Server Costs Leap As AI Demands Bite Into Global Memory Manufacturing Capacity.
Server Costs Leap As AI Demands Bite – Big CCTV users need to invest in servers quickly if they intend to undertake upgrades over the next 18 months as a supply squeeze threatens to double storage costs.
The cause of the price spike is rising demand from AI data centres that’s driving sustained increases in global memory and storage. DRAM, NAND flash and nearline hard drives are all under intense cost pressure as manufacturers prioritise high-margin server components over consumer supply.
Major memory producers including Samsung Electronics, SK hynix and Micron control the majority of global DRAM and NAND output. All 3 have reallocated advanced manufacturing capacity toward server DRAM and high-bandwidth memory for AI infrastructure, constraining supply for PCs, smartphones and consumer storage devices.
Market analysts report that conventional DRAM prices rose by around 55–60 per cent in a single quarter during 2025. Server DRAM pricing is forecast to rise by more than 60 per cent in the first quarter of 2026, with some suppliers planning increases of up to 70 per cent. Combined with increases recorded through 2025, this places memory pricing on track to nearly double by mid-2026.
NAND flash pricing has followed a similar trajectory. After reaching historic lows in 2023 following a prolonged oversupply cycle, manufacturers cut output sharply. By late 2023 those reductions flowed through distribution channels, triggering a rapid rebound. Spot pricing for 512Gb TLC NAND more than doubled within 6 months, with contract pricing following soon after. Retail SSD prices increased accordingly across consumer and enterprise segments and HDD costs are spiking alongside.
Training and deployment of large language AI models requires extremely high memory density and persistent storage. A single AI training node can consume hundreds of gigabytes of DRAM and multiple terabytes of flash storage. At data-centre scale, this crazy demand slurps up most of global output.
A case in point, high-density NAND products are now largely pre-sold months or years in advance and high-bandwidth memory production through 2026 is almost fully committed. Making things harder still, supply contracts that historically covered single quarters now span multiple years, with hyperscalers securing wafer allocations directly from manufacturers. There’s no way things will not get uglier still.
Alongside solid-state storage nearline hard drives used for large-scale data retention are also in short supply, with lead times extending beyond 12 months for high-capacity models. As a result, some data-centre operators are shifting workloads toward QLC flash arrays, increasing pressure on NAND supply and tightening these markets simultaneously.
Just to cap it all off, manufacturers have been cautious in expanding capacity. New memory fabs require multi-billion-dollar investment and several years to reach volume production. At the same time, producers remain wary of repeating previous boom-and-bust cycles that destroyed margins during earlier oversupply periods.
Capital expenditure is increasingly directed toward high-margin products such as HBM and advanced enterprise NAND, accelerating the wind-down of legacy products like DDR4, while demand for those products persists – especially in underfunded high use applications like electronic security, where upgrades can drag out over years.
The current situation is a supply and cost squeeze across DRAM, SSDs and HDDs with enterprise buyers with long-term supply agreements getting priority, while smaller end users facing much higher prices and much longer lead times. With AI demand continuing to expand and supply growth constrained, elevated pricing is expected to persist through 2026 and into 2027 – if not for far longer.
Distributors, integrators and end users of high-end servers should lay in stock as quickly as possible because this situation is only going to get worse.
You can read more about storage manufacturer Seagate’s growth based on growing demand here or read more SEN news here.
“Server Costs Leap As AI Demands Bite Into Global Memory Manufacturing Capacity.”
