HTML5 is the new <applet> which no one will bother install.
There is a common fallacy among developers, we all assume what we like must be also liked by end-users and consumers.
As easy as it is to be pessimistic about the browser technology of the average web user, I'm not so sure that HTML5/CSS3 is such a lost cause. Consider that already something like 1/4 to 1/3 of all web users in the US are using a browser that at least partially supports HTML5 or CSS3. Also consider that the more technologically sophisticated the audience of a particular site the more likely they are to use more modern browsers. Given these two simple facts and existing trends in browser marketshare (away from IE, toward FF, Chrome, Safari) I think there is a very real potential for some future major web venture (either de novo or as an evolution of an existing site) to have a strong case for relying on HTML5/CSS3 technologies.
Indeed, I wouldn't be at all surprised to see just such a thing happen multiple times within the next year.
As well there are many ways, as with html5 video, to take advantage of html5 features without abandoning support for "legacy browser" users. Which is the right way to do it just like any major refactoring/redesign effort.
Look, take any major modern web site and open it up in a browser from 1997 and it just plain won't work, because modern sites are built using technology that wasn't supported back then. How did we get here from there? It was a long, slow, hard slog that was painful nearly every step of the way. Nevertheless, here we are, in all our modern CSS2(ish), AJAX, HTML4(ish) glory. Where will we be in 5 or 10 years? Certainly in a different world founded on different technologies and standards. How will we get there? It'll be a long, slow, hard slog just like it always is. But we'll still get there.