“In many cases of practical interest, only a single addition and a single on-chip copy operation are needed to replace a multiplication.”
Cussen and Ullman want to drop the scalar multiplier from matrix chips entirely. Multipliers eat time, die area and power, so a chip without them packs in more processors. The catch sits in “many cases of practical interest”, which is not all of them. The speedup stays a design argument until someone builds the chip.