So where's the $15k super machine for today? I'd absolutely look to pay that for something with a similar advantage over a MacBook Pro.
Industrial simulation: these days the optimal approach is to get access to CPU is to crunch on headless servers, ideally in the cloud. Expand and contract computing power on a needs basis.
Front-office development: Objective C and Interface Builder was much better for prototyping ideas than C. Modern python/perl/racket/go/c#/etc are much further ahead. There were no mainstream dynamic programming tools more powerful than awk when NeXT was launched. Now the most common expression of hackers is dynamic programming on fabulously powerful free tools. High-quality operating systems are also free. Documentation is free. And it's all far better than what people used to pay for.
Deployment: interactive user interfaces are now delivered through the a web browser. Previously a lot of workstations were sold to run custom internal apps.
Save your money! Get a digital ocean account, and run your macbook until the power supply dies. Perhaps in another ten years google-style web-browser laptops will be ubiquitous.
Users can run a CPU-only version of Pro Tools called HD Native which (along with its competitors like Logic, Sonar, etc) are useful and powerful programs. However they lack the power and depth of plugins available for the hardware-accelerated version of ProTools (and other hardware-accelerated plugins like UAD).
I think the Holy Grail is and has always been to "go native" and get rid of these sorts of accelerator cards as money is better spent on the general-purpose devices that retain their utility better. But we still aren't there yet - when you load up even the best CPUs with lots of convolution processing, too much latency is introduced. So yes, I think there's at least a niche market for the $15K general-purpose computer that truly represents a "bottomless pit" of audio processing power.
I haven't even touched on video editing or 3D rendering, both of which can be even more CPU intensive. When you add all these niches together, they still look like the original power-creative niche Apple was targeting 10 years ago with the Mac Pro and the Macbook Pro, and we still want more power.
> There were no mainstream dynamic programming tools more powerful than awk when NeXT was launched
Maybe I'm missing something, but:
1950s: Lisp and Fortran
1972: Prolog
1977: Bourne shell and Icon
1978: C shell
1979: REXX
1980: Smalltalk-80 (on Macs, even!)
1983: Korn shell
1983: C++
1987: Perl 1.0
1988: Modula-3, and NeXT computer is launched
My answer: your application programming wasn't actually that valuable to real humans. The web leapt 20 years ahead on the things that actually mattered to the world, and we lost some glitz that really didn't.
I think the web actually has far too many capabilities for most applications, and application frameworks are mostly a racket to keep application programmers and professional application users in a lucrative dance of wasting each other's time.
The vast majority of people can't even string together two functions, which js quite possibly the most powerful thing computers can do. For all of our effort we can't even give the people that most basic of capabilities, something five year olds can do with a little guidance. Because programmers are obsessed with GUI animations and digital cockpits and enabling professional app users to bend to the arbitrary whims of their supervisors. Because when we imagine "power" we don't picture collective control of outcomes that matter to us, we think of the sensation we get when we can individually make a fancy tool do arbitrary things.
I realize that's just a rant, so I'll ask: what outcomes matter to you, that you feel are impossible on the web?
In the 90s, most people were packing Windows 95 and if you wanted a Unix box you were going to pay for it. Silicon Graphics, NeXT, etc, were out of the reach of the general population and most developers. Only the CAD guys had them.
But many Windows users still complain today that the Mac is too expensive. So in that sense, if you want a Unix development machine you still have to pay a premium, just thankfully about an order of magnitude lower.
Of course, the democratisation would go much further: it is interesting that around the same time NeXT and Apple were merging, I first heard that some guy called Linus was busy porting Unix to the Intel architecture, and - unbelievably - he was doing it for free!