Semiconductors · resource ledger

Memory (DRAM & HBM)

AI memory demand is squeezing the same production capacity used for ordinary DRAM.

AI-driven3:1HBM-to-DDR5 capacity trade ratio reported by Micron

What the evidence supports

Micron says producing HBM uses roughly three times the capacity needed for the same amount of DDR5. It has also raised its memory forecasts because of AI data-center plans. This is direct evidence from a producer that HBM competes with other DRAM products for limited factory capacity.

How the effect works: HBM uses more wafer capacity and advanced packaging per delivered bit than standard DDR5. When demand grows faster than factories and packaging lines can expand, suppliers devote more of that limited capacity to higher-value AI memory.
Who pays or benefits: AI-server buyers face scarce HBM allocations first. Conventional server and consumer-memory buyers compete for the remaining DRAM supply. Available data does not measure the retail price effect.
What limits supply: Cleanroom construction, lithography tools, manufacturing yields and advanced packaging set the near-term limit.
Attribution boundary: The producer statement measures the production trade-off, not AI’s share of retail RAM price changes.
Evidence that changes the grade: New supply that expands HBM without displacing conventional DRAM lowers the grade. Independent capacity data that contradicts Micron’s trade ratio also lowers it.

Sources

Public data, agency work and company reports

Claims related to this market