100% this. A tube gain stage in isolation isn't especially difficult to imitate using solid-state electronics or DSP, but sounds nothing like a tube amp.

>Utility for guitar amps would be for "hybrid" amps with a tube preamp and tone circuit feeding a solid-state power amp.

The Vox MV50 amps make a decent attempt at this, pairing a Nutube with a class D power stage. It's tiny, it's reasonably loud and it sounds pretty good.

I suspect the lack of interest from other manufacturers is mainly due to cost; for the price of a single Nutube module, you could add a decent DSP chip and gain a whole bunch of extra functionality. An ornamental 12AX7 and an orange LED has better showroom appeal and costs significantly less. Purists will still want real tubes and real transformers, budget-oriented players will want an amp with all the DSP bells and whistles, leaving the Nutube caught in an uncomfortably narrow niche.

Also, a tube has absolutely no latency.

I think that's the primary key here. We can decently emulate a tube amp in software even (and that can improve indefinitely as our knowledge improves), but we'll never create drop-in circuit element emulators either in silicon or software without adding latency.

Subjectively, the one effect often missing from preamp-only tube amps I've tried is power stage sag[1]. I've seen DSP software with it (or rather, a knob for it). In that respect, I'm not sure the solid state / preamp tube / all-tube segmentation can be justified anymore: modern modelling amps sound terrific (subjectively, to my ears), offer sonic variety (because why not), and have a completely different idea of value (comfort, power and size); With that, the NuTube just feels like it's going after an audience who won't be convinced that electrons flowing through a DSP are made of the same quantum stuff, which is unfortunate.

[1] Demo: https://www.youtube.com/watch?v=INi6lepBZt0

Compared to that, the dynamic of a tube are pretty simple, though creating a truthful emulation of a full amplifier is still a hard.

But listen to any Herb Alpert recording, and then listen to any synthesized trumpet. The gulf is as wide as the Atlantic. (Part of the reason is that Alpert plays with the notes and the transitions, while a synthesizer just emits a note.)

Well, maybe a crap one, or one not designed for that. Trumpets have some convenient physics that are (comparatively) simple to model, the problem you're thinking of may have more to do with the control system (i.e. not based on actuation of valves, but skipping right to the oscillators) than some inherent weakness of synthesis.

(I played a trumpet as a kid, so I suppose my ear for it is more discerning than most. My talent proved elusive, but the lasting value was I can appreciate how incredibly good Herb Alpert was. He could play like Fred Astaire could dance and Ernst Udet could fly.)

From my experience, overdriven guitar signal formed by ubiquitous FET-based approach driving a saturated tube power amp can sound gorgeous, tight and punchy, and feel very "natural" and responsive. 12AX7s into your ordinary solid-state power amps - not so much.