It's not like Node.js is even really dominant today (except in the Hacker News community of course), it hasn't even had a 1.0 release yet. Funny to be speculating on what might replace it already :)
Performance is nice (though micro-benchmarks are not the right way to judge that) but if that was the only thing that mattered we would all be writing custom web servers in C for every app. One of the other important things is the community support, which Node.js definitely has going for it at the moment. It has a very healthy open source (this being the key) community that has created a large library of tools.
I think Vert.x is interesting and has a different way of looking at things than Node.js which is great. If they are able to continue to make progress, get more people interested in it, and show that it is easier and faster to write web apps then who knows? Maybe in a few years we can have a discussion on what is going to threaten Vert.x's future.
There are some interesting things brought up by this micro-benchmark that may be worth addressing in Node, but it's very important to remember that nobody doing 100's of millions of hits per day (which these benchmarks show) does so on one machine. They do it on many machines.
There are certain levels where faster languages make a difference (see for example Twitter migrating their backend from Ruby to Java), but it's VERY rare to get to that position.
Most people need a language that is clean and easy to program in, and a good ecosystem.
Vert.x has no ecosystem to speak of. If you want a JVM based system you can get plenty of options there. If you want a Javascript based system the best option right now is Node.js. It's that simple.