research establishments have been running networks of computers that share computing power for years. For instance there is the Large Hadron Grid Collider here in the UK. Very basic forms of utility computing, where companies can rent processing power from supercomputers, such as IBM's Deep Computing Capacity on Demand facility in Montpellier are also now becoming more established. Sun's subscription pricing model will provide more resources for companies wishing to buy computing power on demand.
But all real-world examples of utility and grid computing for large organisations entail tightly controlled computers in tightly controlled environments. Distributed computing projects such as SETI@Home are the exception, but the crucial difference here is that nobody's processing power is charged for, and nobody is asking large companies to spare processing cycles on their mission-critical servers.
For organisations to feel comfortable feeding processing power back to a cloud of computers on the network, two obstacles have to be overcome: The first, as Sun's Schwartz rightly points out, is security; allowing an unknown application from an unknown party to run cycles on your server processors is no trivial matter. Second, nobody -- not even Sun -- has come close to articulating a pricing model for this.
This second hurdle is at least as high as the first; just look at the consternation caused by dual-core processor among those who have to work with, and work out, software licensing models. How do you charge for software running on a combination of single-, dual- and more-core processors on machines from here to Timbuktu? Of course, free open source software neatly bypasses this problem: Sun may not find this entirely to its taste.
Right now, utility computing works for single organisations. In the near future, we're likely to see large companies rent out processing power to their trusted business partners, but a world where we all share each other's computing power with confidence is still a long way off.