With global RAM pricing spiking due to a surge in AI demand, what does this mean for buying a smartphone in 2026?
If you’ve considered building or upgrading a PC lately, you’ve likely noticed the shockingly high price of memory. RAM kits that sold for reasonable sums just a year ago now routinely cost two to three times as much, and prices continue to rise while some sets are increasingly challenging to obtain entirely.
Smartphone owners may feel insulated from these swings, since we don’t buy or install mobile RAM ourselves. However, the same forces driving up PC memory prices are quietly reshaping the smartphones we may plan to purchase next year. Here’s what you need to know.What’s causing extreme RAM prices?
The price of RAM components has increased over the past few months, primarily due to the surge in demand for AI-related products. Data center providers, such as Amazon and Oracle, have been acquiring DDR5 — currently the fastest motherboard memory available — to meet the growing demand for high-performance computing in the cloud. Worse, modern AI servers can feature several terabytes of DDR5, compared to older servers that may have used just 128GB or 256GB of DDR4. AI infrastructure is absorbing an outsized share of global DDR5 output, meaning that PC consumers are directly competing with some of the biggest tech players for RAM modules.
However, the larger culprit is that the AI boom is pushing RAM manufacturers to divert capacity to other memory formats. High Bandwidth Memory (HBM), for example, is used for the large virtual memory pools packed into NVIDIA’s dedicated AI cards, such as the H100 and GH200. HBM is built using the same wafers as DDR4 or DDR5, while the profit margins can be 2x to 5x higher.
As such, Samsung Electronics, SK Hynix, and Micron Technology — the big three that account for at least 75% of global RAM production — have shifted a significant amount of DRAM to the far more profitable HBM since 2023. This has led to a reduction in capacity dedicated to the vast range of other consumer and enterprise RAM types, including DDR, LPDDR, and GDDR, among others.
To highlight just how ravenous the AI behemoth has become, in October 2025, OpenAI signed letters of intent with Samsung Electronics and SK Hynix to supply memory components for its large-scale “Stargate” AI infrastructure project. Industry reports suggest that anticipated memory demand for this initiative could reach as high as 900,000 DRAM wafers per month — a volume that, if fully realized, would command a substantial share of global DRAM wafer capacity.
