The upside for a ROI seems like it would be enormous, and pretty easy to make happen. Industrial seat licensing for products like Solidworks, CATIA, Inventor, and (now) Fusion are enormously expensive. It's not just private organizations either--CAD proficiency is such a basic skill that every serious engineering school has an organizational license for their students too, which I'd imagine also costs a bundle. It's not as if the tools themselves are expanding functionality at some sort of rapid rate, either; I haven't done much CAD in the last two years (so maybe I somehow missed some sort of feature explosion), but while a regular user between 2013-2018, I saw basically no change in the vast majority of my most-used tools for several different CAD programs, with the exception of some improvement in out-of-the-box simulation capability.
The existing FOSS alternatives just aren't at par. I've tried FreeCAD, OpenSCAD, etc., and most mechanical engineers/CAD specialists wouldn't touch a code-based editor. They're certainly better than before, but the rough edges exist and some of them appear in areas that need to Just Work (like assembly and drawings). From my understanding of their contribution graphs, private organizations putting even two or three full-time developers working on those projects could push them much, much closer to being a drop-in replacement for a lot work that gets done, and potentially even save those orgs some money in the short-term by reducing the total number of seat licenses they need in to function.
Edit: Added mention of KiCad, bc that's kind of important too