OpenAI CEO Sam Altman says it is “terrible” for scarce RAM to sit in personal PCs, arguing that memory should be allocated to data centers to maximize utilization for AI workloads. Speaking on the Cheeky Pint podcast, he discusses local inference and the idea that people could run open models on computers they already own or want to build.
Altman’s argument centers on supply constraints: he says having widespread RAM at home leads to low usage overall, compared with using that limited resource in the cloud where it can be utilized more intensively. The coverage links his remarks to OpenAI’s broader approach to deploying closed models in data centers, and some outlets frame the comment as defending that business model.
Across the reports, the consistent angle is Altman’s critique of consumer-side RAM for running AI. Differences are mainly in emphasis and framing: one outlet highlights the cost and scarcity context in sharper terms, while another adds more speculative discussion about what this could mean for future subscription pricing and perceptions of hardware ownership.
