Altman Defends Home RAM as Memory Shortages Persist

Sam Altman said people should be able to buy plenty of RAM for personal computers, while also arguing that shared memory could make better use of limited supply. He linked the debate to rising DRAM prices and constraints affecting OpenAI’s data center buildout.

WTF Index NEUTRAL
◄ Terminator 1 Idiocracy 1 ►

This is a routine discussion of RAM allocation and AI infrastructure, with only a mild lean toward centralized shared computing.

Altman Defends Home RAM as Memory Shortages Persist

OpenAI CEO Sam Altman says people should be free to buy plenty of memory for their own computers. At a Semafor summit, he revisited remarks that had drawn criticism and explained how he sees personal RAM, data center demand and a limited supply fitting together.

Altman clarifies his position on personal computers

“I want to be clear, I want people to have a lot of RAM, I think everybody should be able to buy a lot of RAM for their own computer,” Altman said. He added, “I think the home tinkering is great.”

The comments came after earlier remarks were interpreted as favoring data centers over memory in personal computers. On the Cheeky Pint podcast, Altman had called it “quite terrible” for everyone to have RAM in personal computers rather than in data centers. He later said on X that he wanted people to have “so much ram” and that his earlier comments had been “taken out of context.”

His latest explanation makes room for both personal use and shared infrastructure. Altman acknowledged that people may want to run AI at home, while arguing that shared memory could be used more continuously during a shortage.

Why shared memory entered the discussion

Altman said the point he stood by was about making the best use of scarce supply. If memory is in short supply, he suggested it may be more efficient to share it and keep it working “100 percent of the time” than to have some of it sitting in an edge device at certain times.

He described this as an idea about optimal use, not necessarily a decision OpenAI would make. He also recognized tradeoffs, including the value of people having their own at-home inference. The question, as he framed it, is how to allocate limited memory while preserving choices about where AI runs.

That distinction matters to the debate: supporting home computer upgrades does not erase the pressure on memory supply, and advocating shared use does not mean personal computing has no value. Altman said he wanted the world to have much more memory overall.

Supply constraints reach beyond RAM

Altman connected the discussion to broader bottlenecks facing OpenAI. He said the company had been constrained by different things at different times, including memory, rack assembly, data center capacity and “some random component.” That list suggests the challenge is not limited to acquiring one type of hardware.

He said that a few years ago, the company began arguing that the world needed to build more of the memory supply chain in particular. In his account, the world did not expand it enough, and DRAM prices have since climbed sharply.

Altman estimated that DRAM prices may have risen “5x, 7x,” while acknowledging he did not know the exact figure. The source article also describes memory buying by AI companies as contributing to price hikes affecting products from phones to game consoles.

A shortage with competing priorities

The comments leave two priorities in view: expanding memory supply and deciding how to use what is available. More supply could ease the squeeze Altman described; shared use is one possible response while shortages remain. Neither idea removes the tradeoff between centralized data center resources and memory available in personal devices.

For consumers, the debate connects AI infrastructure spending with the cost and availability of everyday electronics. For companies building AI services, memory is one part of a wider set of data center constraints. Altman’s clarification was that he supports people buying RAM for home computers even as he sees a case for sharing scarce resources.

He also commented on Jensen Huang’s resignation as Nvidia’s CEO, saying he “doesn’t know enough” to say if Huang should be reinstated.