Altman says 'quite terrible' for your PCs to have RAM

In an interview, OpenAI CEO Sam Altman argued that people keeping large amounts of RAM in home computers is an inefficient use of a scarce global resource when that memory is idle much of the time. He said cloud data centers can pool memory and keep it more fully utilized, contrasting their constant workloads with a desktop’s DRAM, which he estimated may be used only about 5% of the time. His comments came during a discussion of running open-source AI models locally, an approach that could let users operate AI on computers they own rather than relying on remote services.

The argument aligns with OpenAI’s business model: its main products run in company-operated data centers, while capable home machines and local models could reduce some users’ need for paid AI subscriptions. The case for centralizing memory also depends on data-center expansion continuing and on their AI workloads being a better use of RAM than everyday computing, gaming, or occasional local inference. The remarks prompted concerns about consumers losing access to hardware and software ownership in favor of renting cloud-based AI and computing. They also drew criticism from readers who saw the efficiency argument as self-serving, particularly amid claims that data-center equipment is often underused and that companies may be buying hardware to limit competitors’ access.