Citrini Research forecasts a 28.7 EB global DRAM shortfall by 2030 — about 18%
of 2030 demand and versus current global capacity of roughly 40 EB. Researcher
Zephyr estimates 2030 DRAM demand (including HBM) at 157.5 EB versus supply of
about 128.8 EB; standard DRAM is the principal bottleneck with supply ~91 EB
versus demand ~120 EB, widening the gap to about 25%. Citrini says capacity
additions from Samsung, SK Hynix, Micron and Chinese fabs are likely to be
quickly absorbed by surging AI-driven demand — large-model training/inference
and HBM-based accelerators — which also boosts server DRAM requirements. Under
this tight balance, DRAM ASPs could remain elevated at roughly $1.5–2 per Gb,
keeping memory costs pressured for servers, PCs and consumer electronics.
