The price war for storage chips may be about to enter a new phase. A recent market outlook from a research institution shows that by 2026, the global commodity memory market will face a "severe supply shortage," and this is not just a short-term supply chain fluctuation but a structural gap driven by the AI revolution.

The data on DRAM is quite crazy. Predictions indicate that the average selling price (ASP) of DRAM could increase by 88% in 2026, directly raising the previous expectations by 35 percentage points. Even more outrageous is server DRAM—mainly used in AI training and inference compute centers—with prices expected to surge by 144% year-over-year. For the mainstream 64GB DDR5 RDIMM, the price could reach $620 per module by Q1 2026, which is 38% more expensive than now.

The expected increase for NAND flash memory has also been adjusted from 44% to 74%. The underlying logic is quite clear— the proliferation of AI Agents is consuming massive amounts of storage space and computing resources. Major data centers have already begun to expand capacity rapidly to meet this wave of demand. However, capacity cannot keep up with the growth in demand, ultimately shifting pricing power entirely to the sellers.

What does this mean? In 2026, anyone looking to upgrade storage configurations or expand computing power will face significantly higher costs. This will have a profound impact on the overall costs of Web3 infrastructure investments.