OpenAI CEO Sam Altman has drawn criticism after discussing the global allocation of computer memory and questioning whether it is efficient for individuals to keep large amounts of DRAM in their personal computers.
The comments came during a video interview with Stripe president John Collison on the Cheeky Pint YouTube channel. At around the 29-minute mark, Altman addressed the growing interest in local inference, referring to users running language models directly on their own hardware.
Altman said some people assume open-source models will solve the need for large data centers because they can run on personal computers. He argued that, from a global allocation perspective, it may be inefficient for everyone to hold substantial amounts of the world’s limited DRAM supply when that memory is used only a small percentage of the time.
The argument has been poorly received by critics, particularly amid higher memory prices. The source report says consumers are paying about five times more for memory than in early 2025, with rising component costs affecting computers, graphics cards, smartphones, game consoles and other products that use memory.
Altman’s position favors concentrating computing resources in large data centers, where AI services such as ChatGPT operate. Critics see that model as reinforcing the shift toward centralized, subscription-based software and hardware services, rather than allowing users to own and run applications locally.
The discussion also comes as the market continues to deal with pressure on memory availability. The report says that, without a major downturn in the AI industry, the current strain on RAM prices could continue for years.
Sources: PCWorld.