I would add a minor quibble about the impedance part. It is only the other half of the equation. After all, impedance is the R in the total power, which is the familiar P = UI, U = RI part of the equation. One consequence is P=U²/R, square of voltage divided by impedance. The higher the impedance, the larger voltage is needed to produce certain level of power, but voltage grows in square so it will not be all that much in the end.

Depending on design, that power can then be translated to acoustic energy at some efficiency or other, and manufacturers either relate that in terms of voltage or milliwatts to achieve certain SPL. From what I can see, most headsets should be deafeningly loud with very low power figures, to the tune that even 1 mW is more than your ears can take without suffering damage. Milliwatts of power are such an astonishingly low figure, and I think it is a crime that not literally everything has enough power behind it to drive headsets well enough. You don't need a separate amplifier to make milliwatts of power, that much is achieved by pretty much anything. I confess I also do not quite understand how headset can have multiple hundreds of ohms of impedance. What is it doing? Do they put big resistors there? A voice coil really shouldn't present that much of resistance.

In any case, I would personally steer away from headsets that require an amp. The Apple USB-C DAC is a great example of USB soundcard that can deliver the few milliwatts more or less perfectly, and it costs all of $10. Hopefully in the future, all DACs on all devices are decent enough, and headsets can be simply driven by any random thing even if it isn't a real amplifier, because there is clearly solution space where the problem is tiny -- just keep impedance anywhere reasonable, say somewhere in 15-50 range, and produce enough SPL per milliwatt, easily achieved by many designs already and there really is no sound quality compromise here as far as I can tell.