Size is a big dictator of it. The other major way of storing in memory is a flip flop, which uses several transistors each in a feedback loop constantly regenerating the 1 or 0 they're storing. A floating gate MOSFET achieves this with just one transistor per bit (and then multiple bits per transistor by storing various voltages on the gate). Same for why they make layers, just to achieve the density necessary. One limiting factor is silicon die size. The larger you make the chip, the more likely there is a defect. Large chips are expensive because of this. Lithography make use of light to project very fine details (and I assume they use something of a smaller wavelength now). That coupled with light-sensitive coatings allows for simple steps to produce a massive number of transistors on a wafer of lots of chips. This lends itself to stacking of layers, as they do massively with these flash memory chips. They just keep adding layers, then etching the features off. Someone I know thinks that people just invent big things outright. I think it mostly is constant refinement of existing technologies and ideas. Chip production is similar to photography, projecting an image, exposing the medium, then developing. Transistor technologies have been developing for many decades. Flash memory is based on dynamic memory (DRAM, used in normal memory sticks in computers), which uses the same gate structure just not floating.