Remarkably, there is nostalgia for the TI-99/4A.
Mark Wills recently wrote a Forth system for it: http://turboforth.net/
Apparently there are people around who still have these machines and keep them going for fun.
It is possible to make a cartridge with TurboForth on it, so you can develop Forth programs for your old machine --- you get interactive Forth development on the target itself --- this is not a cross-compiler.

What I have read (and Mark confirms) is that the TMS9900 was one of the best processors of its day, but TI screwed up the TI-99 and TI-99-4A design, partially in an effort to keep cost down and partially because they just weren't thinking very hard --- unfortunately, this was the only home computer built on the TMS9900.

This was recently posted on another forum:
http://spectrum.ieee.org/geek-life/hist ... oprocessor
(written by a microprocessor manager in TI's MOS division)
The Inside Story of Texas Instruments’ Biggest Blunder: The TMS9900 Microprocessor
The TMS9900 could have powered the PC revolution. Here’s why it didn’t

(I don't know anything about it other than what's in the article from a very credible source, but I'm always glad to see Forth ported to another computer.)

Yes, a good microprocessor, and a poor microcomputer. I think though that more RAM - and much faster RAM - can be added by cartridge. So, once expanded suitably, maybe a good machine.

(TI's corporate structure made it difficult for them to make the right tradeoffs for a successful micro. Same was true for IBM - they only succeeded with Project Chess by getting board-level approval to work like an internal startup and source hardware and software from outside. To get that approval, they made a straw-man proposal to buy Atari and rebadge the 800.)

I always assumed that the x86 was a given for the IBM-PC because it is similar to the 8080 and this would allow easy porting of CP/M programs to the IBM-PC. There was an 8080 assembler available that generated x86 machine-code, so 8080 programs could be converted directly into MS-DOS .com programs.

Wasn't the BIOS pretty similar between CP/M and MS-DOS machines?

A lot of people think the MC68000 was far better than the x86. To a large extent it was.
I always thought the x86 ISA was pretty good for the 1980s though. Using segmented memory it was able to address more than 64KB while still using 16-bit registers --- using 16-bit rather than 32-bit roughly makes your data half the size, so you get less memory usage, which is important because RAM was very expensive in those days. Also, using 32-bit data when you don't really need to slows down the system because you have a 16-bit data-bus.

The advantage of the MC68000 is that it was forward-thinking. The designers expected the limitations of the 1980s (expensive RAM chips and a 16-bit data-bus) to go away --- that was true --- in less than a decade, building a computer with 1MB or more of RAM and a 32-bit data-bus became realistic.