For a lot of people the answer is that they associate computers with work and would rather not use them to make music. There's a definite "I don't hate computers, but I do like hardware" contingent though.
I've pretty much said the same things there that I did here, but I'll quote myself a little bit.
I don't want to overstate that one too much. There are some excellently designed software synths. Some are lean and mean, and immediately feel great to use. Some are a bit more inclusive but still designed in a tight and focused way (hello ArcSyn). Still, it's rare for a software synth to be as laser-focused as something like the 0-Coast or Lyra-8, and more common to give in to market demand and provide 1500+ presets and 37 filter types and 14 assignable LFOs.Hardware often seems more intentionally and thoughtfully designed than software does. A module is more limited and fixed in a lot of ways than software, but that means it's more focused. You can't just tack features on willy-nilly or add another three pages of knobs. There are of course some exceptionally well designed software synths, but many of those take their cues from imitating hardware synths.
I will stand by this assertion, regardless of how woke some people believe they are. And I will gladly paint myself with the same brush, because of how much my universe opened up when I got into modular and it made me feel like a beginner again for a while.The software-based synth community has massive blind spots about methods of synthesis and composition. When you try to point them out, it's like trying to explain rock climbing to a 2-dimensional being from Flatland. I think you have to get into modular for a while to really see what's missing.
When I first started messing with hardware after a long period of 100% in the box music, it was an adjunct to using software synths in the same way as before... except that I had to bounce to audio or just render the whole thing live. Worth it, because of that Microbrute sound that "just happened" as a result of turning knobs. But then I found a lot of the other hardware I tried really wasn't worth it. I hadn't quite figured out what I liked, yet, and it was a process that took a while.
Then when I got into modular, I thought "I can't even do some of this in any software that I've tried." At that point, I could have taken a year or more off of actually making music to get extremely cozy with Reaktor or Max. Instead, I got deeper into modular hardware because it immediately inspired me to make sounds and music every day.
And then as I got deeper still, over the course of about 18 months, my workflow was less "DAW-ish" as I thought of it, and more modular (with the DAW acting as mixer, effects host, recorder, and occasional supplementary synths and samples).
Just all the weird quirks and crosstalk and noise and saturation and signal loss and whatever else... touch plates that depend on skin resistance, noisy PT delay chips, funky DACs, and so on. Things that "just happen" but take a huge effort from genius developers to emulate in software, if they bother to. Look at the heroic efforts Plogue have been putting into emulating old FM chips for instance.There's a lot that goes in in hardware that isn't modeled (or at least, not often or not well) in software -- supposedly undesirable side effects, or less common behaviors that are simply ignored, or because of technical limitations. Those add up to a lot of character. People will speculate about "analog warmth" and oscillator drift and so on and they just run in circles with it.
I'm not going to argue that hardware universally is better than software because of this stuff. Software's f**king cool too. But there's a lot of gear that is just very special to me, for which a software equivalent is nonexistent and unlikely to ever exist.
I'd really rather software developers continued to innovate in new directions, and to forge closer integration with hardware instruments, than try to emulate hardware. Best of both worlds.