I don't agree, for several reasons.
First of all, Apple's ARKit (and by extension, Google's ARCore) is not a good experience. It has some "gee whiz" factor to it, and then once you get over it, you realize that there really aren't very many things you can do with it that you can't do better with a smartphone or a VR headset.
Case in point: I spent a lot of time at my previous employer working on self-service/guided-repair AR apps for the iPad Pro. It's a ridiculously bad concept (though my previous employer is kind of known for chasing ridiculously bad concepts). 1) Your hands are occupied with the tablet, so you're not actually able to perform the repair tasks without putting the tablet down, thus largely negating the usefulness of the overlaid 3D animations we created, and 2) who is going to take a $1000 Apple product with them out to a repair job?
We also spent a lot of time making large posters into 2D animations for conference expos. You could have had the same, exact experience for 1% of the cost if you just bought gigantic TVs and turned them to portrait view. I mean, I didn't complain about getting the money, per se, but it just was not good stewardship for our clients to even propose that contract from the sales team.
There are distance measuring apps that presume the tracking is accurate and precise, which it is not. For a very long time, ARKit had very little concept of how far away any given surface was. Instead, it was good about scaling objects so that visually they did not appear to be moving closer to you as the tracking point jittered wildly on the Z-Axis. At best, it's better than walking off steps to get rough measurements of a space, but anything more than that and you're going to have an easier, faster, more reliable time with a $10 retractable tape measure.
Second, the HoloLens is an amazing device, but it's nowhere near ready for any mainstream use. I think it does have a good use case in industrial repair scenarios (which was originally how we got the iPad work, I had built an app for HoloLens that did the same thing, but then the suits got involved and wanted to push our Apple partnership). There are specific features of the HoloLens that make this work, and there are specific features of the job that make the HoloLens work. Specifically, HoloLens' built-in, completely free voice recognition is a huge boon for being able to build systems that you can use while your hands are occupied with the job. Second, having physical objects (the machine you're trying to repair) on which to hang virtual graphics causes a huge improvement in believability of the visual and usability of the application (primarily related to issues of object permanence being hindered by the limited field of view).
That last point wouldn't be so bad if the HoloLens were visually and performance-wise up to spec with much, much cheaper VR headsets in terms of grpahics and visual fidelity. But so long as you're paying literally 10x as much for half to a quarter as good of visuals (not to mention the input systems being otherwise mismatched), you have a lot of ground to catch up on to make an AR-only experience as valuable and meaningful to a user as you could a VR experience.
AR will continue to be a lackluster consumer user experience for a long time. Most developers making apps for these devices don't understand how to make a real AR app. In AR, the user supplies the context to the developer. In VR, the developer supplies the context to the user. This fundamental dichotomy makes good, meaningful AR application design significantly harder than VR design. In my own evaluation, easily 80% of the extant software in the HoloLens and Magic Leap stores take no cues from the user's immediate vicinity, in their context, to which they are tying their experience. In essence, they are just crappy VR experiences.
In comparison: VR is cheap, it works, it is relatively easy to design for, and there are already a ton of great apps available for it.