Sam Altman Says Your PC Has Too Much Idle RAM: Running AI Models Locally Makes for Terrible Global Resource Allocation

Sam Altman was asked about the ongoing RAM shortage during his appearance on the Cheeky Pint podcast. His answer was blunt: putting RAM in personal PCs instead of data centers is "terrible".

Sam Altman talking about memory allocation on the podcast. Image: Justin Sullivan / Getty Images

For context, the RAM shortage has lasted for a while and prices are on the rise. The surging memory demand from AI data centers has tightened the entire global supply chain. At the same time, the narrative around local AI inference has been growing in popularity: open source models are getting more powerful, they can run on personal computers, deliver better privacy, lower latency, and eliminate the need to send data to the cloud. Altman clearly does not see it this way.

He said: "A lot of people want local inference. Somehow they think the problem is already solved, because there will be open source models that we can run on our own computers, so we don't need all that hardware in data centers. But from a global allocation perspective, in reality, letting everyone..." His original quote cuts off at this point.

We do not know the full content of the truncated segment. The part that was made public is clear enough: his priority is how global memory resources are distributed. For Altman, the RAM sitting inside personal computers is simply in the wrong place.

This opinion will upset a number of people. Local inference is far more than just a tech romantic fantasy, it has tangible practical benefits: it works without an internet connection, keeps data fully local, and no recurring cloud fees are required. But Altman's perspective is centered on resource utilization. RAM in data centers can be partitioned for use by many different users, while RAM on personal PCs just sits idle when the device is not in use. Of course, personal PCs also have idle compute power and unused memory. The core problem is: when the whole world is facing a global shortage, who gets priority access to memory resources.

The Verge's report does not expand on more details, and Altman did not propose any solutions. He only called the current situation "terrible". For developers, this comment acts as a reminder: the dream of widespread local inference is separated from the reality of memory allocation by strained supply chains.

You are free to disagree. After all, your computer is yours to control. But if RAM prices keep rising, this debate will no longer just be about technical routes, it will show up directly on your bills.