Based on the talk at Devoxx, Lars Bak indicated it was a goal for this year. But nothing more concrete. There are still a number of hurdles including completely overhauling the entire GC process in current version of Chrome (Oilpan project). This was only started on at the end of last year and only with a compile time flag. It still has a way to go yet.
Dart is about so much more than just the VM. Even if the language only makes you 5% more efficient (though for me at least, I know the numbers are much higher than that), than writing the equivalent JavaScript code, then dart is worth it.
Truthfully, I'm not sure why you would. Then compilation step really isn't that long, and once transpiled to JavaScript then you can target not just Chrome but most of the current browsers available in some cases at standard JavaScript speeds. Why limit your audience when you don't have to?
It's true the Dart VM in Chrome will give some nice advantages (particularly with snapshots allowing for an increased application startup time). And the running performance may be better than the transpiled version. But that doesn't mean once the Dart VM is released that you should stop building your projects. As I see it, once the VM does start to land in Chrome (and maybe other browsers some day too), then your build process should do two things. 1) Generate working Javascript version of your application. and 2) Generate a minified version of your Dart application in Dart so you save on filesize as well.
In any case, as mentioned in the Dart Devoxx videos, the goal is for this year. But as with all aspects of Dart development (and probably any development within Google by my suspicion) Timelines are not set in stone but fluid and changing due to varying needs. So you'll probably not get a more definite time than that.
=) Matt