Cloud operators face a memory crunch that could force them to allocate up to 68% of their capital expenditure to DRAM and NAND procurement.

Soaring memory chip contract prices could push DRAM and NAND flash to represent more than two-thirds of cloud service providers’ (CSPs) capital expenditure by next year.

Taiwan-based analyst firm TrendForce projects that cloud operators’ total capital expenditure will nearly double this year, rising 98 percent, before expanding another 50 percent in 2027. This accelerated spending is fueled not only by expanded infrastructure investments but also by surging memory procurement costs.

Consequently, TrendForce estimates that DRAM and NAND flash will comprise 47 percent of total hardware spending this year, climbing to 68 percent by 2027. While the firm does not attach specific revenue projections to these forecasts, it anticipates that server DRAM component values will increase by 270 percent year-over-year by late 2026, alongside a 235 percent surge in enterprise SSD pricing over the same timeframe.

Industry observers note that surging demand for high-performance memory from major enterprise clients is compelling DRAM and NAND manufacturers to reallocate production capacity toward server-grade components. TrendForce projects that high-bandwidth memory (HBM), utilized in GPUs, and RDIMMs for servers will collectively consume more than half of global DRAM bit output this year.

These supply constraints are already impacting consumer electronics: PC prices have risen by double-digit percentages amid shortages of standard memory modules, contributing to a 5 percent decline in overall PC shipments. The smartphone sector faces comparable pressures, with annual shipments projected to drop by 15 percent.

The downstream impact on enterprise customers is increasingly difficult to overlook. As memory procurement consumes a larger portion of CSP capital budgets, providers are unlikely to absorb these expenses indefinitely and will inevitably pass the costs along. European cloud provider OVHcloud recently signaled this trend, announcing plans to raise service fees by as much as 87 percent to offset escalating memory costs—a move likely to be emulated by other operators.

Compounding the pressure, TrendForce notes that inflated memory expenses give server and AI accelerator vendors like Nvidia stronger grounds to increase their own pricing. Recent reports indicate Nvidia intends to raise prices by 15 percent on systems incorporating its Vera Rubin and Grace Blackwell chips. To sustain planned infrastructure growth under these conditions, cloud operators may be forced to escalate capital spending further, ultimately triggering additional price hikes for end users.