Sam Altman Thinks You're Wasting DRAM

Sam Altman argues that the global shortage of DRAM makes locally running AI models an inefficient use of scarce memory. In an interview, he said many people expect open-source models to let them run AI on personal computers, but that would leave large amounts of RAM idle much of the time. He favors concentrating memory in data centres, where cloud systems can keep it more fully utilized. His comments come as demand from AI data centres drives up memory prices, raises the cost of consumer devices, and contributes to Micron’s withdrawal from consumer memory production. The account also says OpenAI is permitted to buy as much as 40% of global DRAM output.

Altman describes silicon capacity as a shared constraint and says OpenAI is developing a custom AI processor, Jalapeno, designed with AI for AI. He suggests advances in AI could help create major new technologies or improve hardware efficiency. The central tension is that his preferred allocation directs limited computing resources toward corporate data centres rather than personal ownership and local computing. That prospect troubles consumers already facing higher PC and console prices, as well as the possibility that cloud services will replace devices people can own and control. The piece frames Altman’s efficiency argument as a challenge to individual access: if AI helps produce better hardware, ordinary users may still be excluded from possessing or improving it.