This is a functional AM modulator with no active components (just a bunch of wire and iron oxide):
(This doesn't have voltage gain, but it does have some power gain: With impedance matching and a rectifier, it can amplify signals.)
How is this possible without any transistors, vacuum tubes, or even diodes?
Magnetic cores can saturate: once a certain amount of magnetic field is flowing through the core, it looses it's magnetic properties. For a transformer, this results in in sharp drop in coupling between windings once the total winding current reaches a certain point.
This gets more interesting with 2 matching transformers, connected so that the signal coupled by one is perfectly canceled out by the other:
Normally almost no AC passes to the output, as any voltage induced by one transformer is canceled out by the other. However, if one of the transformers is worse at coupling AC, for instance if the core saturated, then AC will be able to pass through, as the two voltages no longer perfectly cancel out. Saturating just one of the transformers requires adding an additional winding:
When enough DC is present on the control winding that the transformer saturates (for at least part of the AC cycle), then AC *will* be passed through to the output. This allows a small DC current to control a larger AC current, with no moving parts.
To make this into a AM modulator requires only a few changes: adding an audio transformer on the input, and adding a 4th bias winding:
The audio transformer is necessary becuase saturating the core, with the low turn count I used, takes quite a bit of current, but not much voltage. A transformer meant for audio use also presents a large inductance at RF, which stops RF coupling into the audio source, which could not only damage it, but also would also mess up the balance between the other two transformers.
The bias winding serves to keep the core close to saturation during all parts of the audio signal, this not only increases the modulation depth, but also minimizes distortion. The inductor in series with the bias winding keeps RF out of the bias power supply, which could otherwise mess up the balance between the two RF transformers.
I used a small 6mm ferrite core to wind the RF transformers, because a small core will saturate at a low current (in my case it saturates with around 400 mA in the bias coil), and a 100mH inductor as a choke on the bias.
Historically, before tubes and semiconductors were avalable, much larger versions were used to switch hundreds AC from a high frequency generator to send morse code.