Hi --
Just read through the thread. Bachus (and others) have done a good job explaining the current memory situation. Sorry I haven't posted much lately, but I'm just trying to play more and enjoy life near our grandson.
The hardware in Genos and even the SX700/SX900 is divided into two parts: host computer and tone generation. The host computer handles styles and sequencing. The tone generator section makes all the noise.
So, there is memory in the host computer (USER memory) and separate memory in the tone generator (voice expansion memory). The type of memory used in each is specialized for the task: eMMC flash for USER memory and NAND flash memory for the tone generator.
The tone generator is itself a very specialized, parallel computing system. It has its own DSP RAM, etc. The factory waveforms and user waveforms (AKA "samples") reside in the same physical NAND flash memory belonging to the tone generator. It is the same kind of NAND flash used in PC SSDs. Yamaha eliminated the SATA bus to lower cost and increase access speed. The expansion memory is soldered to the printed circuit board and, thus, is not easily removed and replaced. Even if you replaced it, as Bachus pointed out, the software would not recognize the additional memory capacity.
Yamaha's tone generator is designed for the long run. It is used in synths and high-end digital pianos, too. The tone generator has ports and features which are not currently used. They are for future products. You might feel gypped, but that's the way the game is played -- spread design cost over 5 to 10 years and multiple evolving product cycles.
Hope this helps -- pj