Micron Warns Memory Shortage Will Worsen Through 2028

During its fiscal fourth-quarter earnings call, Micron Technology announced that the global memory shortage will tighten significantly in 2027 and 2028. Micron CEO Sanjay Mehrotra revealed that the company has already committed more than 75% of its 2027 output, shifting current customer negotiations to 2028. This supply squeeze has already driven up Micron's DRAM prices by a percentage in the high teens and NAND prices by about 30% in the fiscal fourth quarter. TrendForce expects this upward trajectory to continue, forecasting that conventional DRAM contract prices will rise 10% to 15% in the fourth quarter, while NAND flash prices will climb 15% to 20%.

The persistent shortage is already reshaping hardware configurations and purchasing budgets. IDC previously projected that average PC selling prices would rise 17% in 2026, and Lenovo CEO Yuanqing Yang confirmed that his company will adjust front-end pricing as material costs rise. To cope with these expenses, cloud providers and OEMs have begun downgrading server specifications. TrendForce reported a shift from 96GB and 128GB memory modules down to 32GB and 64GB modules. Meanwhile, Lenovo expects the overall PC market to shrink by about 15% in unit volume during the six months leading to March.

For enterprise practitioners, these market dynamics require a shift in infrastructure management. Counterpoint Research analyst Neil Shah advises organizations to secure multiyear pricing contracts for guaranteed computing needs, noting that shifting workloads to the cloud will not bypass these costs. Shah suggests stretching the lifecycles of non-critical back-office PCs and routine file servers from three years to five years. However, he warns against delaying refreshes for core infrastructure or buying legacy DDR4 memory to save 10% upfront, as manufacturers are converting production lines to high-bandwidth memory and DDR5, leaving older platforms without serviceable parts within two years.

Instead of purchasing new hardware, practitioners should focus on maximizing existing resources. Shah estimates that enterprises routinely waste 30% to 50% of their memory by provisioning for rare peak demands. IT departments can mitigate the shortage by right-sizing virtual machines, quantizing AI models, and batching workloads more efficiently to optimize the silicon they already have in place.