Computing on demand. You can rent one or more virtual machines, each equivalent to a 1.7GHz Xenon with 1.7GB of memory, by the hour for $0.10/hour. That’s about $70/month (plus $0.15/month per GB of storage), which isn’t bad on its own. It’s even better because you can ramp capacity up and down by bringing additional instances on line.

If it works as promised, I think lots and lots of smart, hungry, low capital-outlay startups are going to build using Amazon’s infrastructure. It looks like a no brainer. If you architect your application intelligently, the cost of scaling is going to be rougly linear with load and the lock-in is minimal. From what I understand, the computational environment looks like a linux box, complete with root level access, so the cost of moving to dedicated hardware at some point is very doable. Of course, if Amazon does things right, people won’t ever want to leave, but as a potential customer, it’s nice to know that there is an obvious route to independence if things don’t work out.

Would be interesting to experiment with this for scientific computing. Ideally you’d want an instance manager that could be tuned to the fact that instances are billed on an hourly basis.

Amazon.com Amazon Web Services Store: Amazon EC2 / Amazon Web Services
Amazon Elastic Compute Cloud (Amazon EC2) is a web service that provides resizable compute capacity in the cloud. It is designed to make web-scale computing easier for developers.