Ultimately, it all boils down to "How much heat can the driven device take?"
That is a function of Ohm's Law.
When various components are designed the designer has to take into account how well the various components can dissipate the heat that is produced by Voltage and Current, that is, the Power. Current (I) x Voltage (E) = Power(P). The higher the power-level that any component is subjected to, the more that device has to be capable of dissipating the associated heat.
If the nature of the component (such as, on the inside of a micro-chip) can NOT dissipate the heat produced at a particular voltage, then the designer has to find a way to get the desired effect at a lower voltage. The Current(I) level might end up being the same, but the lower Voltage(E) causes the net effect to produce less heat.
The primary reason for various voltages existing is the "heat" issue. Too much heat destroys things.
Many of the early "chips" operated at 12-Volts. They could do this because there were far fewer heat-producing internal circuits (gates, and such) and many of those "chips" were encased in ceramics. The ceramics worked very well for the particular high-heat situations - although, you sure didn't want to touch one of them!
As chips got smaller and more densely populated the heat became a more critical issue. The designers found a way to get the desired effects with 5-Volts without producing excessive heat.
All the time that this Heat/Voltage issue is going on, the designers are raising the Clock-Speeds, that is, the operating Frequency.
You may have noticed that after the Intel-486 MicroProcessor, all CPU's have (and REQUIRE) a cooling fan just to function. That is, to keep the CPU from Burning Up! At the very least, they become very unstable without the fan.
Time goes on, densities increase, and they find a way to use 3.3-Volts to get the desired effect.
As far as I know, 3.3-Volts is the current, normal, minimum voltage.
Now, the raising of Clock-Speeds causes another major consequence...
XL = 2 * pi * f * L
XC = 1 / ( 2 * pi * f * C )
The XL factor didn't cause much problem... there aren't very many inductors in a chip unless the chip is an ASIC (Application Specific IC).
The real problem showed up in the XC factor. Damned near every circuit within a chip has, at least, some capacitance. As you raise the Frequency, you lower the resistance of the circuit.
That would cause, or rather, allow, more Current to flow, then you get back to the I x E = P situation again. The voltage was brought down, thus reducing current, but the frequency was raised, thus reducing the resistance which in turn, allowed more current which tended to raise the power back where it was!
The designers are trying to raise the frequency to some incredibly large value (approaching infinity)... which in turn causes resistance to drop dramatically (approaching Zero! That is, a Short!) The lower resistance leads to higher currents!
It's a game of Rock-Paper-Scissors!
The game will continue on this way, probably including Super Conductors - someday - with incredibly small voltages, until they find a new technology that eliminates the heat issue altogether.
My thought on this is that they'll need to go to incredibly masssive parallel devices operating in analog fashion rather than digital fashion. That is, an array of partial-devices that drive a final digital device.
Hell, I don't know! What the hell do you think I am? A Rocket Scientist?
NOT !!