Yes, the microcontroller can drive "the cable". Each I/O pin individually is very powerful.

Just that it can't drive all cables, because of limitations on current for a full port or the full chip. And I would think a single FPGA would suffer the same problem. Driving 30-40 outputs when the other side isn't powered up because a fuse is blown or someone forgot to plug in the power connector is no simple task.

I'm currently trying to figure out what the implications would be of having a CPLD or microcontroller at the other end of each ribbon cable, to decode and map the data, while potentially running fewer data signals at a higher speed on the cable. This could allow the signals to be driven low-voltage differential for optimal noise immunity in relation to EMI. Nicest would be if something similar to SATA cables could be used between a main controller and a device-specific adapter board.