FDM was actually developed and commercially marketed quite a long time ago. The patent expired, after which it became fair game for tinkerers and hobbyists. The same is now happening with SLA, although that one turns out to be a bit harder to crack and there doesn't seem to be a coordinated effort to improve the technology like there was with FDM in the form of Reprap.
FDM is never going to have a finish as good as some of the alternatives, but it's relative simplicity and useful printing materials will probably make it a popular choice for a while to come. It's really amazing what you can do with it, although some people fall into the trap of applying it to every problem whether it's a suitable solution or not. I'm glad the days of people printing endless amounts of useless trinkets seem to be passed and people are more and more using their printers for more useful prints.
What was developed commercially was so far from what we call FDM now as to be a completely different product. Numerous products, all suffering from the "SONY trying to own it" disease, each entirely devised as closed ecosystem. All different idiotic mechanisms to try and make 3DPrinting work like conventional printing where you had to buy your supplies from the printer manufacturer, only this time lock out the cartridge cloners for good.
Shocking that nobody could make it commercially viable with the same "consumable supplies that cost more by weight than gold" business model as inkjet printing.
FDM as we know it using off-the-shelf actuators, global-standardized filament sizes and formulas and platform-independent generic controllers was entirely devised by the hobbyist market.
What we really need now is a tertiary product that can directly print molds for metal-casting. I've already done lost PLA at my local Hackspace; while it is theoretically "amateur-able", there really needs to be something more straightforward and single-step to a usable mold, and so things like proper sprue and vent sizing/location can be part of the prep software and safe defaults preprogrammed.
Would be a total game-changer, even if you could only get down to like .4mm layer height, with something the size of a Tornado you could do real, usable machinable castings in a usable framework of time & assache. Many applications could STILL be usable roughcast. We have small, cheap inductive kilns that can do a couple pints of non-ferrous metals on a benchtop already.
mnem
ZZZzzzZZZzzz...