The world has a memory problem, thanks to artificial intelligence.<\/p>\n
\nThe explosion in AI-related cloud computing and data centers has led to so much demand for certain types of memory chips that now there’s a shortage. The imbalance is expected to start affecting prices of all sorts of products powered by technology.<\/p>\n
\n“I keep telling everybody that if you want a device, you buy it now,” said Avril Wu, a senior research vice president at TrendForce, a Taiwan-based consultancy that tracks markets for computer components. “I myself bought an iPhone 17 already,”<\/p>\n
\nThe chips are known as RAM, or random access memory, and are crucial to making sure that things like smartphones, computers and game consoles run smoothly. Chips allow you to keep multiple tabs open in browsers, for instance, or watch videos without them being choppy.<\/p>\n
\nWu said TrendForce’s data indicates that demand for RAM chips exceeds supply by 10% \u2013 and it’s growing so fast that manufacturers are having to shell out a lot more to buy them each month.<\/p>\n
\nWu said this quarter alone, they’re paying 50% more than the previous quarter for the most common type of RAM, known as DRAM \u2013 dynamic random access memory. And if producers want the chips sooner, they’re paying two to three times more.<\/p>\n
\nWu expects DRAM prices to rise another 40% in the coming quarter, and she doesn’t expect the prices to go down in 2026.<\/p>\n
How AI is gobbling up memory<\/p>\n
\nAI data centers require huge amounts of memory to accompany their cutting-edge graphics processing unit (GPU) microprocessors that train and operate AI models.<\/p>\n
\n“AI workloads are built around memory,” said Sanchit Vir Gogia, CEO of the tech advisory firm Greyhound Research.<\/p>\n
\nWhat’s more, AI companies are spending billions of dollars constructing data centers at warp speed around the world. It’s the reason why Gogia says the demand for these chips isn’t just a cyclical blip.<\/p>\n
\n“AI has changed the nature of demand itself,” he said. “Training and inference systems require large, persistent memory footprints, extreme bandwidth, and tight proximity to compute. You cannot dial this down without breaking performance.”<\/p>\n
More chips for AI means fewer\u00a0 chips for other products<\/p>\n
\nIdaho-based Micron Technology is one of the world’s top makers of RAM and it’s benefited from this increase in demand. It reported better-than-expected quarterly earnings last week on the back of higher memory chip prices.<\/p>\n
\nCEO Sanjay Mehrotra said the company expected the market to remain strong, as the AI boom continues apace. “We believe that the aggregate industry supply will remain substantially short of the demand for the foreseeable future,” he said on a webcast after the earnings report.<\/p>\n