AI
“OpenAI CEO Sam Altman said it's a terrible use of the world's limited RAM supply to have it sitting in personal computers, where it might only be used 5% of the time.”
Plain restatementSam Altman publicly stated that keeping DRAM in personal computers, where he estimates it is used about 5% of the time, is a poor allocation of a globally limited memory supply, and that the memory would be better utilized in cloud data centers.
Sam Altman really did say this. It comes from episode 39 of Stripe's Cheeky Pint podcast, released on 29 October 2026, about 29 minutes in, and the wording appears in the official published transcript. He said it is "quite terrible" from a global allocation perspective for everybody to have some of the world's limited DRAM supply sitting in their own computer being used 5% of the time, and he argued that cloud data centers would keep scarce memory more fully used. The post's paraphrase is close, though it drops the word "quite" and leaves out that he was specifically arguing against the idea that running open-source AI models on home computers would reduce the need for data centers. The 5% figure was his own estimate offered in conversation, and he gave no data or method behind it. The "greedy home PC users" framing in the images is the post's own satire, not something he said. The reported backlash is real, and outlets also noted OpenAI's own very large memory supply agreements, though the often-cited figure of up to 40% of global DRAM output is a projection from 2025 reporting rather than a measured amount.
[drifted from the evidence:] OpenAI CEO Sam Altman [drifted from the evidence:] said it's a terrible use of the world's limited RAM supply to have it sitting in personal computers, where it [drifted from the evidence:] might only be used 5% of the time.
Sam Altman [added by the neutral restatement:] publicly stated that keeping DRAM in personal computers, where [added by the neutral restatement:] he estimates it [added by the neutral restatement:] is used [added by the neutral restatement:] about 5% of the time, [added by the neutral restatement:] is a poor allocation of a globally limited memory supply, and that the memory would be better utilized in cloud data centers.
Red-tinted words in the claim drifted from the evidence. Green-tinted words are what a neutral restatement needs.
The trace / claim to source
- Altman did make the statement. It appears in the official transcript of Cheeky Pint episode 39, published by Stripe on 2026-10-29, and is quoted consistently by many outlets with a matching ~29-minute timestamp.
- The substance of the paraphrase matches: he framed DRAM sitting in individual personal computers as a poor outcome from a global allocation standpoint.
- The "5% of the time" figure is his, said in that sentence about memory in personal computers.
- The caption's statement that he argues the memory would be better used in cloud data centers is supported. He said maximum utilization in the cloud is what a supply-constrained environment calls for.
- The caption's statement that the remarks sparked backlash online is supported by multiple outlets reporting widespread social media criticism.
- The caption's statement that critics point to AI companies as a cause of rising RAM prices reflects what outlets reported critics saying, and OpenAI's large reported memory commitments give that criticism a factual anchor.
- Altman said "quite terrible," and the claim renders it as "terrible." This is a small softening of a hedge rather than a reversal, and it does not change the thrust of the remark.
- The claim drops the conversational context. Altman made the remark while pushing back on the specific idea that open-source models running locally would remove the need for data centers, not in response to a question about gaming PCs, laptops, or ordinary consumer computing. The sentence he spoke is phrased generally, about "everybody," which is why the general paraphrase is defensible, but a reader of the caption alone would not know the subject under discussion was local AI inference.
- The post's images frame this as being about "greedy home PC users." That is satirical commentary added by the post, not language Altman used. No source shows him calling consumers greedy or calling for people to own no RAM.
- Whether the "5% of the time" figure is accurate. It was an estimate offered in conversation, and no measurement method, dataset, or definition of "used" was given with it. Outlets noted he provided no supporting data. The claim correctly attributes it to him, so this is uncertainty about the underlying fact, not about the attribution.
- How much consumer DRAM is genuinely idle, and how much could realistically be redirected to data center workloads, is not established by any source found.
- The reported figure that OpenAI's projected demand could reach 40% of global DRAM output is a projection attributed to Reuters and Bloomberg reporting from October 2025, not a verified figure for actual consumption in 2026.
- Reported magnitudes of recent memory price increases vary substantially between outlets, so no single price-increase number is relied on here.
The statement is real and comes from an identifiable primary artifact. Altman was interviewed by Stripe co-founder John Collison on episode 39 of Stripe's Cheeky Pint podcast, released October 29, 2026, titled "Why Sam Altman is against a single superintelligence," with the full transcript published on the episode page. The official transcript contains the decisive sentence: "quite terrible to have everybody have a bunch of the world's limited supply of DRAM sitting in their own computer". Multiple outlets that watched or read the episode quote the same sentence ending with the phrase "and being used 5% of the time", placing it at roughly the 29-minute mark. The surrounding context is a discussion of local AI inference. The exchange occurred about 29 minutes in, in a discussion of local inference, meaning AI models run on people's own computers, where Altman said "A lot of people want local inference." He was discussing the assumption that open-source models running on people's own computers would reduce the need for data centers, and said you want maximum utilization in the cloud when dealing with a supply-constrained environment, adding that many people's intuition runs the opposite way. On the caption's secondary points: multiple outlets reported that the comments drew widespread social media backlash, with critics arguing that AI companies such as OpenAI are driving the memory shortage in the first place. On OpenAI's own memory footprint, Tom's Hardware, citing Reuters, reported in October 2025 that Samsung and SK hynix reached preliminary agreements to supply memory for OpenAI's Stargate project, with projected demand climbing to 900,000 DRAM wafers per month, potentially as much as 40% of total global DRAM output. That is a projection in reporting, not a verified consumption figure.
Complete reasoning
The reply is formatted for pasting into the thread where the claim is circulating.