The reason SGI failed, and eventually Sun too, isn't because the world "chose wrong", but because their performance simply did not keep up with x86.
When these RISC-based workstations were initially released their performance, especially at graphics, was well beyond what a PC could do, and justified their high prices. A "workstation" was in a class by itself, and helped establish the RISC mystique.
However, eventually Intel caught up with the performance, at a lower price, and that was pretty much the end. Sun lived on for a while based on their OS and software ecosystem, but eventually that was not enough especially with the advent of Linux, GCC, etc, as a free alternative.
Ivan Sutherland described the reason [1] why PCs won a long time ago. Basically a custom tool may do a thing "better" than a general purpose tool for a while, but eventually, because more resources are spent improving the general tool, the generalized tool will be able to do the same thing as the specialty tool, but more flexibly and economically.
Sun had the perfect opportunity with Utility Computing around the mid-2000s but when cloud took off we had Oracle buying SUNW. They killed Sun Cloud which had the opportunity to be big, vast, and powered by JAVA hardware.
Sun Microsystems was a company like no other. The last of a dying breed of "family" technology companies.
I was at the MySQL conference when it was announced that Oracle was buying Sun. It just took all the life out of the conference. All the Sun folks were super pissed off. Truly the end of an era.
I was there too. It certainly felt "timed" to maximize the sense of deflation for people working on MySQL. Perhaps it was just coincidence. IIRC Larry Ellison said that the crown jewel in the deal was actually Java.
I remember that time. It felt like Sun was on death's doorstep since the dot-com crash. On the hardware side, the market was flooded with used Sun hardware. On the software side, Linux was "good enough" for most workloads.
Sun really struggled to make full use of their multicore systems. That m:n process model is coming back with fibers and libuv, but we have programming primitives and a deeper roster of experienced devs now than we did then. Back then they caused problems with scalability.
There were times when Java ran better on Intel than on Solaris.
When these RISC-based workstations were initially released their performance, especially at graphics, was well beyond what a PC could do, and justified their high prices. A "workstation" was in a class by itself, and helped establish the RISC mystique.
However, eventually Intel caught up with the performance, at a lower price, and that was pretty much the end. Sun lived on for a while based on their OS and software ecosystem, but eventually that was not enough especially with the advent of Linux, GCC, etc, as a free alternative.