Global Memory Shortage Driven by AI Infrastructure Demand

A global shortage of semiconductor memory, including DRAM and NAND flash, has emerged as manufacturers prioritize high-margin components for artificial intelligence infrastructure [1]. This supply constraint, which tech media has dubbed "RAMmageddon" or "RAMpocalypse," is expected to persist until at least 2030 as production capacity is reallocated away from consumer electronics [1].

Key takeaways

• The shortage is driven by a structural shift in manufacturing capacity toward High Bandwidth Memory (HBM) required for AI accelerators and data centers [1].
• DRAM prices rose by 172% throughout 2025, leading some manufacturers to halt new orders or exit consumer product lines [1].
• Major PC manufacturers, including Dell and HP, have reported unprecedented cost escalations, with memory costs for HP rising to 35% of total build materials [1].
• Geopolitical tensions and trade barriers, including restrictions on selling manufacturing equipment to Chinese entities, have further constrained supply chains [1].

The Shift to AI-Driven Production

The current supply crisis differs from the 2020–2023 chip shortage, which was primarily caused by pandemic-related disruptions [1]. Instead, the current scarcity is a result of manufacturers like Samsung, SK Hynix, and Micron Technology reallocating wafer capacity to meet the surging demand for AI-ready data center hardware [1]. Because HBM production requires significantly more wafer capacity per bit than standard DRAM, the supply of conventional modules for smartphones and consumer PCs has contracted sharply [1].

This reallocation has created significant operational challenges for major electronics firms. By late 2025, Dell Technologies reported that it had never witnessed costs escalating at such a rapid pace, while Lenovo noted that its memory inventories were 50% above normal levels to hedge against further price hikes [1]. Smaller firms have faced even greater difficulties, as suppliers have begun requiring prepayments or shorter payment windows to manage market unpredictability [1]. Furthermore, the supply of essential manufacturing components, such as the glass cloth used for high-speed data transfer, has been unable to keep up with the demands of major chipmakers like Nvidia, Apple, and AMD [1].

Why it matters

The long-term outlook suggests that the memory market will remain strained for several years. Analysts and industry CEOs anticipate the shortage will last until at least 2030, as AI workloads are projected to consume roughly 70% of total data center capacity by the end of the decade [1]. While some companies have attempted to mitigate these costs through long-term supply agreements or aggressive pricing strategies, the broader market continues to face inflated costs and supply volatility [1]. As manufacturers continue to prioritize server-grade components, the consumer electronics sector faces a sustained period of higher prices and limited availability for standard memory modules [1].