I/O for chips come from the chip's edges, and two of the four edges of Nvidia GPUs are dedicated to HBM, according to Patel and Ontiveros. That leaves two edges for connectivity, where Nvidia is more likely to
Nvidia and AMD are known to be voracious for high bandwidth memory chips, but there are other players in the highly competitive AI market who might be keen to snap up HBM stock to avoid being left high...