I think the pull is that Google is a very academic place (based on visits and talking to people who work there), and if Google's interests align with yours, you're basically doing the research you would have already done PLUS you get paid better PLUS you don't have to fight for funding on an almost daily basis PLUS you actually have data. Google is a fabulous, exciting, wonderland place for lots of software engineering professors/students.
The question is whether Google is a career. Places like Google, Microsoft Research (and, perhaps Facebook) allow for the freedom that academics crave, but that's only three places. If research-style engineering goes out of fashion in favor of enterprise middle-management hell, you're going to be left high and dry if you don't have a publishing record to fall back on when you try going back to the Ivory Tower.
Fortunately, those places have enough cachet that you can say "I worked at Google" and no publishing record for those years are forgiven, but if you don't have one (say, you went half way through your PhD), coming back is going to be hard. That's why the decision is so difficult: you're taking a bet on whether the benefits taken for granted in academia are going to continue in industry. A broader shift would have to show more companies than those three, in order to provide (for want of a better word) "safety" that your worklife goals aren't in jeopardy.
That shift happened back in the .com boom of the late 90s. It has never really reverted back. It's still the case that top-flight PhDs are going to work as developers at companies (versus academic or industry researchers).
I think it is a testament to how interesting the work in our field is.