However, the opportunity cost of a high bandwidth user is much greater, because their use might be denying 100 other "normal" users access to the network. If 25% of those users decide to cancel because their service sucks, then AT&T has lost significant revenue.
*Note I am not apologizing for AT&T at all, in fact I think they are evil.
It's a tough trade off to make. The way I see it is that you can plan for the increase in data usage and save lots of money on support staff.
The commoditization of wireless access is happening and it seems like AT&T/Verizon are making it easy for someone else to enter the market. Clearwire & Sprint are quickly rolling out a 4G network in major metropolitan areas offering unlimited service for $30/month.
I know integrated services are good, but i'm honestly not sure i want a device which packs several components each with a per unit cost of > $20. I'd much rather look for competitive deals. Same goes for GPS.
Now, if we couldn't only get them to talk better to each other....
I think that plan is still available until next monday, unless I'm mistaken and it ended last monday...
But you can't sign up for the unlimited plan and turn it off whenever you please, which is how Apple is still advertising the iPad despite these changes being announced. [1] At this point nobody can a new iPad and still get this old pricing, given that the iPad is backordered 7-10 days and the changes take effect in 4 days.
AT&T is moving from an all-you-can-eat data structure to one where the people who consume the most network pay more.
Of course this wouldn't be a problem if Steve Jobs hadn't made an implicit promise that you'd be able to get all-you-can-eat data from AT&T for $30 on an iPad, motivating many people to buy one. I almost bought one last weekend based on this assumption, now I'm glad I didn't.
But does anyone know whether a high-bandwidth user really costs AT&T significantly more than one wandering around with their phone on downloading just a trickle of data. AT&T already paid for their infrastructure, right?