Video codecs generally use one or another type of YUV format so lots of display hardware can already use it, on something like a mobile SoC or settop box no one can afford the memory bandwidth to do the conversion anyhow.

Carmack is giving an extremely simplified version of the world as it exists today. Your hardware can already use YUV formats, maybe it is even outputting YUV because you chose a resolution and bit depth that forced it to for bandwidth or clock constraints. Complex apps like browsers that do their own compositing can already choose to use YCbCr formats when supported for textures.