Precision manufacturing at scale. The physics of merging a hundred-gigahertz-scale circuit board track into a waveguide are very unforgiving. The physics governing the tolerances of said waveguide are similar.
For a moment I thought this was actual co-ax, which would be supremely ironic (it was used in the early days of Ethernet, but twisted pair proved cheaper).
But it looks like neither candidate has a conductive core, although they do have the conductive shell.
But one of the companies mentioned has been attempting this for 9 years.
Why hasn't this taken off already? It doesn't seem we're in need of any breakthroughs. So where do the economics break down?