As for the cost of a hypothetical neo geo cart being ludicrously expensive, well... krikzz would probably sell them for $500+ but his profit margins are very large. They wouldn't cost nearly that much to manufacture.
You can buy high speed low power 64MB DRAMs for like $3 and low end FPGAs/CPLDs are similarly cheap. You'd spend more money on voltage shifters, they're even cheaper but you'd need so damn many of them for all 200 pins! I can't foresee any part that costs more than $5 unless you need a beefier FPGA to control all the little routing FPGAs/CPLDs and SD interface, and those are about $15.
You can check PCB printing prices at pad2pad. I don't know what the exact dimensions are, but for 4x8" @ 50 units they're only $8.
Teh assembly is where it can get expensive unless you do it yourself with masks and an oven.
And this assumes we can get the necessary engineering help for free or at a steep discount. If you bribed krikzz or went to a firm you're looking at tens of thousands of dollar. And the parts/manufacturing to build a prototype or two would run you $500 - $1000.
But once you have the design drawings the parts themselves are inexpensive. And those are today's prices. By the time it goes into production a non-profit venture could probably break even at $100 each if the development and assembly costs can be kept down.
I'm sure it will happen eventually and have purchased an AES in anticipation while they're cheap. The only question is if it's going to be a crummy chinese device, an expensive proprietary cart, or a crowdfunded opensource project.
Just throwing that out there since the idea's been kicking around in my head awhile. The technology has gotten to the point where something like this is starting to look feasible. If a couple of engineers more competent than me can get onboard, and people are willing to donate for the development costs, it can happen.