“Quite terrible”: Sam Altman says PCs waste the world’s limited DRAM supply

OpenAI CEO Sam Altman has argued that keeping large amounts of RAM in personal computers is an inefficient use of limited memory supplies, saying data centers can put the same resources to better use. He said local AI inference, which is becoming an increasingly popular alternative to cloud-based AI services, leaves memory unused most of the time, making it a poor allocation of scarce DRAM.

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. You want maximum utilization in the cloud if you’re trying to deal with this very supply-constrained environment, - Sam Altman.

In other words, Altman believes memory would be put to better use in centralized data centers, where it could continuously serve multiple users and workloads. However, his argument focuses on local AI inference and does not account for the many other reasons people need RAM in their PCs, including gaming, everyday applications and multitasking.

Altman's position also aligns with OpenAI's business model. Although the company offers some open-weight models, its main products are closed-source models that run in data centers and are accessed through subscriptions and cloud services. Wider adoption of local AI models could reduce users’ dependence on cloud platforms and subscriptions, although data centers would remain essential for training large models.

Remarks come as AI infrastructure providers compete for increasingly limited memory supplies. Memory manufacturers are prioritizing demand from data centers, putting additional pressure on the consumer market and contributing to higher prices for PCs and other hardware. His comments add to the debate over whether the world’s limited DRAM supply should be concentrated in large AI data centers or remain available to individual users.