i agree with the 'buy thin' mantra. As you mention, it's exciting to see something like apex come to the market in the future. i've also been keeping my eye on some other upcoming platforms like those offered by Apprenda (http://www.apprenda.com) and Etelos (http://www.etelos.com). who knows, apex may only be the start of things to come in making thin clients the new 'in'

Buy thin? Not if you're talking investments in stocks. Consider:
A thin client has to run at the speed of your internet connection. That's a five year old Pentium III with a (still working) 20 GB hard drive and 128 MB RAM. Faster is a waste. Windows is a waste too, so you're going to be looking at Linux as more "congenial" installation packages come along, i.e., Ubuntu's latest.
But maybe not EVERYTHING is going to be done server-side. There'll be a few client-side processes, so maybe you'll need a bit more than the thinnest of thin clients.
And there'll always be some data that you don't want to store on the net. You know, that stuff your ex-wife's lawyer will want to subpoena. A good sledge hammer on a hard drive is far better at evading a subpoena that fighting her lawyer and your SaaS provider in court. So maybe you'll need more that 20 GB locally.
And so on.

I agree with some of your sentiment. However at my company we are changing from a thin accounting/web store (Netsuite) to a traditional fat, locally hosted app running on a network. Why? The thin approach has been getting a fatter and fatter price over the term of our three year lease. It's now over 100% more than three years ago... and it wasn't cheap then. Waiting 5 seconds for a response to every key press has it's downside too, and this is with 5 MBit DSL. There's more to the story of course. But you'd think a software service could be priced so attractively that it would be a no-brainer vs hosting your own application. Economy of scale and all that. But in at least some thin client implementations today this is not the case.