The global surge in artificial intelligence infrastructure has pushed DRAM and enterprise SSDs into a period of deep shortage and rising costs. According to Apacer CEO Chang Chia kun, the top 3 memory manufacturers are dedicating almost all new capacity to high margin AI hardware like HBM. This leaves module makers struggling to secure basic inventory as supply deficits are projected to last through the first half of 2027.
Memory manufacturers are prioritizing premium products like HBM and DDR5 for servers to capture maximum profit from AI data centers. At the same time, massive language models and AI inference workloads are driving up demand for high capacity NAND in enterprise SSDs to offset memory loads. This double squeeze means that standard memory allocations for personal computers and industrial systems are shrinking rapidly.
Module manufacturers have shifted their strategy from price negotiations to pure supply acquisition. Buying inventory at high prices is now a necessity to keep businesses running. Chang noted that buying expensive stock is better than having no stock to sell. This defensive stockpiling is keeping prices elevated across the entire market.
Consumer devices like smartphones and budget laptops are already hitting a wall. Brands in these categories are highly sensitive to bill of materials costs and have started to delay or reduce their orders. Niche sectors like edge AI, industrial PCs, and networking equipment are holding steady. These specialized buyers can easily pass the increased memory costs down to the final consumer.
The high price of DDR5 is pushing some system builders back to older technology. DDR4 is seeing a quiet recovery as Intel resumes supply of processors that support the older platform. Since DDR4 provides better cost efficiency, many budget sensitive projects are opting for this route. Limited production lines will likely keep DDR4 prices steady instead of dropping.
