Generally they are, but they don't scale well. For example, with an FFT you could decode every radio station simultaneously with the same receiver and only log

Generally they are, but they don't scale well. For example, with an FFT you could decode every radio station simultaneously with the same receiver and only log more compute. So if you only want a single high SNR narrowband channel (even tunable) it's easier to do in analog, but if you want all 100 possible channels with stereo and/or some digital processing to add features or reduce noise... it's much better to do digitally. However, doing it with an FPGA/DSP would probably be significantly more efficient than a general purpose CPU.