I go to any $5/month host, check "install wordpress", and it's done. When Clojure CMS/blog apps have even 1% of the installability (as in, ability to be installed) by non-developers, perhaps they'll have a fighting chance.
I say this as someone who develops Grails apps all day long.
That is a function of popularity, not language. If some blogging app had a userbase the size of wordpress's, it wouldn't matter if it were written in PHP or brainfuck, shared hosting providers will set it up. This is why appealing to techies makes sense. They will be the ones setting it up in the initial "its not popular yet" phase, and if they don't use it, it will never get off the ground. Once they start setting it up for their friends/family/etc, a few companies start offering it as an option for their hosting. Then a few more, and it snowballs as the app gets more and more popular. Trying to start from "use PHP so end users will use it" doesn't work, because installing a PHP app is just as hard for them as installing anything else.
Keeping JVM engines going, and keeping a lot of compiled apps in RAM, on the offchance that someone makes a request, means fewer apps/sites can coexist on the same hardware. My $5/month PHP plan (hypothetical - I don't have one) - only compiles/executes the PHP when the request is made. If no one requests my site for 3 days, it's all just sitting there, not using any resources.
Until there's another technology that follows a similar model (or something else which provides good economies of scale for hosting providers), PHP will continue to dominate large segments of certain problem spaces.