OpenAI chief executive Sam Altman said in a Cheeky Pint interview that putting scarce DRAM in personal computers for local AI inference is a poor use of global memory supplies compared with data centers, Tom's Hardware reported¹.

Discussing local inference with open-source models, Altman said many people assume they will run AI on machines they already own. From a global allocation view, he said it is quite terrible to have limited DRAM sitting in home computers and used 5% of the time¹.

He said operators should seek maximum utilization in the cloud in a supply-constrained memory market, and that public intuition often runs the other way¹. The remarks were in an interview with John Collison's Cheeky Pint channel, Tom's Hardware said¹.

Tom's Hardware noted OpenAI sells cloud subscriptions and that local AI could reduce paid use of its models. The outlet said Altman's comments assume data-center buildouts continue and that always-on data-center RAM is more worthwhile than RAM used for ordinary PC work¹. Whether memory makers will keep prioritizing data-center DRAM over consumer modules is not addressed with evidence in the report¹.