I don't think this is true at all. Tracker software in the 90s largely did not have MIDI support, it was an entirely different category of software.
Although it is true that trackers in the PC era grew out of the Amiga (MOD) scene where sample playback was standard, there were also trackers for the 8-bit home computers which were used to create actual chiptunes on SID or the AY-3-8910 and its derivatives.
In the context of demoscene, chiptune-influenced compositions were still extremely popular, because including full-length instrument samples cost far too much space to distribute inside an intro or cracktro. Songs were made up of tiny bursts of white noise and tightly looped single-cycle waveforms, and the bulk of tracker "effects" (pitch bends, arpeggios etc) came directly from the techniques used to coax more tonal variety from sound chips of the 8-bit era.
Tracker music used for intros might not literally be "chiptunes", but they definitely took a lot more influence from chip music than from popular music. For example, this is the track that was bundled with Scream Tracker 3: https://www.youtube.com/watch?v=R3hpnmANSMw (64-Mania by Edge)
I'm talking explicitly about the PC era. The 8bit machines were far more limited and that music was chiptune in every meaningful way.
But when you get to PC's you didn't have one set in stone hardware FM synth as a standard. Your hardware varied and it was absolutely controlled via MIDI even if you lacked the hardware to interface with external MIDI devices.
Depending on the hardware you had at the time your soundcard gave you either FM Synthesis such as with the classic SoundBlasters, Adlib etc. A fixed sample library in the case of devices like the Sound Canvas. Or with the later more advanced soundcards you had wavetable synthesis starting with the AWE32 which opened you up to soundfonts.
If you had a Sound Canvas or similar "sound module" as they were called, there were many, you absolutely did have external MIDI capabilities because it was a requirement for the device to work. You had to have a MIDI interface card or box. I had a 4 port Roland card in the mid 90's. I don't remember the model but they stopped making drivers after Win98 for it.
I think you are confusing trackers with the more general capabilities that appeared in PC sound cards of the 90s.
Although there were sound cards that supported connecting external MIDI instruments, and there were sound cards that implemented the so-called General MIDI set of instrument sounds on the card itself, tracker software did not use any of this functionality. Trackers loaded samples into memory and mixed the audio in software. This is how Scream Tracker (for example) could output directly to the PC speaker and did not require any particular sound card.
Because everything was mixed in software, the complexity of a composition was limited by the CPU and memory of the computer itself. It didn't matter what sound card you had. On under-powered PCs, loading very large samples or going above 4 or 8 channels of polyphony was not a viable option. This meant that tracker musicians operated under similar restrictions to the chip musicians of the 8-bit era. In fact, quite a few tracker musicians got their start in the 8-bit era, which is why a lot of demoscene music sounds similar to it.
(Edit to add: in case it's not clear, later trackers did use hardware mixing and effects if they were available, but I'm trying to explain more about the culture of the scene and how it influenced the type of music that came out of it.)
I used to make music on modtracker on a celeron, i could load more than 8 tracks but not very long or not playing them simultaneously, so i resourced to downmixing some parts and re-loading them as one single sample, then combining that with hardware synths on a multitrack (portastudio actually).
That's cool! I remember seeing those sorts of tricks in Amiga mods especially.
I jumped from the 8-bit (3 channel) world straight to a 486sx, so getting 16 channels seemed incredible at the time. I soon found out that going above 8 was ill-advised :) The sample size was limited to 64k too, so even if you did bounce tracks together you might only be able to load a few seconds.
File size in general felt like a big deal back then. Going over 100k for the whole mod was considered pretty excessive. I think Fast Tracker 2 ushered in the era of larger samples. I suppose it coincided with modems getting faster too, so people were less reluctant to download songs that got up into the megabyte range.
The demoscene stuff always seemed especially clever to me, because they didn't have the benefit of dedicating the entire computer just to mixing and playback - they needed to display graphics too! I think you can often tell when composers came out of the demoscene by listening for stuff like "J37" arpeggios and the sort of breakdown like in 64-Mania where you play the same sequence in two channels with slightly different settings to create a phasing or chorus effect.
Amiga and Atari ST weren't 8-bit and overlapped the PC era. These were the platforms where the term chiptune emerged and thrived as part of the demo/cracking scene.
The original SoundBlaster, SB Pro & SB 16 could and did output PCM sound, and there were mod players & trackers written for them that enabled the creation of chiptune style music-- generally music with very small PCM samples to sound like the C-64 or Atari ST. These uses on PC had nothing to do with MIDI.
Although it is true that trackers in the PC era grew out of the Amiga (MOD) scene where sample playback was standard, there were also trackers for the 8-bit home computers which were used to create actual chiptunes on SID or the AY-3-8910 and its derivatives.
In the context of demoscene, chiptune-influenced compositions were still extremely popular, because including full-length instrument samples cost far too much space to distribute inside an intro or cracktro. Songs were made up of tiny bursts of white noise and tightly looped single-cycle waveforms, and the bulk of tracker "effects" (pitch bends, arpeggios etc) came directly from the techniques used to coax more tonal variety from sound chips of the 8-bit era.
Tracker music used for intros might not literally be "chiptunes", but they definitely took a lot more influence from chip music than from popular music. For example, this is the track that was bundled with Scream Tracker 3: https://www.youtube.com/watch?v=R3hpnmANSMw (64-Mania by Edge)