OpenAI CEO argues DRAM belongs in data centers, not personal computers

OpenAI chief Sam Altman believes global memory chips should remain in data centers rather than sitting idle in personal computers.

Altman made the comments during an appearance on the Cheeky Pint podcast, where he addressed the ongoing shortage of RAM in the consumer electronics market. He called it “terrible” for individuals to hold limited supplies of DRAM in devices that run at low capacity. His view directly challenges the growing trend of running open-source AI models locally on home hardware.

The CEO explained that global allocation is more efficient when hardware operates at maximum capacity. He noted that personal computers typically use their memory resources only five percent of the time. In contrast, cloud infrastructure allows for continuous, high-volume processing of AI tasks without the wasted potential found in individual desktops or laptops.

This strategy aligns with OpenAI's aggressive purchasing of memory components. The company is securing its supply chain to support its massive computing needs. Altman’s comments suggest that the current push for local inference may not be the optimal solution for a constrained global market of DRAM.

Memory manufacturers Samsung and SK Hynix have already locked in long-term contracts with cloud service operators. These agreements reserve much of the world's RAM production for the next few years. Projects like OpenAI’s Stargate are central to these bulk purchases, which have driven up the cost of laptops, phones, and other consumer technology.