But how to create an encryption that can be transmitted over digital network ? There will still be interfaces that will quantify qubits to simpel bits, and no way to restore qubit states on the reception side: all you'll be able to get is a probability of getting 1 or 0.

So to transmit only 1 qubit, you'll need to transmit thousands bits (remember that strengh of qubits is that this strength only occurs when their probabilistic states are near 50%/50%, and only if a qubit has a probability of 33%=0 and 66%=1, it encodes 1 bit and is not formally more secure). It is then just more complex to implement, but it still offers significant advantages however in terms of inherent parallelism.

So what will quantum processing chip will look like: a 3D crystal not working with binary gates but with refraction of light in the crystal, equivalent to transform binary processing to a superposition of waves with a Fourier-like transform: the information is not localized in each binary gate, but diffused in the crystal, it is inherently fault tolerant, largely immune to noises or defects in the crystal and this is probably a major step for making chips as we've almost reached now the maximum integration level (density of gates on a mostly 2D layout with just a few 2D layers) where quantum physics applies everywhere and starts producing some unpredictable states on all surrounding gates (complicated by ther unpredictability of disorders produced by heat and artefacts where individual electrons create superposition of probabilistic lobes).

Let's go further, all the research will be on how to design internal crystals geometries. And instead of using electrons (or absence of electrons, i.e. "holes" in semiconductore), we'll use tiny photons (and absence of photons) and their capabilities of creating interferences over very large volumes without being blocked completely by the material or defects of crystals (the same defects that cause now many rejections of built dies which are now too expensive to build and that cause severe problems with hot temperatures in power-inefficient CPUs).

A binary logic could still control the internal geometry of atoms in the crystal (equivalent to "programming" it), using for example magnetic fields to reorient the atoms and rearrange interference gates, then data will be fed as polarized photons (emitted by a laser diode on one side of the crystal, and photon detectors all around the crystal to detect the produced interferences and quantify them to produce a binary result with some probabilities. And the crystal will still work to do the processing even if it has some internal defects or is perturbated locally by temperature.

But this will only be for local processing (not extending outside the new chip, unless the photons are transported over fibers to longer distances to feed another remote processing crystal.

For long distance (the Internet) we don't have any fiber network connected this way, all routers use binary logic and binary quantification (exact polarization of photons is completely lost at the reception site). We don't have any pure photonic network to build the worldwide internet we know today.