The explosive global demand for AI semiconductors is creating a shortage of traditional memory chips, driving prices sharply higher and forcing manufacturers and consumers alike to adjust. As leading chipmakers shift focus to high-bandwidth memory (HBM) — the essential component for training AI models like those used by Nvidia — supply of standard DRAM and DDR5 server modules has plunged.
This supply squeeze has pushed the memory industry into a potential “super cycle”, according to analysts, as firms race to stockpile inventory. “There’s definitely scrambling going on,” said Fusion Worldwide president Tobey Gonnerman. “We’re seeing double and triple ordering just like in past shortages.”
Rising AI investment — with companies such as Amazon, Meta, Microsoft, and Alphabet expected to spend $400 billion on AI infrastructure this year — is further straining production capacity. Meanwhile, data center and smartphone upgrades are compounding the pressure, sending DRAM spot prices up nearly 200% year-on-year.
The surge has boosted profits for major memory producers: Samsung, SK Hynix, and Micron have all seen share prices soar by up to 170% this year. Yet the trend also threatens higher costs for electronics makers, who are already grappling with tariffs and rare earth supply issues. Companies like Raspberry Pi have raised prices after memory costs jumped 120% in a year.
Despite optimism, experts such as Dan Hutcheson of TechInsights caution that the “super cycle” may be short-lived, predicting a cooling period by 2027. Still, for now, the AI boom has turned the once-overlooked world of DRAM into one of tech’s hottest commodities.
