I half-agree, Gene. There is the obvious push by the big companies, because groups like Microsoft have realized that they've pretty much saturated the market and run out of killer features to add. The only way to get ongoing revenue is, in fact, a subscription model.
However, I see a few problems that I don't think they'll be able to iron out.
First, there's the Open Source groups. OpenOffice is annoying to use, but if it's a choice between it and the next iteration of LiveDocuments (or whatever it was called), guess which one I'll choose. And if they kill Windows, well, ReactOS is finally shaping up into something that's nearly releasable. And I can confirm that OpenOffice actually does run well enough for general use on cheapo netbooks.
Second, there are all the organizations that rely on secrecy. The Pentagon, for example, is not going to allow all its sensitive data to be intercepted during upload or copied from a server farm in another country. Hedge funds aren't going to release their plans somewhere where a guessed password loses them their advantage.
And don't forget to add in the user licenses. To serve your data to you wherever you are, they (like Facebook's Terms of Service, for example) need you to give them the right to copy and otherwise use your content however they see fit. Without that, you could technically sue them for releasing your data to any machine other than the one you uploaded to. But that also means that they own the rights to use and publish your documents without your input. Sure, they usually won't, but if you're an admiral, a day trader, or a criminal, are you going to trust that they won't make an ethical exception? Probably not.
Third, there's latency and reliability. Imagine the day traders at Goldman-Sachs using 3G Wireless for their trades instead of the dedicated NYSE line.
Along similar lines, there are also applications that simply don't make sense anywhere but your computer. Home automation, for example, can only be programmed from your machine. Dealing with that as a "cloud application" is complication for its own sake, and not a likely investment.
And last, there's competition. We don't live in IBM's world anymore, or even Intel's. Hardware is cheap to produce in just about any configuration. If desktops disappear entirely, I wouldn't be surprised to see some Chinese company producing, say, docking stations that'll connect your iPhone to your television, a keyboard, and an external hard drive to turn it into a "real" computer.
So yeah, if Microsoft, Oracle, and Google have their way, the desktop will vanish and we'll be forced to pay them monthly fees for the use of their software. But the resistance is alive and well in the same market pressures that kept picoJava microprocessors from going in your refrigerator fifteen years ago (and kept Windows 95 from being a "cloud operating system"--remember all that "the web is the desktop" propaganda?), so I'm not overly concerned.
(And notice that Microsoft isn't even taking it that seriously. Office 2007's ribbon is a desktop idea, and not web-friendly, for example, and their new file formats are designed to be usable outside of Office itself. That wouldn't make much sense, if their direction is to provide mainframe-based services where nobody would ever see their data.)