Altman: Cloud efficiency beats local RAM storage

OpenAI CEO Sam Altman stated that it is inefficient to keep memory on personal computers when global supply is limited. In an interview on the Cheeky Pint podcast, the chief executive explained his stance on hardware allocation and the growing demand for AI infrastructure.

Altman described the idea of running local inference on individual devices as problematic from a global perspective. He noted that many people assume open-source models will solve the need for centralized hardware, but he argued the opposite. He called it "quite terrible" for the world’s limited supply of DRAM to sit in personal computers where it is used only five percent of the time.

The executive emphasized the need for maximum utilization in the cloud to manage the constrained environment. He observed that many people’s intuition runs contrary to this efficient allocation strategy, favoring local storage despite the scarcity of components.

This debate comes as OpenAI continues to secure hardware through major purchasing agreements. Memory manufacturers Samsung and SK Hynix have established long-term contracts with cloud operators and projects like OpenAI’s Stargate. These deals reserve a large portion of the world’s RAM output for the coming years.

The high cost of these chips is already affecting consumer prices. The billions spent by tech companies on memory components are contributing to the rising cost of laptops, phones, and other devices. Altman’s comments suggest he views centralized data center storage as the practical solution to this market pressure.