The semiconductor industry has spent decades pursuing more compute. Smaller transistors, higher clock speeds, larger processors, and increasingly sophisticated architectures have defined competitive advantage.

While compute performance remains important, a new constraint is emerging across AI infrastructure: memory.