Sam Altman says it is terrible your PC uses RAM while AI data centres run short of supply

Amid the ongoing global RAM shortage, OpenAI CEO Sam Altman has a rather unusual take. He believes that people keeping large amounts of memory in their personal computers, where it may sit unused most of the time, is not the best way to use the world's limited supply. Instead, he argues, that memory would be better used in cloud data centres, where it can serve many users.Altman made these comments during an appearance on the Cheeky Pint podcast, where he was asked about supply chain constraints and the growing demand for memory. Discussing the popularity of running AI models locally, rather than accessing them through cloud services, he questioned whether this approach makes sense when memory supplies are already tight.“A lot of people want local inference,” Altman said, referring to the process of running an AI model on a personal computer or other local device. He argued that open-source models have encouraged people to believe they can run AI on their own machines and reduce the need for large data centres.“But from a global allocation perspective, actually, it’s quite terrible to have everybody have a bunch of the world’s limited supply of DRAM sitting in their own computer and being used five percent of the time,” he said. Why Sam Altman wants more RAM in data centresAltman's argument is largely about utilisation. A personal computer may have 16GB, 32GB or even more RAM, but it does not necessarily use all of that memory throughout the day. In a cloud data centre, resources can be pooled and allocated to different users and workloads as demand changes.“You want maximum utilization in the cloud if you’re trying to deal with this very supply constrained environment,” Altman said. His comments come at a time when AI companies are expanding data centres and competing for memory chips. OpenAI itself has made large infrastructure commitments. In October 2025, it announced partnerships with Samsung and SK hynix to expand advanced memory production for its Stargate initiative. The companies said they were targeting production capacity of up to 900,000 DRAM wafer starts per month as part of the accelerated rollout. The pressure is already being felt by consumer electronics makers. TrendForce expects conventional DRAM contract prices to rise 10–15 per cent in the fourth quarter of 2026. IDC, meanwhile, says the memory shortage could push global smartphone shipments down 16.7 per cent this year, while average selling prices rise 27.6 per cent to $581 (roughly Rs 55,144).However, there is another side to the debate. Running AI locally can offer users more privacy, offline access and greater control over their data. And memory used in consumer PCs cannot simply be pulled out and transferred to an AI server. Different workloads require different types of memory, even as manufacturers shift production capacity towards higher-demand server and AI products.- Ends