Because it's not that easy due to Apple Silicon architecture display controller limitations with the current chips. Note that none of the PCI-e cards in the dem

Because it's not that easy due to Apple Silicon architecture display controller limitations with the current chips. Note that none of the PCI-e cards in the demonstration were GPUs. They were all network/storage/accelerators/etc. In the short term, I could see shoving an Nvidia GPU in a slot for offloading CUDA and GPU compute, but it wouldn't be really suitable for video gaming and such.