FPGAs for instance can emulate analog behaviour with a finer resolution than DSPs (some ns) overcoming the limits of the sample rate and it's granularity. The output does not necessarily have to be 192kHz, but can be much higher directly driving the AA filter with some MHz -> less tradeoff regarding the edge frequency and non linearity of the the anti aliasing filter, but the advantage is small only. The available resources are better invested into the number of voices.
The biggest advantage of FPGAs is pipelined processing, when used correctly. A current FPGA runs on e.g. 200 MHz and produces around >4000 channels for sample rate of 48kHz. To use 192kHz DACs, e.g. 1024 channel are processed during one sample period.
My current synth offers in double structure 8192 voices. You might get impression here: