AI’s memory problem raises worrying CES buzz

The most consequential conversations at the Consumer Electronics Show (CES) this year were not about foldable screens, humanoid robots, or cars that talk back. They were about memory — or rather, the lack of it.

As artificial intelligence (AI) seeps into everything from laptops and smart rings to cars and home appliances, the semiconductor industry is running into a bottleneck that threatens to reshape how gadgets are designed, priced, and sold: a global shortage of memory chips.

Dynamic Random Access Memory, or DRAM — the component that allows devices to run apps, multitask, and perform real-time computations — has gone from a cheap, invisible commodity to one of the most contested resources in tech.

Supplies are tightening, prices are rising, and manufacturers are scrambling.

“Everybody is screaming for more supply… they just can’t find enough,” said Sangyeun Cho, who oversees Samsung Electronics’ semiconductor business in the United States. “And AI demand is still mounting.”

The culprit is AI itself. Memory manufacturers have redirected production toward data centers powering large language models and other AI workloads, where demand — and margins — are far higher.

That shift has left fewer chips available for consumer electronics, even as those same products now require more memory to support on-device AI features.

Supply chain woes

The result is a squeeze felt across the global supply chain. According to market tracker IDC, the memory crunch is unprecedented and could extend well into next year.

“What began as an AI infrastructure boom has now rippled outward, with tightening memory supply, inflating prices, and reshaping product and pricing strategies across both consumer and enterprise devices,” IDC analysts said.

“For consumers and enterprises alike, this signals the end of an era of cheap, abundant memory and storage.”

Consumers are already feeling it. Prices for computers, storage drives, and memory cards have risen, said Larry O’Connor, founder and chief executive officer of Other World Computing. Behind the scenes, device makers are making harder trade-offs.

Some are paying premiums to secure memory. Others are redesigning products to use less of it — or quietly scaling back features.

“Maybe the (robotic) dog you’ll get will sniff around and roll over, but it’s not going to bark a serenade because it doesn’t have enough memory,” said Michal Siwinski of Arteris, which designs chip connectivity systems.

The shortage is also forcing long-overdue discipline in software development. Engineers are writing leaner code, optimizing performance instead of relying on ever-cheaper hardware to mask inefficiencies.