> - You may not distibute[sic] this for a fee, not even "handling" costs.
Also I think the GPL was pretty important to the kernel, since many companies, especially in the 90s, probably would have kept contributions private and their code closed without that gentle push.
OTOH maybe eventually most people would have switched to FreeBSD (or whatever free *BSD would have been the "mainstream" choice), just like they switched to Linux in our universe, since they thought that whatever value add provided by some proprietary unix wasn't worth it anymore.
In a hypothetical copyleft-free universe, sure, there would be a lot more companies using OSS to create proprietary products without having to think about what is a derivative work, linking and distribution restrictions. OTOH all those proprietary companies playing the "commodify your complement" game against each other would ensure that the quantity and quantity of OSS would continually be increasing as well, forcing those companies to continually innovate lest they lose their market to the free OSS alternatives. To repeat, hypothetically speaking, as we don't have an alternate universe to run such experiments in.
This should be "quantity and quality"
And in what concerns x86, they would just support it as well, as Solaris did. HP-UX and Aix also have supported a couple of CPU architectures, as did some of the others.
Just Irix was kind of married with MIPS.
Or Windows would just have won the x86 server room instead.
In any case, there are a couple of major surviving GPL projects, all competition against Linux on the IoT space are BSD/MIT FOSS POSIX clones, ironically one of them being sponsored by Linux foundation (Zephyr), so those around in 20 years will get to appreciate how much of GPL will still be left around.