"The global RAM shortage is the result of AI data centers gobbling up high-bandwidth memory. "AI workloads are built around memory," Sanchit Vir Gogia, CEO of Greyhound Research, told NPR in late December. "AI has changed the nature of demand itself. Training and inference systems require large, persistent memory footprints, extreme bandwidth, and tight proximity to compute. You cannot dial this down without breaking performance.""
"Samsung says AI data center-fueled RAM scarcity could raise the company's prices. Wonjin Lee, Samsung's global marketing leader, sounded the alarm in an interview with Bloomberg on Tuesday at CES 2026. As recently as early December, Samsung told Reuters that it was monitoring the market but wouldn't comment on pricing. So, the change of tune can be seen as a deliberate signal to soften the ground ahead of an official announcement."
"It's been more than three years since ChatGPT launched and kicked off the AI craze. During that time, companies have hyped chatbots and other generative AI tools as a technology that will take us to the promised land, making life easier as machine learning automates our daily lives. It isn't yet clear if an AI bubble is set to burst, but some financial forecasters have sounded the alarm."
Samsung warned that semiconductor supply issues driven by AI data center demand for high-bandwidth memory are causing a global RAM shortage and could lead to higher product prices. Memory manufacturers have redirected production to service AI workloads, creating downstream shortages that affect even low-bandwidth RAM used in automobiles. AI training and inference require large, persistent memory footprints, extreme bandwidth, and close proximity to compute, making demand inflexible. Samsung signaled possible repricing as supplies tighten despite reluctance to pass costs to consumers. The shift accelerated after ChatGPT's debut, and financial forecasters caution of potential market corrections while immediate burdens fall on consumers and workers.
Read at Engadget
Unable to calculate read time
Collection
[
|
...
]