Sam Altman: Personal-PC RAM Is a "Terrible" Use of Supply

Altman's remarks came during an appearance on the Cheeky Pint podcast, where he also commented on AI's impact on math. The discussion touched on a common assumption among enthusiasts: that open-source models running on local computers will reduce the need for data-center capacity. Altman pushed back on that intuition, framing the issue as one of global allocation rather than local capability. He argued that in a supply-constrained environment, spreading limited DRAM across many personal machines that use it only five percent of the time is worse than concentrating it in the cloud for maximum utilization.

The context he described is already visible in the market. Memory companies Samsung and SK Hynix have signed long-term deals with cloud service operators and projects like OpenAI's Stargate, reserving much of the world's RAM output for the next few years. Those commitments, and the billions being spent on chips, are why consumer devices like laptops and phones are getting more expensive now. Altman's comments suggest that tension between local and cloud computing may shape how the industry allocates scarce memory.