Sam Altman Says Idle PC RAM Is Wasted Amid Memory Shortage

OpenAI CEO Sam Altman has stirred debate over the global memory chip shortage, arguing that users keeping large amounts of RAM in personal computers, while barely using it most of the time, is an inefficient allocation of limited supply.

Speaking on the Cheeky Pint podcast, Altman discussed supply chain challenges and surging demand for memory chips as AI adoption accelerates. He noted that running AI models locally on personal computers is a growing trend, helped by the spread of open-source models. But in his view, fitting every device with large amounts of DRAM that sits mostly idle is not the best answer when supply is tight. He stressed the importance of getting the highest possible utilization out of memory in cloud data centers.

Shared resources versus fixed memory

Altman's argument rests on resource sharing. A data center can distribute memory and processing capacity among many users according to demand, rather than reserving fixed memory in each PC that may not be fully used during the day.

The remarks come as AI companies expand data center construction. In October 2025, OpenAI announced partnerships with Samsung and SK hynix to support advanced memory production under its Stargate initiative, targeting capacity of up to 900,000 DRAM wafer starts per month.

Prices already rising

The demand is already hitting consumer electronics. TrendForce expects contract prices for conventional DRAM to rise 10% to 15% in the fourth quarter of 2026. IDC estimates that the shortage could cut global smartphone shipments by 16.7% this year, while the average selling price climbs 27.6% to $581.

The case for local AI

Running AI locally still offers clear benefits: stronger privacy, offline use, greater user control over data and less reliance on cloud services. PC memory also cannot simply be moved to AI servers, because memory types and operating requirements differ between consumer devices and data centers.

Altman's comments open a wider discussion about how computing resources should be distributed globally: local AI that gives users more independence, or pooled resources in data centers serving millions. Memory has become one of the key battlegrounds in the race to develop AI.