Memory wall
The growing gap between how fast processors compute and how fast memory can feed them data. Often the real limit on performance.
Also: von Neumann bottleneck, memory bottleneck
Processor speed grew much faster than memory speed for decades. In 1995 William Wulf and Sally McKee called the result the memory wall: eventually processors would spend most of their time waiting for data. Caches, prefetching and out-of-order execution hide some of the wait. For AI, the problem is sharper, since large neural networks need huge amounts of data moved for each calculation, which is why accelerators use HBM stacked beside the chip. Moving data now often costs more energy than computing on it.