In fact, the loudest possible point in a digital signal is rated at 0 dBFS – that's decibel full scale. 0 is the maximum level, and the minimum level is defined by the amount of bits you have to represent sample amplitude in pulse code modulation. For example, the loudest sample in 8 Bit PCM would be 1111 1111
and at the other end of the scale, 0000 0001
.
This means there is an absolute maximum that you can transmit – simply speaking – from the file, to your sound card, until the digital signal is converted into an analog one.
In fact, 0 dBFS are hit quite easily and it's very common. Peak amplitude normalization for audio files works this way: you scale the audio levels in a file in such a way that the peak(s) sit at 0 dbFS.
Therefore, the "loudest" possible digital file is a sine wave hitting 1111 1111
constantly in pulse code modulation, at the frequency where it is most perceptible by humans. This doesn't mean it has to be the most damaging. Certain speakers will react differently to different frequencies. If you create a sound that hits the fundamental resonance frequency of a speaker, you could introduce vibrations that could physically damage the speaker itself.
Considering all that, the digital volume of a file does not matter though. If a digital file is not loud enough, you can always normalize it to 0 dBFS via digital signal processing. It's the analog volume, i.e. whatever makes your speaker cones move, that determines the actual loudness.