OpenAI's Sam Altman says your PC's RAM is a waste of scarce DRAM that belongs in data centers

OpenAI chief executive Sam Altman has argued that the memory inside personal computers is a poor use of a scarce global resource, and that the world's DRAM supply would be better placed in data centers. The comments, reported by Tom's Hardware and amplified on Hacker News, wade directly into the debate over whether AI inference should happen locally on consumer hardware or centrally in the cloud.

What Altman said

Speaking with the YouTube channel Cheeky Pint, Altman described local inference — and, by extension, DRAM sitting in home computers — as "quite terrible." According to Tom's Hardware, his fuller point was about utilization: "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 5% of the time."

The argument is essentially utilitarian. In this framing, consumer machines tie up memory that sits idle for most of the day, while data centers can keep the same modules busy around the clock and share them across many users. Viewed that way, steering DRAM toward centralized facilities extracts more useful work per chip.

A hostile reception

The Hacker News discussion that followed was overwhelmingly critical. Many readers took the statement as self-interested — the CEO of a company whose business depends on centralized inference suggesting that consumers do not really need memory in their own machines.

One commenter inverted Altman's logic, arguing that by the same reasoning it would be equally terrible for a small number of people to hold most of the world's money while it sits "used 5% of the time." Another pushed back on the utilization claim itself, citing an unnamed report that supposedly found AI data centers idle over 70% of the time for lack of demand, with operators buying hardware largely to keep it away from competitors. That report was not identified in the thread, so the figure cannot be verified — but it highlights the central weakness in Altman's case: whether real-world AI facilities actually sustain the high utilization his argument assumes.

Others in the thread read the remarks as a sign that local inference is viewed as a competitive threat, with several commenters defending the right to keep as much memory as they like in their own machines, used or not.

Why it matters

Coming from the head of the company that defines mainstream AI access, this is an unusually candid statement of where the industry's incentives lie. If AI builders treat consumer DRAM as competition for a constrained supply, ordinary hardware buyers feel the consequences: heavy data center demand for memory can tighten the market and raise component prices, a dynamic commenters in the thread accused AI operators of already driving.

The remark also draws a sharp line in the local-versus-cloud inference debate. Advocates of on-device models point to privacy, offline capability and freedom from recurring fees; Altman's counterargument is that shared infrastructure is simply more efficient at a global scale. Which view prevails will shape what hardware consumers are sold — and whether the memory in a future PC is treated as yours to fill or as a misallocated resource.