Why are we optimizing for "true hybrid cloud"?
I find VMware's response amusing because it shows how little VMware understands about actual, good cloud architecture. (Of note, Mathew Lodge, who works for VMware in this space, is one of the smartest people I've ever interviewed on the cloud and other modern technologies--they should definitely put him in charge of more stuff there, and perhaps they wouldn't be so far behind).
The idea that "true hybrid cloud" (that is, a cloud where you are splitting applications across public and private clouds) is an enviable "nirvana" is insane. True hybrid is very, very hard to do properly, and, because of this, should generally be avoided if at all possible, Instead, having separate public and private clouds (have functions entirely contained within one or the other, like dev/test in public cloud and production in private), or having cloudbursting scenarios where you go from private to public only in exceptional circumstances, are far preferable to trying to maintain a constant, solid connection between public and private clouds today.
Amazon, in contrast, is consistent and adamant that you're going to be better off if you go entirely public, using Amazon's regions as your failover. The problem with Amazon's approach is that it leads to AWS vendor lock-in, but to date, that hasn't really been a huge issue, since they've been the most reliable (by machine hours, certainly) and essentially the cheapest (as far as price per performance has gone, and please don't quote me some benchmarketing that compares AWS's 2008-era m1 family to what some other provider released 6 months ago). And what you do get with AWS is very functional. As opposed to where you'll be trying to implement a true hybrid cloud today.